The association between indoor air pollution from solid fuels and cognitive impairment: a systematic review and meta-analysis

Search strategy

The meta-analysis followed the PRISMA guidelines [18] and involved an extensive search of the PubMed, Web of Science, EMBASE, and Cochrane Library databases from their inception to December 2023. A sensitive search strategy was developed using keywords related to household air pollution, solid fuel, dementia, cognition disorders, Alzheimer’s disease, and cognitive impairment. The specific search strategy used in four databases can be found in Supplementary Material Table S1. In the meta-analysis, only studies conducted on humans were included, without any language restrictions. Two researchers screen and include eligible studies separately, and if a disagreement is encountered, a joint decision is discussed with a third researcher, ensuring the consistency of the results. The study protocol has been registered in the PROSPERO sites (http://www.crd.york.ac.uk/PROSPERO), CRD42022385486.

Inclusion criteria and exclusion criteria

Studies that explored the association between IAP through solid fuel use and cognitive function were included. These studies measured both the fuel type for cooking and heating and the cognitive function status in the same individuals. The study methods covered cross-sectional, longitudinal cohort, and case-control, focusing on participants aged 45 years and older.

We excluded duplicated research that provided additional data on already included studies. Furthermore, the following studies were excluded: (1) those not relevant to indoor air pollution or cognitive dysfunction; (2) non-original papers such as reviews or editorials; (3) studies that could not provide specific values required for the analysis.

Data extraction

Two researchers separately collected the publications’ important details with a standardized form. The data gathered comprised the name of the first author, the publishing year, participant features (country, age, gender), sample size, cognitive assessment method, type of fuel used for cooking/heating, study design, major results (including effect estimates such as β, hazards ratio (HR), or odds ratio (OR), and their corresponding confidence intervals (CI)), and any other pertinent information. In cases where the included article did not directly provide β, HR, OR, or 95 % CIs, we calculated these values using the original data.

Assessment of study quality and risk of bias

The selected studies’ quality and risk of bias (ROB) were assessed separately by two researchers, and any discrepancies were settled through communication with a third author. The Newcastle-Ottawa Scale (NOS) was utilized to analyze the quality of cohort and case-control studies [19]. It assigns stars based on three main criteria: study group selection (up to 4 points), group comparability (up to 2 points), and outcome ascertainment (up to 3 points), for a total of 8 items. Except for the item of comparability, which obtained a maximum of 2 points, the other items obtained a maximum of 1 point. Every research project received a NOS score (0–9), with a score of 7 or higher denoting “high quality,” while a score below 7 denoted “low quality.” For the assessment of cross-sectional studies, we employed the scale suggested by the Joanna Briggs Institute (JBI) (http://www.joannabriggs.edu.au/) [20]. The scale consists of 10 items evaluating the quality of literature based on research objectives, rationale, participant selection, inclusion criteria, sample characteristics, tool reliability and validity, data authenticity, ethical considerations, statistical methods, and the significance of results and research value. A score of 0 is given for non-compliance with requirements, 1 point for mentioning but not providing detailed descriptions, and 2 points for comprehensive and accurate detailing. Each study was assigned a JBI score (0–20), with a score exceeding 16 considered “high quality”. The studies included in our analysis were exclusively cross-sectional or cohort studies, and we adopted the JBI and NOS scales to assess their quality, ensuring good applicability of these assessment tools. The Office of Health Assessment and Translation (OHAT) Risk of Bias Rating Tool for Human and Animal Studies was utilized to evaluate the ROB in cohort and cross-sectional studies [21]. Key criteria include exposure and outcome assessment, as well as confounding bias. Additional criteria encompass selection bias, attrition/exclusion bias, selective reporting bias, conflict of interest, and other sources of bias.

Assessment of publication bias

Publication bias was evaluated by examining funnel plot asymmetry [22]. Asymmetrical data distribution may point to the existence of minor study effects, which could be a sign of publication bias. Additionally, we conducted Egger’s test, a statistical test that utilized weighted linear regression analysis to evaluate funnel plot asymmetry. Significant asymmetry was determined to be present when the p-value was below 0.1 [23].

