Household polluting cooking fuels and adverse birth outcomes: An updated systematic review and meta-analysis

Background and aim The current study aimed to clarify the association between household polluting cooking fuels and adverse birth outcomes using previously published articles. Methods In this systematic review and meta-analysis, a systematic literature search in PubMed, Embase, Web of Science, and Scopus databases were undertaken for relevant studies that had been published from inception to 16 January 2023. We calculated the overall odds ratio (OR) and 95% confidence interval (CI) for adverse birth outcomes [low birth weight (LBW), small for gestational age (SGA), stillbirth, and preterm birth (PTB)] associated with polluting cooking fuels (biomass, coal, and kerosene). Subgroup analysis and meta-regression were also conducted. Results We included 16 cross-sectional, five case–control, and 11 cohort studies in the review. Polluting cooking fuels were found to be associated with LBW (OR: 1.37, 95% CI: 1.24, 1.52), SGA (OR: 1.48, 95% CI: 1.13, 1.94), stillbirth (OR: 1.38, 95% CI: 1.23, 1.55), and PTB (OR: 1.27, 95% CI: 1.19, 1.36). The results of most of the subgroup analyses were consistent with the main results. In the meta-regression of LBW, study design (cohort study: P < 0.01; cross-sectional study: P < 0.01) and sample size (≥ 1000: P < 0.01) were the covariates associated with heterogeneity. Cooking fuel types (mixed fuel: P < 0.05) were the potentially heterogeneous source in the SGA analysis. Conclusion The use of household polluting cooking fuels could be associated with LBW, SGA, stillbirth, and PTB. The limited literature, observational study design, exposure and outcome assessment, and residual confounding suggest that further strong epidemiological evidence with improved and standardized data was required to assess health risks from particular fuels and technologies utilized.


Introduction
Household air pollution (HAP) is often considered to be a major public health problem in low-and middle-income countries (LMICs) (1). HAP has a direct impact on human health and is an important risk factor for increased morbidity and mortality (2). The inefficient combustion of solid fuels (wood, coal, charcoal, dung, and crop waste) and kerosene in simple stoves and devices is a major source of HAP (3-5). Approximately 2.4 billion people cook mainly with polluting fuels (solid fuels and kerosene) globally (6). The inefficient combustion of polluting fuels often emits a high level of air pollutants (7). It is estimated that nearly 3 million people die from HAP exposure every year, the vast majority of whom live in LMICs (1). Research has shown that exposure to HAP is associated with chronic obstructive pulmonary disease, lung cancer, acute respiratory infections, cerebrovascular disease, ischemic heart disease, and adverse birth outcomes (8)(9)(10)(11). The mechanism involved in the cardiorespiratory effect includes inflammation and oxidative stress induced by reactive oxygen and nitrogen species generated by inhaled pollutants (12,13). HAP has also been associated with epigenetic adverse effects, which change DNA expression and potentiate the inflammatory effects of pollutants (14).
Perinatal morbidity and mortality are majorly associated with adverse birth outcomes, such as low birth weight (LBW), small for gestational age (SGA), stillbirth, and preterm birth (PTB) (15). For example, 21.9% of neonatal deaths were attributable to being born SGA (16). The leading causes of death in children under 5 years old were PTB complications (17). The risk factors of adverse birth outcomes include both individual (e.g., smoking, diet, antenatal depression, and antenatal care) and environmental factors (e.g., air pollution, occupational exposure, and pesticides) (18)(19)(20)(21)(22). For example, evidence suggests that the combustion of polluting cooking fuels emits high levels of air pollutants, such as fine particulate matter (PM 2.5 ), carbon monoxide (CO), and polycyclic aromatic hydrocarbons (PAHs), which can influence fetal development (23,24).
A growing number of epidemiological studies focus on the association between HAP and adverse birth outcomes. Two metaanalyses have pooled the effects of solid fuel use and adverse birth outcomes (25, 26). Both mainly focused on LBW and stillbirth, and only one study conducted a subgroup analysis for LBW (26). In addition, the study did not address the question of whether the strength of any association between polluting fuel use and the risk of adverse birth outcomes was affected by different types of polluting fuel (26). Since then, many new studies have been conducted, focusing on a broader range of fuel types and adverse birth outcomes, and the conclusion is still inconsistent. One cohort study (27), two case-control studies (28,29), and six cross-sectional studies (30-35) suggested that polluting fuel use was related to increased risk of adverse birth outcomes, while other studies (23, 24, [36][37][38][39][40][41][42][43][44][45][46][47] failed to reveal a significant correlation between them. A recent review of the subject qualitatively assessed the impact of unclean cooking fuels on adverse birth outcomes, but the review was short of quantitative combined research data (48). The HAP field is evolving rapidly. Confirming or disputing earlier findings is vital in this circumstance. Furthermore, a timely review of research methods and results can assist or guide further research as well as public health policy. Based on these considerations, we carried out a systematic review and meta-analysis to determine whether HAP was associated with LBW, stillbirth, PTB, and SGA. We also explored whether the strength of any association between polluting fuel use and the risk of adverse birth outcomes was affected by different types of polluting fuel. Subgroup and meta-regression analyses were conducted to explore the sources of heterogeneity.

