Airborne Endotoxin Concentrations in Homes Burning Biomass Fuel

Background About half of the world’s population is exposed to smoke from burning biomass fuels at home. The high airborne particulate levels in these homes and the health burden of exposure to this smoke are well described. Burning unprocessed biological material such as wood and dried animal dung may also produce high indoor endotoxin concentrations. Objective In this study we measured airborne endotoxin levels in homes burning different biomass fuels. Methods Air sampling was carried out in homes burning wood or dried animal dung in Nepal (n = 31) and wood, charcoal, or crop residues in Malawi (n = 38). Filters were analyzed for endotoxin content expressed as airborne endotoxin concentration and endotoxin per mass of airborne particulate. Results Airborne endotoxin concentrations were high. Averaged over 24 hr in Malawian homes, median concentrations of total inhalable endotoxin were 24 endotoxin units (EU)/m3 in charcoal-burning homes and 40 EU/m3 in wood-burning homes. Short cooking-time samples collected in Nepal produced median values of 43 EU/m3 in wood-burning homes and 365 EU/m3 in dung-burning homes, suggesting increasing endotoxin levels with decreasing energy levels in unprocessed solid fuels. Conclusions Airborne endotoxin concentrations in homes burning biomass fuels are orders of magnitude higher than those found in homes in developed countries where endotoxin exposure has been linked to respiratory illness in children. There is a need for work to identify the determinants of these high concentrations, interventions to reduce exposure, and health studies to examine the effects of these sustained, near-occupational levels of exposure experienced from early life.


Research
The use of solid or biomass fuels to cook and to heat homes is widespread in large parts of the developing world, with an estimated 3 bil lion people exposed to smoke from burning these fuels in their own home (International Energy Agency and Organisation for Economic Cooperation and Development 2004). The World Health Organization estimates that bio mass fuel smoke exposure is responsible for about 1.5 million early deaths per year (Prüss Ustün et al. 2008), with a global burden of disease of approximately 2.5% of all healthy lifeyears lost. Most of the burden of disease arises from respiratory infections, especially in children < 5 years of age, with a dispropor tionate amount of health problems falling on women and children, who are more likely to be at home or to have responsibilities for cooking and heating activities (Rehfuess et al. 2006).
Research into indoor air pollution in homes burning biomass fuels has tended to focus on airborne concentrations of fine particulate mat ter (PM) (Albalak et al. 2001;Edwards et al. 2007;Ezzati et al. 2000;Fullerton et al. 2009;Kurmi et al. 2008), but airborne endotoxin may also play an important role.
Endotoxin or lipopolysaccharide is part of the cell wall of Gramnegative bacteria and has been measured in airborne PM in occupational settings, office buildings, households, and ambient air. Once inhaled, endotoxin stimu lates an amplifying series of endotoxin-protein and protein-protein interactions, sequentially binding to a range of proteins and receptors, leading to production of chemotactic cytok ines and chemokines (Hađina et al. 2008) and lung inflammation and resultant oxidative stress. Studies have shown associations between household endotoxin concentrations and diag nosed asthma, asthma medication use, and severity of asthma symptoms (Michel et al. 1996; Thorne et al. 2005). Respiratory illness in endotoxinexposed working populations has been frequently documented (Smit et al. 2008; Thorne and Duchaine 2007). In asthma, endo toxin exposure has been shown to be protective of development of allergic disease at low levels while also producing nonallergic asthma and/ or aggravating symptoms of existing asthma (Douwes et al. 2003;Smit et al. 2009). Thorne and Duchaine (2007) tabulated data on endotoxin levels in a wide variety of industries and home environments, which indicated geometric mean (GM), inhalable fraction, personal exposures of 12-8,300 endotoxin units (EU)/m 3 in agricultural occu pations and 5.8 EU/m 3 in homes of rural asthmatic children (n = 326).
Airborne endotoxin concentrations in homes in Boston, Massachusetts (USA), have been shown to be significantly associated with the presence of dogs, moisture sources in the home, and the amount of settled dust (Park et al. 2001). Endotoxin has also been identi fied in tobacco smoke (Larsson et al. 2004) and in homes where smoking takes place, pets are present, and/or dampness or mold is found (Rennie et al. 2008;Tavernier et al. 2006). In the large U.S. National Survey of Endotoxin in Housing (Vojta et al. 2002), increased household endotoxin was most strongly associ ated with living in poverty, number of people in the home, pet ownership, and household cleanliness (Thorne et al. 2009). Most studies have used endotoxin levels in settled dust as a surrogate for personal exposure. There are very few studies that have measured airborne endo toxin concentrations in household settings.
It seems probable that the burning of com mon biomass fuels such as wood, charcoal, dried animal dung, and crop residues within small and poorly ventilated homes will pro duce high endotoxin exposures. The only avail able report in the scientific literature comes from a small study in the Ladakh region of India, where shortterm sampling (< 60 min) of two homes produced average endotoxin Background: About half of the world's population is exposed to smoke from burning biomass fuels at home. The high airborne particulate levels in these homes and the health burden of exposure to this smoke are well described. Burning unprocessed biological material such as wood and dried animal dung may also produce high indoor endotoxin concentrations. oBjective: In this study we measured airborne endotoxin levels in homes burning different biomass fuels. Methods: Air sampling was carried out in homes burning wood or dried animal dung in Nepal (n = 31) and wood, charcoal, or crop residues in Malawi (n = 38). Filters were analyzed for endotoxin content expressed as airborne endotoxin concentration and endotoxin per mass of airborne particulate. results: Airborne endotoxin concentrations were high. Averaged over 24 hr in Malawian homes, median concentrations of total inhalable endotoxin were 24 endotoxin units (EU)/m 3 in charcoal burning homes and 40 EU/m 3 in woodburning homes. Short cookingtime samples collected in Nepal produced median values of 43 EU/m 3 in woodburning homes and 365 EU/m 3 in dungburning homes, suggesting increasing endotoxin levels with decreasing energy levels in unprocessed solid fuels. conclusions: Airborne endotoxin concentrations in homes burning biomass fuels are orders of magnitude higher than those found in homes in developed countries where endotoxin exposure has been linked to respiratory illness in children. There is a need for work to identify the determinants of these high concentrations, interventions to reduce exposure, and health studies to examine the effects of these sustained, nearoccupational levels of exposure experienced from early life. concentrations of 24 and 190 EU/m 3 (Rosati et al. 2005). These concentrations are within the range of those found in occupations involved in the handling and processing of large volumes of biological material.
In this article we present results from a study to measure endotoxin levels within the main living area of 69 homes in Malawi and Nepal and to explore differences in these con centrations based on the fuel type being used.