Assessment of heterogeneity

Clinical and methodological evaluations were conducted to determine the suitability of combining the data from the included studies, such as population characteristics, exposure variables, statistical analysis methods, and outcome measures. Visual inspection of forest plots and statistical methods were employed to assess statistical heterogeneity. Homogeneity was analyzed by the I² statistic and Cochran’s Q-test [24]. A value exceeding 50 % was considered indicative of high heterogeneity [25].

Statistical analysis

For the analysis of heterogeneity, if I²≥50 %, it was considered as “high” heterogeneity and the effect estimates were combined using the random effects model. In contrast, the fixed effects model was used [26]. The effect sizes for continuous variables were estimated using β coefficients and 95 % confidence intervals (CIs), while for binary or categorical variables, HRs or ORs and their 95 % CIs were applied. The quantitative measurements of solid fuel exposures were not studied, as most of the studies used self-reported questionnaires rather than actual exposure doses. Analyses of subgroups and meta-regression were carried out to investigate possible causes of heterogeneity. Subgroup analyses were used based on publication year (before 2021 or after 2021), geographic area (China or India), and sample size (>10,000 or ≤10,000), and cognitive assessment methods (total cognitive score: 30–31 points, 21–22 points or 43 points). Sensitivity analyses were conducted by systematically excluding every individual research to evaluate the stability and reliability of the pooled results. Stata version 15 was used to carry out each statistical analysis (Stata Corp, College Station, Texas).

The features of literature retrieval and study

Through the search formula, we obtained a total of 3,740 hits (Figure 1). A total of 201 articles were determined to be appropriate for this analysis after duplicates and other publications were disqualified using the criteria for screening by abstract. By reading through the full text, 13 papers were included in the research. One of the studies involved 3 separate cohorts [27]. These studies included 277,001 participants from 3 countries, which are China, India, and Mexico. The two study types that are most frequently utilized are cross-sectional and cohort. Fuel types include biomass, kerosene, wood, coal, etc. According to the study quality assessment, for cross-sectional studies, the scores ranged from 12 to 19 points. For cohort studies, scores ranged from 7 to 9 points. Totally, 12 studies were judged to be of “high quality”, and another study was of low quality. Table 1 displayed the specifics of the included studies. According to the ROB assessment, 13 studies were defined as “high quality” and 3 studies were defined as “low quality”. See Table S2 for more details.

Figure 1:

Flow chart of participant selection for this meta-analysis.

Association between solid fuel use for cooking and total cognitive score/episodic memory/cognitive impairment/cognitive decline

For total cognitive score, 11 studies were selected, totaling 318,329 participants, with coefficients ranging from −2.070 to −0.321 (more details see Table S3). In the study by Chen et al. [17], Cao et al. [16] and Du et al. [32], cognitive assessment consisted of the following three areas: orientation and attention, episodic memory, and visuo-construction, which were performed using TICS10, immediate and delayed word recall, and figure drawing, respectively, with a total cognitive score of 22 or 31 points. In the study by Luo et al. [30] and Wang et al. [37], the assessment of cognitive functioning consisted of two components: mental state and memory, and was accomplished through the TICS/MMSE, figure drawing, and immediate and delayed word recall tasks with a total score of 21 points. Deng et al. [28] and Tseng et al. [36] assessed cognitive functioning using the MMSE and MoCA, respectively, with a total score of 30 points. Dakua et al. [31], Jana et al. [33], Jin et al. [34], and Krishnamoorthy et al. [35] assessed cognitive functioning using five components: memory, orientation, arithmetic function, executive function, and object naming, with a total score of 43 points. Given the substantial statistical heterogeneity observed in these studies (I²=92.8 %, p＜0.0001), a random effects model was employed. A negative association was found between solid fuel use for cooking and total cognitive score, with the β=−0.73 (95 % CI: −0.90 to −0.55). The forest plot for the results was displayed in Figure 2A.

Figure 2:

Meta-analysis of the association between solid fuel use for cooking and (A) total cognitive score, (B) episodic memory score. Figure legends: Each diamond represents the beta coefficient (β) from an individual study, the diamond at the bottom represents the pooled effect size, and horizontal lines indicate the 95 % confidence intervals (95 % CI). A positive beta coefficient indicates those who used solid fuel for cooking had higher total cognitive scores and episodic memory scores, while a negative beta coefficient indicates those who used solid fuel for cooking had lower total cognitive scores and episodic memory scores.

