The results of our study indicate a strong, exposure-related inverse association between maternal use of charcoal as cooking fuel during pregnancy and birth weight of the newborn. The average birth weight of babies born among exposed mother was 243g (95% CI: 496, 11) lower compared with the babies of mothers using LPG as cooking fuel. Garbage burning at home was an important risk factor for LBW. Garbage burning was associated with a 195% (RR=2.95; 95% CI: 1.10, 7.92) increase in the risk of LBW. Joint evaluation of these two exposures resulted in further reductions in birth weight and additional increase in the risk of LBW.
Validity of results
We selected consecutive mothers giving birth in a teaching hospital and thus the study population represents a defined catchment area. We achieved a high response rate (91.5%) which minimizes selection bias. Mothers giving birth at KBTH as against other facilities in the study area where not a distinct cohort from the source population but do so mainly because of the hospital's proximity to their homes and/or the comprehensive and specialist services on offer. We also carefully excluded mothers referred to the facility for whatever reason from the study. The outcome of interest was measured and recorded independently from the study and represent a well defined and objective variable with a negligible measurement error. Information on exposure and potential confounders was collected retrospectively. Use of cooking fuel choices and garbage burning practices represents quantitatively well-defined entity and it is reasonable to assume that retrospective data collection resulted in relatively reliable data on maternal exposure during pregnancy. The ordinal scale exposure variable per the use of maternal report of fuel type, and cooking and garbage burning practices was sensitive to certain amount of measurement error but was however not likely to be related to the outcome of interest. In summary, it reasonable to assume the exposure assessment reflects reasonably well the exposure conditions during pregnancy.
Although we applied a cross-sectional study design, we were able to collect reasonably valid exposure information from the time period relevant for causation of reduction in fetal growth and therefore our results support a causal relation between the exposures and the outcome of interest. Establishment of temporal relation may be problematic in some research settings with the use of cross-sectional study design. This however should not be a concern in our study because it is clear that exposure to combustion pollutants was present during pregnancy among mothers who cooked with charcoal and those who reported garbage burning at their homes. Also it was increasingly clear from the information collected from the mothers and summarized by the investigators that the choices of cooking fuels and garbage burning activities remained relatively stable in our research settings. It is possible that some women using charcoal during most of the duration of pregnancy might have reported use of LPG as their primary fuel, but this information bias would rather tend to underestimate the true effect. We were not able to undertake air quality measurements in homes of the mothers. The potential for exposure misclassification nonetheless is reduced in our study because the quantitative assessment of exposure to combustion pollutants from charcoal use and garbage burning comprised several types of information collected from participants and summarized by the investigators, including duration of cooking, time spent in cooking area, type of ventilation applied as well as frequency of garbage burning at home and how often mothers were present during combustion.
The study adjusted for the effect of age, social class, marital status and gravidity of mothers, and sex of neonate in the analysis. We had reliable information on gestational age of mothers but did not consider it as a covariate in the analysis based on a recent work by Wilcox and colleagues
 which reported gestational age as a collider and provided evidence of the likely bias produced by adjustment for gestational age in statistical analysis. We were however unable to examine the effect of other determinants of birth weight such as maternal nutrition and anthropometry, malaria and sexually transmitted infections. IAP exposure during pregnancy is not expected to be dependent on these factors hence the estimated effect of IAP exposure on birth weight is not likely to be confounded by these factors. Maternal smoking is another important determinant of birth weight, but in Ghana only few women smoke. The 2008 Ghana Demographic and Health Survey
 estimated the proportion of women smoking cigarettes and other tobacco products to be 0.4%. Maternal smoking can therefore not be considered as a serious threat to validity in this study. Cooking fuel was related to social class (p<0.0001), educational level (p<0.0001) and occupation (p<0.0001) of mothers and likely to influence IAP exposure experience during pregnancy. LPG was the primary fuel of high social class mothers with charcoal the preferred fuel of low social class mothers. Uneducated and semi-literate (educated up to junior high school) mothers preferred charcoal with tertiary level educated mothers patronizing LPG. Traders, street vendors, hairdressers and seamstresses had preference for charcoal with office workers preferring LPG. Studies in Ethiopia
 and Kenya
 also associated household cooking fuel choices with employment, income, educational level and social class of women. We adjusted for the effect of social class but not educational level and occupation of mothers in the analysis due to the well known fact that education and occupation determines social class of an individual, and also the fact they were unrelated to birth weight in our analysis. Controlling for confounding by social class is always problematic especially in our study where more than half (58.8%) of the study participants were low social class. This is due to the strong effects of social class on health outcomes. We do not therefore overrule the possibility of residual confounding by social class in our study, but we think residual confounding does not solely explain our observations on the adverse effects of combustion products from charcoal and garbage burning.
