This meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA) guidelines (26). To retrieve relevant articles, international electronic databases such as PubMed, Science Direct, Global Health, African Journals Online, HINARI, Semantic Scholar, and Google and Google Scholar searches were used. Gray literature was also identified from different university's digital libraries. The following key terms were used to search the studies: Indoor air pollution”, “household air pollution”, biofuels OR “household fuel” “biomass”, “domestic fuel”, “coal”, “cooking fuel”, “wood”, “cooking smoke”, “charcoal”, “solid fuel”, “cooking firewood”, “biomass fuel”, “biomass smoke”, “wood fuel”, “wood smoke”, “charcoal smoke”, “Air pollution”, “Particulate matter”, “PM₁₀”, “PM_2.5”, “ozone”, “nitrogen dioxide”, “sulfur dioxide”, “carbon monoxide” “Polycyclic Hydrocarbons”, “Pregnancy outcome”, “birth weight”, “low birth weight”, “low birthweight”, “premature birth”, “premature infant”, “fetal growth retardation”, “fetal development”, “gestational age”, “small for gestational age”, “small gestational age”, “fetal mortality”, “fetal death”, “perinatal mortality”, “stillbirth”, “embryo loss”, “spontaneous abortion”, “congenital abnormalities”, “neural tube defects”, “low and middle income country”, “low income country”, “middle income country”, low and middle income countries”. All key terms were combined using the Boolean operators “AND” or “OR” as appropriate. The search was carried out up to December 1, 2023, by four authors independently (CD, YS, AE, and SDK).

In this meta-analysis, we included observational studies (cross-sectional, case-control, and cohort studies) on indoor air pollution exposure and adverse pregnancy outcomes in low and middle-income countries. Studies published between 2000 and December 1, 2023, were included in the meta-analysis. However, qualitative studies, unretrievable studies, editorial letters, studies with poor methodological quality, and studies that did not report the outcome of interest were excluded from the meta-analysis.

The primary outcome of the study was to estimate the pooled association between indoor pollution exposure and adverse pregnancy outcomes in low and middle-income countries, calculated by dividing the number of adverse pregnancy outcomes by the total sample size and multiplying by 100.

After all searched articles were exported into the Endnote X20 version, and duplicate articles were removed, data was extracted by using a standard data extraction template by four authors (CD, KM, ET, and EB). The standard data extraction template consists of the author's name, country, publication year, exposure assessment, study design, prevalence, type of adverse pregnancy outcome, and sample size. Five reviewers (LA, CD, FD, MA, and AM) screened the relevant articles for eligibility, and the quality of each article was evaluated using the Joana Brigg Institute (JBI) critical appraisal checklist (27). The quality of each study was independently assessed by the four authors (CD, AAT, AK, and NK), with scores measured on a scale of 100%. A quality score of <50% was used to include articles for further analysis (28, 29). In the case of any discrepancies encountered during the quality assessment, the mean score was computed from the evaluations of all reviewers to address and resolve any differences.

DerSimonian and Liard's method of random effect model was used to determine the pooled association between indoor air pollution exposure and adverse pregnancy outcomes using STATA 14 (30). The Higgs I² statistic model was used to determine the heterogeneity of the included studies, with values of 25%, 50%, and 75% indicating low, moderate, and high heterogeneity, respectively (31). A p-value of less than 0.05 was considered indicative of the presence of heterogeneity. The publication bias was assessed using a funnel plot and Egger's test with a p-value less than 0.05 suggesting a publication bias (32).

Subgroup analysis was carried out based on various study characteristics such as sample size (large or small), year of publication (2020 and after or before 2020), quality of the study (high or low quality), and study setting (nationwide, healthcare facility or community level). Moreover, a sensitivity analysis was also performed to assess the influence of a single study on the pooled prevalence estimates.

Low birth weight: a birth weight of <2,500 g (33).

Stillbirth: “a baby who dies after 28 weeks of pregnancy but before or during birth”(34).

Neonatal death: “deaths among live births during the first 28 completed days of life” (35).

Preterm birth: “babies born alive before 37 weeks of pregnancy” (36).

Small for gestational age: those smaller in size than normal for their gestational age, most commonly defined as a weight below the 10th percentile for the gestational age (37).

Using an international electronic database, a total of 2,002 articles was identified. Out of these, 296 duplicate articles were excluded using the Endnote reference manager and 817 articles were excluded based on their titles and abstracts. Besides, 10 articles were also excluded based on the quality assessment and outcomes of the interest. Finally, 30 full-text articles were eligible for this meta-analysis (Figure 1).

A total of 30 articles (4, 9, 10, 13, 15–17, 20–22, 38–57) were included to determine the association between exposure to indoor air pollution and adverse pregnancy outcomes in low and middle-income countries. In this meta-analysis, a total 2,120,228 of study participants were included. From the included studies, Pakistan had the highest at least one adverse pregnancy outcome (35.4%) (38), and the lowest adverse pregnancy outcome was reported in India (1.7%) (44). Regarding the study setting, 12 studies were conducted in healthcare facilities (4, 13, 16, 21, 38–41, 43, 44, 46, 50), 14 in nationwide (country level) (4, 13, 16, 21, 22, 38–41, 43, 44, 46, 50, 52), and four in community level (17, 47, 51, 54). Likewise, 28 studies had scored 75% and more of JBI the quality assessment (4, 9, 10, 13, 15–17, 20, 22, 38–57), while two studies scored 62.5% (21, 51). Nine studies were carried out in India (10, 13, 16, 44, 47, 52–54, 57), three in China (17, 45, 48), three in Bangladesh (22, 41, 43), three in Ghana (9, 15, 56), two in Pakistan (38, 51), one each in Ethiopia (4), Zimbabwe (46), Nepal (42), Zambia (20), five African countries (India, Pakistan, Guatemala, Kenya, and Zambia) (49), fifteen African countries (39), Nigeria (50), Uganda (40), Indonesia (21), and Sri Lanka (55) (Table 1).

