This study is the first to investigate the associations between various types of cooking fuels and risk of LBW and SGA in the Chinese population. Our results support the hypothesis that use of biomass for cooking is associated with an increased risk of LBW compared to use of gas, and suggest that the association between biomass and LBW is likely due to prematurity but not IUGR.
Our study population primarily resided in an urban area. We noted a relatively high prevalence of gas stove use and low prevalence of biomass use compared to what we might expect in rural areas. The unbalanced prevalence of exposure may impact the accuracy of model estimation. However, the observed increased risk of LBW associated with biomass use was consistent with previous epidemiologic studies [1, 5, 15, 23, 33, 34, 36]. A cross-sectional study from India involving 14,850 infants found that using coal, kerosene, and biomass fuels for cooking was associated with a significant decrease in mean birth weight and increased risk of LBW . Another Indian study including 47,139 infants reported that using biomass as cooking fuel was associated with a slightly increased risk of smaller size at birth; however, birth weight was measured indirectly, thus the results may be biased by the mother’s subjective recall . A matched case-control study from the Gaza Strip including 446 births suggested that wood smoke was associated with an increased risk of LBW . A retrospective population-based cohort study from Pakistan with a sample size of 634 suggested use of wood for cooking was associated with increased risk of LBW . A Zimbabwean study based on 3,559 births reported that cooking with biomass was associated with reduced birth weight . A study from Guatemala involving 1,717 women also showed a reduction in birth weight in association with using wood for cooking . A birth cohort study of 9,604 participants reported that biomass was associated with increased risk of LBW and SGA . A recent study from India that included 1,744 pregnant women found that wood fuel use was associated with non-significantly increased risk of LBW .
All of these published studies were conducted in areas where there was a very high prevalence of LBW, with rates as high as 33 %. In addition, the percentage of households that used biomass as a primary fuel source was also very high (up to 79 %), suggesting that the study populations had a very low socioeconomic status (SES). Malnutrition, which is common in these areas, plays an essential role in birth weight particularly in IUGR. However, neither these important confounders nor gestational age were controlled for in these prior studies.
Increased risk of LBW associated with biomass has been consistently reported in the literature, supporting our findings. Combustion of biomass fuel emits high concentrations of airborne particulate matter (PM) and toxic chemicals including carbon monoxide (CO), nitrogen dioxide (NO2), sulfur dioxide (SO2), and polycyclic aromatic hydrocarbons (PAHs) [17, 42]. When these pollutants are absorbed into the maternal bloodstream, the O2 content of maternal blood is reduced. Subsequently, O2 delivery to placenta is reduced, resulting in preterm delivery and subsequently LBW. Both epidemiological [9, 11, 13, 35, 38] and animal studies  have linked these pollutants to LBW. Therefore, it is biologically plausible that exposure to biomass smoke increases the risk of LBW.
Our study found that the effect of biomass on LBW was attenuated after adjusting for gestational age, which indicated a negative confounding effect. After stratification by preterm and term births, a significant association with biomass was only observed among preterm births but not term births. Our study also found no significant association between biomass and SGA, consistent with results from a recent study in India . All these findings may suggest that prematurity rather than IUGR plays a major role in the association between biomass and LBW. An early study from southern India examined the effect of biomass use for cooking on LBW and SGA, and found that it was associated with both LBW and SGA, with a stronger association with LBW. In their study, the rates of LBW and SGA were approximately 33 % and 62 %, respectively , much higher than our study. These results suggested that if incidence of LBW was higher than 10 %, LBW was most likely caused by IUGR rather than prematurity, while if LBW incidence was less than 10 %, preterm infants constituted the majority of LBW . The observed associations among preterm births need to be confirmed in future studies.
The literature has described a rapid rise in caesarean sections in China in the past decades. The current national rate is nearly 40 %, irrespective of geographic location or SES . In our study population, the caesarean section rate was approximately 37.9 %, comparable to the national rate. Since the correlation between caesarean and preterm birth was relatively low (correlation coefficient = 0.08), caesarean section was unlikely a major contributor to preterm births.
A study from India reported a significant association between LBW and using coal for cooking, though they did not adjust for gestational age . This was consistent with our study results prior to adjustment for gestational age; however, after controlling for gestational age the association diminished. A suggestive positive association between coal and SGA was also observed in our study. Because we cannot know if adjustment for gestational age would affect Epstein et al.’s results , it is unclear if the association with coal is unique to India or an artifact of uncontrolled confounding. It is possible that the level of pollutants released from combustion of coal in China is lower than those in India. A large number of studies have shown that concentrations of pollutants (i.e., CO and particular matters) released from coal are lower than those from biomass in China [14, 17, 22, 31, 32]. In addition, households in China usually use honeycomb briquettes with relatively high combustion efficiency, smoke is removed via chimney, and cooking generally occurs in a separate room or building . As the majority of our study population came from an urban area, we would expect a higher percentage of households to be equipped with a chimney or hood in our study population. Our population is distinct from India, where most cooking stoves are simple (often made from mud as a U-shaped construction or three pieces of brick), have poor combustion efficiency, and are poorly ventilated .
We found that electromagnetic stoves were not associated with an increased risk of either LBW or SGA after controlling for gestational age. Electromagnetic stoves, more commonly known as induction cookers, use the electromagnetic induction principle to heat and cook food. Because the electromagnetic stove has many attractive features such as high energy efficiency, low noise, and no open flame, more families are beginning to replace coal, biomass, and gas stoves with induction stoves. To our knowledge, no previous study examined the association between electromagnetic stoves and LBW and SGA. Several studies have examined the association between electromagnetic fields (EMFs) and fetal growth. However, the results were inconsistent . Occupational exposure to EMFs has been suggested to be associated with adverse birth outcomes (e.g., LBW, preterm, and birth defects) as reviewed by Robert . Yet, studies have reported that among pregnant women, exposure to EMFs did not increase risk of LBW [3, 6, 20]. A non-significant but suggestive association between induction stove use during pregnancy and SGA in our study warrants further investigation.
The study population was recruited from the largest maternity and child care hospital in Lanzhou, the capital city of Gansu Province. The majority of study population came from Lanzhou City. Approximately 20 % of the remaining study population came from other cities and towns in Gansu Province. Although the study was hospital-based, which might impact generalizability, the LBW rate (6.2 %) in our study population was similar to the previously reported LBW rate (5.0 %) in all of Gansu Province .
Strengths and limitations should be considered when interpreting the study results. Information on birth weight was obtained from medical records and birth weight was measured in grams by trained professional nurses within the first hour of life, minimizing potential misclassification of the outcome. Information on gestational age was available, which allowed us to not only control for gestational age when studying the relationship with LBW, but also examine the association with SGA. One concern is that information on household heating source was not collected in the study. Since the majority of the study population came from urban areas where heating was centralized, the number of households using other fuels (such as coal or biomass) for heating was expected to be minimal. Generally households using coal or biomass for heating were likely to have lower SES; therefore we adjusted for SES (using education and family income as proxies) in our models. Lack of information regarding whether study participants were the primary person in charge of food preparation, number of meals cooked per day, and time spent in the kitchen might result in exposure misclassification. However exposure misclassification would likely be non-differential, if any, resulting in an underestimation of the observed associations. Given the various sources and factors that might influence indoor air pollution, it is important for future studies to employ more accurate methods, such as household or portable air quality monitors, to assess indoor air pollutants exposure.