Methods
The current systematic review and meta-analysis scrutinized the published adverse birth outcomes data related to polluting cooking fuel use from previous research publications using previously published search. We used the standard method of the preferred reporting items for systematic review and meta-analysis (PRISMA) protocols 2009 for conducting this review study (49). The study protocol was registered with PROSPERO (CRD42021269660).

Search strategy
We searched the PubMed, Embase, Web of Science, and Scopus databases for articles reporting HAP and adverse birth outcomes published up to 16 January 2023. In the present review, HAP is defined as chemical, biological, and physical contamination of house air and derived from the use of polluting fuel (wood, dung, crop residues, charcoal, coal, and kerosene) for cooking (31). An infant weighing <2,500 g was defined as LBW (23), and SGA referred to a baby weighing less than the 10th percentile for a certain gestational age (35). Stillbirth was defined as delivering a baby without any sign of life after 20 weeks of gestation (27, 50), while PTB was defined as an infant born before 37 weeks (24). The search process has been focused on the following terms: (indoor air pollution or household air pollution or cooking fuel or unclean fuel or solid fuel or biomass or wood or coal or kerosene or cooking) and (pregnancy outcome or pregnancy complications or low birth weight or premature birth or stillbirth or small for gestational age). The detailed search strategies are presented in Supplementary Table 1. We also scanned the reference lists of retrieved articles and previous meta-analyses to identify additional studies.

Selection criteria
All studies obtained from the aforementioned resources were independently evaluated by two reviewers for inclusion and exclusion. First, the title and abstract of each study were reviewed, and the full text of the relevant studies was retrieved and assessed for inclusion eligibility. Studies were considered for inclusion if they were (1) original studies, (2) conducted in the human population, and (3) quantified the association between HAP exposure during pregnancy and adverse birth outcomes (LBW, SGA, stillbirth, and PTB). Studies were excluded for the following reasons: (1) the outcome was due to other factors (e.g., maternal age, educational level, and the house renovation) but not polluting cooking fuels (e.g., wood, dung, crop residues, charcoal, coal, and kerosene); (2) studies reported other outcomes those were not of interest; (3) the subjects from the control group were not exposed to clean fuels (e.g., electricity, liquid petroleum gas, natural gas, and biogas) or were exposed to polluting fuels; (4) studies did not report relative risk (RR) or odds ratio (OR); (5) conference abstract, letter, or protocol; (6) not published in English. Based on the inclusion and exclusion criteria, we excluded several unqualified studies from previous meta-analyses. The reasons for exclusion are as follows: reported the impact of other factors (e.g., cooking smoke and chimney stove) on adverse birth outcomes (n = 3) (51-53); the outcomes were not of interest (n = 2) (54, 55); did not report RR or OR (n = 3) (56-58); published in non-English (n = 2) (59, 60).