Study population and sampling strategy.
Samples of airborne PM were collected from homes in two studies that assessed indoor air pollution and health in Malawi and Nepal. In Nepal the Dhanusha district was selected. This is a flat, lowlying area of the country close to the border with India. Two villages were sam pled: one in the south of the district (Lohana), where dried cow or buffalo dung is burned, and one in the north (Dhalkebar), where wood is burned. Fifteen homes were sampled in each village, during cooking time in the morning or evening in December 2008. After consent was given by an adult householder, air sampling equipment was placed in the main living area of the home and sampled air between 90 and 180 min. This study had ethical approval from the Nepal Health Research Council.
For the Malawi study, details of methods used and results of PM concentrations meas ured have been previously published (Fullerton et al. 2009). In summary, a total of 75 homes were recruited from around Blantyre and rural Chikwawa villages during April 2008. Sampling equipment was placed in the main living area of each of these homes for a period of approxi mately 24 hr, except in six homes where short term sampling similar to that used in Nepal (60-200 min duration around the time of a cooking event) was carried out (all respirable samples; n = 4 wood burning; n = 2 maize crop residue burning). Not all homes received an instrument capable of providing a sample for the measurement of endotoxin concentrations; we therefore present a subsample of data from 38 (19 rural and 19 urban) of the 75 Malawian homes. This study had ethical approval from the Research Ethics Committee of the College of Medicine, University of Malawi, and the Liverpool School of Tropical Medicine.
Sample collection. Air sampling was con ducted by placing a small Apex air pump (Casella, Bedford, UK) attached to either a cyclone sampling head (2.2 L/min) or an Institute of Occupational Medicine sampling head (2.0 L/min) to sample the respirable (median aerodynamic diameter, 4 µm) or the total inhalable (defined as anything that can be breathed into the nose and mouth and is broadly particulate with an aerodynamic diameter < 100 µm) particle size fraction of PM, respec tively. Both types of sampling heads were loaded with preweighed 25mm glass fiber filters with a 0.7µm pore size. All samples in Nepal were collected using an IOM sampling head, whereas 32 of the 38 Malawian samples were collected using a cyclone. Sampling was performed in accordance with Methods for Determination of Hazardous Substances (MDHS) no. 14/3 (Health and Safety Executive 2000). The equip ment was placed in the main living area of the home at a height of approximately 1.0 m and, where possible, at about 1.0 m from the main stove or cooking area. After sampling, each filter was placed in a sealed metal tin and sent back to the United Kingdom for reweighing before being further transported to the United States for endotoxin analysis. Field blanks were used to correct the data for changes in filter weight associated with manipulation.
Endotoxin determinations were based upon the maximum slope of the absorbance versus time plot for each well.
The arithmetic mean (14.4 EU/sample) for the six Malawi filter field blanks was subtracted from each of the other Malawi filter results. The analytical limit of detection (LOD) was derived from using a value of three times the standard deviation (9.24 EU/filter) of the field blank measurements (Malawi filter analyti cal LOD = 27.7 EU/filter). Where corrected filter values were less than the LOD (n = 12), the filter was assigned a value of onehalf the LOD (13.9 EU/filter). A similar process was applied to the Nepal filters based on results from four field blanks (arithmetic mean = 4.4; SD = 0.95 EU/filter; analytical LOD = 2.85 EU/filter). For the Nepal filters with corrected values less than the LOD (n = 4), a value of 1.43 EU/filter was assigned.
Statistical analysis. Data were double entered to a Statistical Package for the Social Sciences (SPSS), version 17.0 file (SPSS Inc., Chicago, IL, USA), and summary statistics and box plots were generated directly. Mean endo toxin concentrations measured in Nepalese total inhalable dust samples from wood and dungburning homes were compared using a Mann-Whitney Utest. A similar test was used to test for differences in respirable endo toxin concentrations in Malawian charcoal and woodburning homes.