Synthesis with previous studies
A systematic literature search identified six previous studies conducted in Guatemala
[9, 10], Zimbabwe
 and India
[13, 14] that have assessed the relationship between indoor pollution from solid fuel use and birth weight. The study in Guatemala
 found babies born to mothers who used wood to be 63g lighter than babies born to mothers who used gas or electricity. The other study in Guatemala
 was a randomized control trial and found infants born to mothers who used open fires (control group) to be on average 89g lighter than infants whose mothers used a chimney stove (intervention group). The study in Zimbabwe
 found babies born to mothers cooking with wood, dung, or straw to be on average 175g lighter compared with babies born to mothers using LPG, natural gas, or electricity. In the Pakistan study
, infants born to wood users were on average 82g lighter than infants born to natural gas users. This study also estimated the population attributable risk for LBW explained by wood use to be 24%. The earlier Indian study
 found infants born to women from households using wood and/or dung as primary cooking fuel to be 104.5g lighter than infants born to mothers from households using biogas or kerosene. This study also reported exposure to biomass fuel to be associated with an adjusted 49% increased risk of LBW. The recent Indian study
 found children born in households using high pollution fuels (wood, straw, animal dung, crop residues, kerosene, coal and charcoal) to be 73g lighter than those born in households using low pollution fuels (electricity, LPG, natural gas and biogas). A recent meta-analysis
 of five studies examining this relationship also estimated a reduced mean birth weight of 95.6g (95% CI: 68.5, 124.7) and an increase risk of LBW of 38% (OR = 1.38, 95% CI: 1.25, 1.52) among women exposed to IAP. This study also estimated the population attributable risk for LBW explained by IAP to be 21%.
The findings of our study are consistent with these previous studies albeit our effect estimates were quite larger than any previously reported. The similar results produced from the sensitivity analysis means use of charcoal as cooking fuel and garbage burning at home represents an important threat to optimal fetal growth. We do not by any means imply with the large effect sizes reported that charcoal is a high polluting and more potent fuel than other biomass like wood, dung and crop residue. Garbage burning on the other hand releases dioxins, hazardous chemical substances that have been shown in animal studies to severely impair fetal growth even at low levels of exposure. Studies in human populations have also reported associations of low level dioxin exposure during pregnancy with decreased birth weight
[21–24]. Our study and the previous reviewed however did not actually measure the quantity of biomass combusted and the amount of pollutants released for which mothers were exposed. It is therefore reasonable to assume that our study participants might have on average combusted large quantities of charcoal and garbage during pregnancy with the cumulative adverse effect reflected in the large effect estimates reported. The significant exposure-response relationship observed by our study to some extent confirms this assertion. We must however emphasize that unmeasured confounding, and residual confounding by social class as already noted could contribute to the large effect sizes reported in spite of efforts to eliminate this potential confounding from our study.
The studies highlighted however had some limitations which our study was purposely designed to address and strengthen the epidemiological evidence. Firstly, the studies in Zimbabwe
 and India
 relied on mothers self-report of child size at birth in respectively estimating birth weight of 47% and 60% of their study infants. This could have resulted in under or over estimation of birth weight of these infants. Secondly, the other studies cited
[9, 10, 12, 13] were community-based and obtaining timely measurement of birth weight of infants delivered at home was problematic. The time of measurement in some cases raises doubt about their acceptability as true reflection of birth weight of these infants. A baby's weight can fluctuate within the first week of life with newborns losing up to 10% of their birth weight during the first 3–5 days of life. Some of the studies in an attempt to address this limitation, restricted birth weight analysis to newborns weighed within 48–72 hours after birth. Lastly, majority of households in developing countries use a combination of cooking fuels with those in urban areas especially usually using a combination of polluting and clean fuels as a way of reducing household fuel bills and to at times hasten the preparation of meals. In Ghana for instance, majority of urban households mostly use LPG for preparing sauces, stews, soups and continental meals with charcoal used mainly for preparing staple foods such as banku, fufu and ampesi. Most of the households that use LPG for cooking are also compelled to rely on charcoal when LPG is in short supply. All the studies highlighted collected information on primary cooking fuel of participants without attempting to identify participants using a combination of cooking fuels. This distinction is important for proper quantification of exposure experiences of study participants. For instance, in settings where use of combination of polluting and clean fuels prevails, assessing only primary cooking fuel could result in under or over estimation of exposure. About 19% of the mothers who participated in our study used a combination of charcoal and LPG for cooking.
Our study to the best of our knowledge is the first to examine the contribution of garbage burning at home to the indoor exposure experience of pregnant women, and its relation with birth weight of newborns. Over a quarter (27%) of the mothers studied reported garbage burning in their homes during pregnancy. Garbage burning is a frequent practice in a number of Ghanaian urban households as a way of managing their solid waste. This is because most urban areas especially the secluded and deprived zones are usually not reached with waste collection services. Areas receiving these services are also faced with untimely and irregular service provision. In rural Ghana, the situation is different with younger members of the household tasked with disposing of the household waste at designated sites in the community each morning. The sight of garbage burning in rural households is therefore uncommon.
Burning of charcoal, other solid fuels and garbage emits smoke which contains a number of air pollutants including carbon monoxide (CO) and particulate matter (PM). Inhaled CO and PM impairs fetal growth in two ways; (1) CO combines with hemoglobin to cross the placenta decreasing oxygen supply to tissue which limits the ability of the placenta to transfer nutrients to the fetus, and (2) PM reduces maternal lung function thereby increasing the risk of maternal lung disease, and in turn reducing oxygen delivery to the fetus as well as causing cell damage in the fetus through oxidative stress
[25, 26]. Impaired fetal growth subsequently leads to reduced or low birth weight
[26, 27]. Also, reduced oxygen transport across the placenta and fetal uptake due to reduced oxygen supply to the placenta can result in preterm delivery and consequently reduced or low birth weight
. The fetus in particular is considered to be highly susceptible to environmental pollutants because of its differential exposure pattern and physiological immaturity
[28, 29]. The high cell proliferation and changing metabolic mechanisms during the critical phase of fetal development have been identified as the physiological process that renders the developing fetus extremely vulnerable to environmental toxicants