Data extraction and quality assessment
Data from all included studies were independently extracted by two reviewers and cross-checked to avoid errors. The following information was extracted from the publications: author, publication year, location of the study, study population, study design, study period, exposure, outcome, covariates adjustment, the frequency distribution of exposure and outcome, comparator, exposure, and outcome assessment method. We also extracted risk estimates of association relating polluting cooking fuel to adverse birth outcomes for pooled analysis. RR or OR with precision [95% confidence interval (CI)] was extracted from included studies. The fully adjusted effect estimates were used for analysis when both unadjusted and adjusted estimates were provided.
The methodological quality of cohort and case-control studies was assessed by the Newcastle-Ottawa scale (range, 0-9 stars) (61). Each study was awarded stars based on three dimensions [selection, comparability, and outcome (cohort studies) or exposure (casecontrol studies)] (61). A study awarded seven or more stars was considered high quality (62), whereas one awarded three or fewer stars was considered low quality (63). The 11-item checklist was used for methodological quality assessment of crosssectional studies, which had been recommended by the Agency for Healthcare Research and Quality (range, 0-11 scores) (64). A study with an eight or higher score was deemed to be of high quality and that with a three or lower score was of low quality (64).

Statistical analysis
We assumed that RR was approximately equivalent to OR for our rare adverse birth outcomes (65, 66). To examine the association between polluting cooking fuels and adverse birth outcomes, we calculated ORs and corresponding 95% CIs. The I 2 statistic was defined to assess the heterogeneity across studies. We reported the heterogeneity as low, moderate, or high with I 2 values of 25, 50, or 75%, respectively (63). Publication bias was quantitatively assessed using Egger's tests. Moreover, the trim and fill method was adopted when publication bias existed. Meta-regression analysis was considered when there were 10 or more studies (67). Subgroup and meta-regression analyses were conducted by location (Asia, America, or Africa), type of study design (cohort, case-control, or cross-sectional), sample size (<1,000 or ≥1,000), cooking fuel type [biomass fuels (wood, charcoal, crop residues, and animal dung), fossil fuels (coal and kerosene), or mixed fuels (biomass plus fossil fuels)], and outcome assessment method [direct assessment (measure and health/hospital records) or indirect assessment (maternal recall)]. Finally, sensitivity analyses were conducted to examine the influence of each study on the overall pooled estimate by the omission of each estimate one at a time (67). All the statistical analyses were performed using Stata statistical software version 14.0 (Copyright 1985-2015 StataCorp LP), and P-values < 0.05 were considered significant. Figure 1 shows the literature screening processes in detail. In brief, 2,604 publications were identified through PubMed (n = 441), Embase (n = 421), Web of Science (n = 943), and Scopus (n = 798). An additional article was identified by a manual search of reference lists of included studies (n = 1). After excluding duplicates (n = 959), 1,536 records were excluded based on title and abstract. We obtained 109 articles for in-depth evaluation, of which 77 were excluded after reviewing the full text. Ultimately, 32 studies were enrolled in our final systematic review and meta-analysis.

Pooled analysis
In this updated meta-analysis, we found polluting cooking fuels to be associated with LBW, SGA, stillbirth, and PTB. A total of 24 studies were included for investigating the relationship between polluting cooking fuels and LBW; the pooled OR was 1.37 (95% CI: 1.24, 1.52) with high heterogeneity (I 2 = 75.0%, P < 0.001) ( Figure 2). We found evidence of publication bias (Egger's test P < 0.001) (Supplementary Figure 1). After applying the trim and fill method, the corrected OR was 1.13 (95% CI: 1.01, 1.26). A total of 11 studies were explored for SGA; the pooled OR was 1.48 (95% CI: 1.13, 1.94), with high heterogeneity (I 2 = 88.7%, P < 0.001) (     LBW Environmental tobacco smoke, sex of the child, birth order of the child, mother's age at childbirth, mother underweight, mother's anemia status, antenatal care during pregnancy, pregnancy intention, mother's tobacco use, mother's education, social groups, wealth quintiles, and area of residence

Subgroup and meta-regression analyses
To explore possible sources of heterogeneity across eligible studies, we performed subgroup analyses for LBW and SGA (Table 2). Notably, the association between polluting cooking fuels and the risk of LBW was consistent in all subgroups examined, except for the subgroup conducted according to fuel types. In the subgroup analysis of cooking fuel type, biomass (OR: 1.57; 95% CI: 1.34-1.84) and fossil fuels (OR: 1.49; 95% CI: 1.10-2.02), but not mixed fuels (OR: 1.14; 95% CI: 0.98-1.32), were associated with LBW. Subgroup analysis based on location, the association between polluting cooking fuels, and risk of SGA was significant in Asia (OR: 1.47; 95% CI: 1.03-2. 10  subgroup analysis of study design, the association was significant in cohort studies  Table 6), we found that study design (cohort study: P < 0.01; cross-sectional study: P < 0.01) and sample size (≥1,000: P < 0.01) were the covariates associated with heterogeneity. Cooking fuel types (mixed fuel: P < 0.05) were covariates associated with the heterogeneity in the SGA analysis.