For episodic memory, 4 studies (5 reports) were selected, totaling 63,064 participants, with coefficients ranging from −0.38 to −0.18 (more details see Table S4). Immediate and delayed word recall were used to evaluate the episodic memory, with a total score of 10 or 20 points. The heterogeneity test revealed a high level of heterogeneity (I²=56.5 %). Hence, a random-effects model was utilized to aggregate the results. The outcomes revealed that individuals using solid cooking fuels was associated with episodic memory scores (β=−0.23, 95 % CI: −0.30 to −0.17, Figure 2B) [16], 17], 30], 37].

For cognitive impairment, 2 studies (3 reports) were selected, totaling 12,451 participants, with hazard ratios (HRs) ranging from 1.19 to 1.82 (more detail see Table S5). Cognitive impairment was defined when the MMSE was <24 or 18 points in Deng and Du’s studies [28], 29], respectively. The overall pooled effect revealed that the use of household solid fuel for cooking was strongly associated to an elevated risk of cognitive impairment (HR=1.31, 95 % CI: 1.09 to 1.57) (Figure 3A).

Figure 3:

Meta-analysis of the association between solid fuel use for cooking and (A) cognitive impairment and (B) cognitive decline. Figure legends: The diamond represents the individual study hazard ratio (HR), and the diamond at the bottom signifies the pooled HR with 95 % confidence intervals (95 % CI). An HR greater than 1 indicates an elevated risk of cognitive impairment and cognitive decline in those who used solid fuel for cooking, while an HR less than 1 indicates a decreased risk of cognitive impairment and cognitive decline in those who used solid fuel for cooking.

For cognitive decline, 2 studies (4 reports) were selected, totaling 19,805 participants, with hazard ratios (HRs) ranging from 1.18 to 1.30 (more detail see Table S5). In both studies, a drop of three or more points at the follow-up was defined as cognitive decline. The results of meta-analysis suggested that the use of solid fuel for cooking was associated to a higher risk of cognitive decline (HR=1.24, 95 % CI: 1.18 to 1.30, Figure 3B) [27], 28].

Associations between trajectories of cooking fuel types and cognitive decline

One study (3 separate cohorts) was selected, with 15,644 participants (more details see Table S6). A drop of three or more points at the follow-up was defined as cognitive decline. Relative to those who continuously used solid fuel, the sustained use of clean fuel was associated with a lower risk of cognitive decline (OR=0.55, 95 % CI: 0.42–0.73, Figure 4A) [27]. Similarly, relative to those who continuously used solid fuel, switching from solid fuel to clean fuel was associated with a lower risk of cognitive decline (OR=0.81, 95 % CI: 0.71–0.93, Figure 4B) [27].

Figure 4:

Meta-analysis of the association between trajectories of cooking fuel types and cognitive decline, (A) sustained use of clean fuel and (B) switch from solid fuel to clean fuel. Figure legends: The diamond represents the individual study odd ratio (OR), and the diamond at the bottom signifies the pooled OR with 95 % confidence intervals (95 % CI). An OR greater than 1 indicates that, compared to continuous solid fuel use for cooking, sustained use of clean fuel and switching from solid fuel to clean fuel are associated with a higher risk of cognitive decline. Conversely, an OR less than 1 indicates compared to continuous solid fuel use for cooking, sustained use of clean fuel and switching from solid fuel to clean fuel are associated with a lower risk of cognitive decline.

Association between solid fuel use for heating and total cognitive score and episodic memory

For total cognitive score, 3 studies (4 reports) were selected, totaling 18,597 participants, with coefficients ranging from −0.75 to −0.32 (more details see Table S7). The association between solid fuel use for heating and total cognitive score was β=−0.43 (95 % CI: −0.59 to −0.26, and I²=0.0 %) (Figure 5A) [16], 17], 36].

Figure 5:

Meta-analysis of the association between solid fuel use for heating and (A) total cognitive score and (B) episodic memory score. Figure Legends: Each diamond represents the beta coefficient (β) from an individual study, the diamond at the bottom represents the pooled effect size, and horizontal lines indicate the 95 % confidence intervals (95 % CI). A positive beta coefficient indicates those who used solid fuel for heating had higher total cognitive scores and episodic memory scores, while a negative beta coefficient indicates those who used solid fuel for heating had lower total cognitive scores and episodic memory scores.