Our analysis of the 30 included studies revealed significant heterogeneity among them (I² = 100%; p < 0.001). Hence, DerSimonian and Liard's method of random effect model was carried out to estimate the pooled association between exposure to indoor air pollution and adverse pregnancy outcomes. The results of the random effect model indicated the pooled prevalence of at least one adverse pregnancy outcome was 15.55% (95% CI: 12.61–18.49) (Figure 2).

Meta-analysis showed that there is a significant association between indoor air pollution exposure and adverse pregnancy outcomes. The high pooled prevalence of adverse pregnancy outcomes was small for gestational age (23.77%) followed by low birth weight (17.74%). The pooled prevalence of preterm birth among pregnant women exposed to indoor air pollution was 16.56%. Likewise, there is a significant association between indoor air pollution exposure and stillbirth (6.11%) (Table 2).

A funnel plot was used to determine the publication bias, revealing an asymmetric distribution that strongly indicated the presence of publication bias (Figure 3). Further, statistical analysis employing the Egger regression test verified the absence of publication bias (p = 0.159). Similarly, the publication bias was also carried out using a funnel plot and statistical Egger test for low birth weight. The finding showed there was an asymmetric distribution that strongly indicated the presence of publication bias (Figure 4). However, the Egger regression test confirmed the absence of publication bias (p = 0.191).

Besides, the publication bias was also determined for small gestational age, preterm birth, and neonatal death using a funnel plot (Supplementary material 1). The Egger regression test confirmed the absence of publication bias among included studies for small gestational age (p = 0929) and preterm birth (p = 0.891), neonatal death (p = 0.322).

To pinpoint the source of heterogeneity among included studies (I² = 100%, p < 0.001), subgroup analysis was conducted based on study setting (nationwide, healthcare facility or community), sample size (small or large), quality of the study (high or low) and year of publication (2020 and after or before 2020). The study conducted at the community level had higher pooled adverse pregnancy outcomes (20.71%), with extreme heterogeneity among included studies (I² = 99.9%, p < 001) followed nationwide (16.06%) (Table 3). Regarding the year of publication, the highest pooled prevalence of adverse pregnancy outcomes was observed among studies conducted before 2020 (17.01%; 95%CI: 13.39–29.63) than studies conducted after 2020 (12.48%; 95% CI: 9.15–15.80). In addition, the high pooled prevalence of adverse pregnancy outcomes was observed among small sample size studies (16.71%; 95%CI: 12.46–20.96) as compared to studies conducted with large sample size (15.28%; 95%CI: 12.04–18.52) (Table 3).

A sensitivity analysis was performed to evaluate the impact of individual studies on the overall pooled estimate of adverse pregnancy outcomes, and the results indicated that no single study exerted a significant effect (Figure 5). Similarly, we also evaluate the impact of individual studies on the overall pooled estimate of low birth weight, small for gestational age, preterm birth, neonatal death, and stillbirth, and the findings suggest that there is no evidence of a single study's effect on the overall pooled prevalence (Supplementary material 2).

Exposure to indoor air pollution, such as biomass fuel, particulate matter, and kerosene was statistically significantly associated with adverse pregnancy outcomes. Twenty-three studies were included to determine the association between biomass fuel exposure and adverse pregnancy outcomes (4, 9, 10, 13, 15–17, 21, 22, 40–44, 48–52, 54–57). Thirteen of the included studies had a positive association (9, 13, 17, 21, 41, 43, 44, 48–51, 54, 55), while negative association in 10 studies. The pooled results from random effect analysis showed that exposure to biomass fuel would increase the risk of adverse pregnancy outcomes by 1.16 (OR = 1.16; 95% CI: 1.12–1.20), with significant heterogeneity (I² = 88.8%; p < 0.001) (Figure 6).

The association between particulate matter exposure and adverse pregnancy outcomes was determined based on four studies (20, 39, 47, 56). Two of the included studies had a positive association (39, 47), while negative association in the remaining two studies (20, 56). The odds of adverse pregnancy outcomes were 1.28 times higher among pregnant women who had particulate matter exposure than those who had no exposure (OR = 1.28; 95% CI: 1.25–1.31), with significant heterogeneity (I²= 70.1%; p = 0.018) (Table 4).

Similarly, the association between kerosene exposure and adverse pregnancy outcomes was determined based on the finding of two studies (13, 44). There is a significant association in one study (44) and a non-significant in another study (13). Meta-analysis showed that the pregnant women who had kerosene exposure were 37% times more likely to have adverse pregnancy outcomes than counterparts (OR = 1.37; 95%CI: 1.09–1.66) (Table 4).

The study followed the updated preferred reporting items for systematic review and meta-analysis. In this meta-analysis, all types of adverse pregnancy outcomes as a result of exposure to indoor air pollution were pertinently assessed. One of the limitations of this meta-analysis was that it did not establish causality between independent and dependent variables since the majority of the included studies were cross-sectional and case-control study designs.