Sensitivity analysis
Sensitivity analysis results showed that excluding one study at a time did not significantly alter the overall effect of polluting fuel use on LBW (OR altered between 1.33 and 1.41) (Supplementary Table 7) and SGA (OR altered between 1.26 and 1.56) (Supplementary Table 8). Regarding the studies reporting SGA, heterogeneity was greatly reduced when Chaudhary's study (34) was excluded (I 2 reduced to 15.9%) (Supplementary Table 8).

Discussion
In a previous meta-analysis, Pope et al. reported that indoor air pollution from solid fuel use was associated with a 38% increased risk of LBW and a 51% increased risk of stillbirth (25). Amegah et al. reported that solid fuel use was associated with a 35% increased risk of LBW, a 30% increased risk of PTB, and a 29% increased risk of stillbirth (26). These results are consistent with ours.
This is the first meta-analysis to summarize the available evidence relating to polluting cooking fuels and SGA, and we found a positive association between them. The combustion of polluting cooking fuels emits high levels of pollutants, such as particulate matter (PM) and nitrogen dioxide (NO 2 ), which have been shown to be associated with SGA in several meta-analyses .
/fpubh. . (74-76). We included 11 relevant studies exploring the effect of polluting cooking fuels on SGA; further eligible studies are needed to verify the association.
In the subgroup analysis, the summary OR varied between different locations for SGA. This may be attributed to the different cooking fuel choices in the studies. For example, the majority of women in South Asia use wood (49.1-89.7%), and most of the women in Africa use charcoal (85.4-93.5%), whereas women in Latin America mainly use liquefied petroleum gas (69.1-97.6%) (77). However, this study only included five studies in South Asia (11, [33][34][35][36], two studies in Africa (44,45), and one study in Latin America (28) to detect the association between polluting fuels and SGA. Therefore, more studies are needed to explore regional differences.
Different fuel types may influence the strength of the association between polluting fuels with LBW and SGA. The results demonstrated that biomass fuels had a larger pooled OR for the association with LBW than fossil fuels. However, when various polluting cooking fuels were grouped together, the association was not significant. Studies have shown that the concentrations of pollutants released from coal are lower than those released from biomass (78,79). A laboratory assessment has also shown that the concentrations of pollutants released from fossil fuels (including kerosene) are lower than those from biomass (80). Therefore, the pooled OR could be higher between biomass fuels and LBW. However, we found the pooled OR of the association between fossil fuels and SGA was higher than that with biomass fuels. The result was consistent with the included studies (24, 37). In addition, the results of the included studies also showed higher OR values between biomass and LBW than fossil fuels. The two cohort studies were both conducted in China, and more studies are needed to confirm this conclusion and explore the potential biological mechanism. Studies have reported that the concentrations of pollutants are reduced when people use mixed fuels (78,81). This could be attributed to the occasional use of relatively low polluting fuels (e.g., kerosene) (81), which masked the relationship between one type of polluting cooking fuel (biomass or fossil fuels) and adverse birth outcomes.
There was high heterogeneity in the present study, although the meta-regression analyses found several covariates associated with heterogeneity, which was still uncontrolled after subgroup analyses. Moreover, sensitivity analysis was performed to explore the potential source of heterogeneity in LBW and SGA. However, the source of heterogeneity was not found in LBW. This may be attributed to the number of included studies that were not sufficient to detect the source of heterogeneity. In the future, more qualified original research is needed to explore the source of heterogeneity. For SGA, sensitivity analysis revealed that after excluding Chaudhary's study (34), the heterogeneity was reduced greatly. In this cross-sectional study, mothers of SGA infants had higher rates of exposure to polluting fuels during pregnancy, and mothers with appropriate size for gestational age infants had lower rates of exposure. Therefore, the population with different exposure rates may be a source of heterogeneity. Furthermore, .
/fpubh. . the adjustment for different covariates may be a source of heterogeneity. Compared with other studies, Chaudhary's study had additionally adjusted high carbohydrate snacks, pregnancyinduced hypertension, gestational diabetes, cardiovascular diseases, polyhydramnios, hypothyroid, and anemia. Combustion of polluting cooking fuels emits high levels of pollutants, such as PM, CO, NO 2 , sulfur dioxide (SO 2 ), and PAHs (37). Exposure to PM induces maternal systemic and placental oxidative stress and inflammation (82), which could result in suboptimal placentation, and subsequent fetal growth restriction (83). It is known that PAHs are linked to developmental and reproductive toxicity (84). They can result in inadequate transplacental nutrient exchange (85). Carbon monoxide binds to hemoglobin to form carboxyhemoglobin (86), which leads to inadequate oxygen supply to the fetus and hence fetal growth retardation (35). Moreover, it may cross the placental barrier, where it can act directly to affect fetal health and development (44). In short, both PM and toxic chemicals affect the growth and development of the fetus through various mechanisms, leading to adverse birth outcomes.
The current study had some strengths. In this updated metaanalysis, we provided the first quantitative assessment of the association between polluting cooking fuels and SGA. Compared with the previous study (26), this meta-analysis included 23 new studies and conducted subgroup and meta-regression analyses for LBW and SGA. In addition, we explored whether different types of polluting fuels use had a different impact on the risk of adverse birth outcomes.