For episodic memory, 2 studies (3 reports) were selected, totaling 18,196 participants, with coefficients ranging from −0.38 to −0.18 (more details see Table S7). Episodic memory was assessed by immediate and delayed word recall, with a total score of 10 or 20 points. The association between solid fuel use for heating and total cognitive score was β=−0.22 (95 % CI: −0.34 to −0.10, and I²=0.0 %) (Figure 5B) [16], 17].

Subgroup analysis and meta-regression analysis

To explore possible causes for this heterogeneity, we conducted additional meta-regression analyses and subgroup analyses. The meta-regression results suggest that region (p=0.043) and sample size (p=0.037) may be sources of heterogeneity, independent of publication year (p=0.276) and cognitive assessment methods (p=0.086). However, the results of subgroup analysis illustrated that the publication year, area, sample size and cognitive assessment methods were not sources of heterogeneity (Figure 6).

Figure 6:

Subgroup analyses of the relationship between solid fuel use for cooking and total cognitive score. (A) Publication year, (B) area, (C) sample size, (D) cognitive assessment method.

Sensitivity analysis and publication bias

In sensitivity analysis, the overall results remained robust, indicating the stability of the findings. Removing any individual study did not significantly impact the strength and significance of the main findings (Figure 7A). The funnel plots showed that there was no publication bias for studies on cooking with solid fuel (Figure 7B). Additional quantitative analysis using Egger’s tests revealed p-values of 0.565 (Figure 7C), indicating no publication bias.

Figure 7:

Sensitivity analysis (A), funnel plot (B) and Egger’s test (C) of included studies.

Important results

This research included 13 studies from three countries with a total of 277,001 participants, in which we thoroughly analyzed the information that was already available on solid fuel use and cognitive function. Nearly all studies found an association between cognitive function and solid fuel usage for cooking and heating [16], 17], 36]. The results of the meta-analysis revealed important correlations between total cognitive score and episodic memory score with solid fuel use for cooking and heating. Cooking with solid fuel is associated with a higher risk of cognitive impairment and cognitive decline. However, the high heterogeneity across studies in these meta-analyses significantly undermines the validity of the gathered evidence.

Comparison to previous meta-analyses

Most of the meta-analyses [38], [39], [40], [41] examining the relationship between air pollution and cognitive impairments primarily focus on pollutants such as particulate matter ≤2.5 µ (PM2.5), PM10, carbon monoxide, and nitrous oxides, and their association with the risk or incidence of dementia (2–5). These studies have found that PM2.5, nitrogen dioxide, and nitrogen oxide increase the risk of dementia. However, based on our limited knowledge, only one meta-analysis [42] has summarized five studies published before April 2021, specifically investigating the relationship between using solid fuel indoors and cognitive dysfunction. The findings indicated a strong association between exposure to IAP from cooking and heating fuels and cognitive dysfunction in elderly women. Nevertheless, this study solely reported mean cognitive scores and did not encompass other dimensions of cognitive impairment, such as episodic memory. There is a lack of dynamic assessment of changes in cognitive function following the use of solid fuels. Moreover, the selected papers, methodologies, and findings vary significantly, resulting in substantial heterogeneity (I²=100 %).

In comparison, we conducted an updated literature search until December 2023, which led to the inclusion of more research and participants in our analysis (Table 1). Additionally, we performed a systematic analysis that encompassed other dimensions of cognitive assessment, such as episodic memory, and the longitudinal associations between solid fuel use and changes in cognitive function. As a result, our results provide a more comprehensive understanding of the topic. However, the meta-analyses revealed significant heterogeneity, which greatly diminishes the credibility of the overall results.

Our efforts to examine possible origins of heterogeneity by incorporating variables such as the publication year, area, sample size, and cognitive assessment methods in subgroup analyses and meta-regression analyses, the meta-regression results suggest that region (p=0.043) and sample size (p=0.037) may be sources of heterogeneity. However, subgroup analyses did not reveal positive results. This suggests that other undisclosed or unrecorded variables, such as genetics, nutrition status, and physical activity by countries, may play a more significant role in contributing to the heterogeneity [43]. Therefore, it is important to deal with the findings cautiously when investigating the effect of solid fuel use for cooking on the total cognitive score in previous studies.