Study limitations and future perspectives
Our study also had some limitations. First, because of the observational nature of included studies, recall and selection bias cannot be eliminated. Although cohort studies are less vulnerable to such bias, only 11 cohort studies were included in the final analysis. In addition, cross-sectional and case-control studies have difficulties in determining the temporality between exposure and outcome. Consequently, our current results should be interpreted cautiously, and more prospective studies are needed to thoroughly investigate the association between polluting cooking fuels and adverse birth outcomes. Second, there is a possibility of misclassification of exposure and outcome assessment. All included studies collected information regarding primary cooking fuels through interviews. However, there is considerable variability in exposure, such as multiple fuel use and temporal changes in fuel use. Using only qualitative indicators (such as reported cooking fuel use) could produce considerable exposure misclassification for exposure settings and inaccurate correlation assessments between exposure and outcomes. By directly measuring and recording the concentration of kitchen air pollutants, we can avoid misclassification of exposures and obtain accurate exposureresponse relationships. However, only three of included studies directly measured the concentration of kitchen air pollutants (28,38,40). It is important for future studies to employ more accurate methods, such as household or portable air quality monitors, to assess household air pollutant exposure. Moreover, the baby's birth size was used as a proxy for birth weight when .
/fpubh. . the information could not be retrieved from a health card or maternal recall. However, previous findings demonstrated that the demographic and health survey data on birth size could be used as an alternative to birth weight (23). Third, although some potential confounding factors were taken into account in the original studies, other confounding factors, such as the type of cooking stoves, personal exposure, and the availability of windows or chimneys, still remained uncontrolled in some studies. A study conducted in Indonesia demonstrated that pollutant concentrations remained high even when the houses were adequately ventilated (31); however, the residual confounding effects of other important factors could not be eliminated, which might cause inaccurate evaluations of their effects on the risk of adverse birth outcomes. Finally, there was considerable heterogeneity in the current study.
After conducting subgroup and meta-regression analyses, the heterogeneity was still uncontrolled. Moreover, sensitivity analysis suggested that no individual study significantly affected the pooled effect size. Therefore, the heterogeneity is the cause for caution concerning the conclusion. Strong epidemiological evidence estimating health risks from particular fuels and technologies utilized, and improved and standardized data collection capturing the fuels and technologies used in the home for all key enduses, such as cooking, heating, and lighting, are needed. The understanding of exposure to total household air pollution will be improved through increased monitoring efforts combined with future modeling that takes stove stacking into account. This will better inform policy and programmatic decision-making as well as the global monitoring of health and environmental impacts.

Conclusion
Our findings suggest that polluting cooking fuel use is associated with an increased risk of LBW, SGA, stillbirth, and PTB. Different polluting fuel types may influence the strength of the association between polluting fuel use and LBW and SGA. Moving forward, we encourage more prospective studies to thoroughly investigate the association between polluting cooking fuel use and adverse birth outcomes. On exposure assessment, a more direct method of measuring, such as household or portable air quality monitors, will help quantify exposure information and an exposure-response relationship.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.