Interpretation of findings

While the causal relationship between air pollution and cognitive impairment is not entirely accepted, air pollution is an essential cause of cognitive impairment, according to a growing amount of research. Whether it is cooking or heating, the use of solid fuels in households increases the concentration of pollutants in the air, including PM2.5, PM10, and carbon monoxide [44]. These air pollutants may act on the body through a series of reactions, leading to cognitive impairment. The following points provide possible explanations for the association between the use of solid fuels and cognitive impairment.

Firstly, the pollutants released by solid fuel combustion may directly enter the brain through active transport, the blood-brain barrier, or olfactory sensory neurons [45], leading to systemic inflammation, oxidative stress, and neuroinflammation. Systemic inflammation can result in cellular damage and alterations in the balance of reactive oxygen species/cytokines in the brain, and may potentially change the biological composition of the brain’s innate immune cells [46]. Moreover, systemic inflammation might facilitate the deterioration of the olfactory, respiratory, and blood-brain barrier, thereby increasing the chances of particles entering the central nervous system and creating a vicious cycle [47]. Individuals who experience inflammation and oxidative stress changes caused by air pollutants have a higher potential for cognitive impairments [48]. Secondly, airborne particle exposure is believed to be associated to reduced resting cerebral blood flow velocity and increased resting cerebral vascular resistance in older adults [49]. Abnormal cerebral hemodynamics are independently related to cognitive function decline [50]. Thirdly, exposure to air pollution has been associated to smaller brain volumes and β-amyloid accumulation, considered markers of brain atrophy and neuronal dysfunction, respectively [51], 52].

Strengths and limitations

The thorough and methodical evaluation of the most recent epidemiological data showing the link between IAP from solid fuel use for cooking and heating and cognitive impairment is a key strength of our review. We went beyond individual studies or isolated reviews by covering other aspects of cognitive function assessment, such as episodic memory. The dynamic assessment of changes in cognitive function by solid fuel use further enhances the overall understanding of the topic. Furthermore, we rated the credibility of the combined evidence after carefully assessing each study’s quality and bias risk. Researchers and decision-makers can benefit from this assessment since it can be used to pinpoint the shortcomings and gaps in the current body of knowledge. Moreover, our inclusion of many nationally representative cohorts such as the China Health and Retirement Longitudinal Study (CHARLS), the Longitudinal Aging Study in India (LASI), the China Family Panel Studies, and the Mexican Health and Aging Study (MHAS) contributes to the strength and generalizability of our conclusions. These large-scale, representative populations enhance the external validity of our findings.

However, it is important to acknowledge some restrictions. Firstly, most studies didn’t assess exposure quantitatively, which is crucial for determining the extent to which exposures contribute to cognitive impairment or decline and for establishing significant health harm. Secondly, although our analysis included some longitudinal data, the majority of the studies had a cross-sectional design. Longitudinal studies have the potential to provide more insights into cognitive deterioration over time and offer a better understanding of disease outcomes. Thirdly, while analyzing subgroups to find probable reasons for heterogeneity, the limited availability of data prevented us from capturing all key factors that may contribute to the observed heterogeneity, such as the type of solid fuel (most researches are fixed fuel types). Lastly, the current assessment of exposure to solid fuel use is rudimentary and lacks precise quantification. However, rigorous quantification of exposures in the next investigations would allow for a more accurate and reliable assessment.

Conclusions and future outlook

In conclusion, our analysis provides evidence of an association between exposure to IAP from solid fuel usage and an increased risk of cognitive impairment. Due to the exacerbated adverse effects of indoor air pollution (IAP) on older adults [53], our results emphasize the significance of comprehensive air control policies to protect the health of older individuals. For instance, transitioning to cleaner fuel from solid sources could be considered. Furthermore, we recommend that future research focus on investigating specific types of solid fuels and exposure levels, and emphasize the need for large-scale prospective cohort studies to further explore the correlation between air pollution and cognitive function. With the global aging population, taking effective measures to prevent age-related cognitive decline holds significant public health implications.