Results
Tables 1 and 2 provide summary statistics of the measured total inhalable and respirable  directly comparable with total inhalable endo toxin concentrations and will be an underes timate of total inhalable levels, the respirable data are broadly supportive of the increasing gradient in endotoxin concentrations: charcoal < wood < cow dung < maize crop residues. Figure 1 is a box plot of airborne endotoxin concentrations from the directly comparable samples taken during cooking from wood burning and dungburning homes in Nepal. There is a statistically significant difference in airborne endotoxin concentrations in Nepalese homes burning dung compared with those burning wood (Mann-Whitney Utest, z = 4.0; p < 0.01). The much larger endotoxin concen trations in dungburning homes do not appear to be simply a function of the increased PM produced in this type of fuel. Figure 2 illustrates the amount of endotoxin per mass of PM meas ured in the Nepalese villages and demonstrates that dunggenerated smoke tends to contain much more endotoxin than does a similar mass of woodgenerated smoke (z = 2.2; p = 0.024). Figure 3 presents data on 24hr respirable endotoxin concentrations from the Malawian homes burning charcoal or wood. The differ ence between fuel types is not statistically sig nificant (z = 0.46; p = 0.647). Figure 4 shows the median endotoxin concentration per mass of respirable PM, again for the 24hr samples collected in homes in Malawi. The median concentrations of endotoxin per mass of dust is higher in woodburning homes than in char coalburning homes, although this is not statis tically significant (z = 0.243; p = 0.808).

Discussion
Endotoxin concentrations reported in this study are high and much higher than those found in a recent study measuring airborne endotoxin in 10 homes in northern California (Chen and Hildemann 2009), where mean concentrations were generally < 1 EU/m 3 , and in a study of homes of rural asthmatic children, where the GM inhalable endotoxin was 5.8 EU/m 3 (n = 326) (Thorne and Duchaine 2007). They were also considerably higher than those measured from a large study of the homes of 332 children in Canada (Dales et al. 2006). The mean air borne endotoxin concentration in the Canadian study was 0.49 EU/m 3 , almost 100 times less than the 24hr average levels measured in this study for charcoalburning homes and close to 1,000 times lower than the average level during cooking with dried dung in homes in Nepal. However, results from the Canadian study showed that even at the relatively low levels of exposure experienced by the Canadian study population, there was a statistically significant relationship between airborne endotoxin and respiratory illness in the first 2 years of life.
The only previous study of endotoxin con centrations in biomassburning homes was car ried out in two homes in the Ladakh region of India (Rosati et al. 2005), where endotoxin lev els of 24 and 190 EU/m 3 were found, broadly in line with our data. The Indian homes were small, portable tentlike structures with little in the way of ventilation or extraction of smoke generated from burning dung and crop residues.
A healthbased guidance limit of 50 EU/m 3 has been recommended for occu pational settings in the Netherlands (Heederik and Douwes 1997) for an 8hr timeweighted average exposure. The median value of 24hr samples collected from charcoalburning homes (using respirable dust size selection and hence conservative compared with the total inhalable dust sampler used for the limits proposed in the Netherlands) was approximately 20 EU/m 3 .   Figure 4. Box plot of 24-hr airborne respirable endotoxin by fuel type per PM mass on the filter in Malawian homes. The line inside the box represents the median value, the lower and upper box lines represent the limits of the interquartile range (25th and 75th percentiles), and the "whiskers" represent the 5th and 95th percentiles of the distribution. The circle indicates an outlier observation as described in Figure 2; the asterisk indicates an observation more than three times the interquartile range from the 25th or 75th percentile. Difference in means p = 0.808. Scaling this to an 8hr timeweighted average would produce levels of around 60 EU/m 3 , exceeding the concentration deemed to be acceptable for a healthy workforce. From our results, we would anticipate much higher 8hr timeweighted average values from wood and dungburning homes, and it seems likely that many of these would approach or exceed the healthbased guidance limit value. The health effects of exposure to the endo toxin concentrations measured in the homes in this study may be considerable, particularly because exposure is sustained and occurs from birth in most homes. Personal exposures of women who carry out cooking and fire light ing have the potential to be even higher than the static or area measurements made in this study because of regular close proximity to the smoke plume. There is a need for personal exposure data in these settings.
We acknowledge that this study has several important weaknesses. We did not design the study to collect samples for analysis of endo toxin, but rather "piggybacked" it onto two studies that set out to characterize PM con centrations in homes in Malawi and Nepal. As a consequence, our results present data from both short cooking periods and longer 24hr samples and also a mixture of total inhalable and respirable PM size selection. In addition, there was an extended period between the col lection of the filters and analysis for endotoxin, and we believe that this led to the high levels of contamination of some of the field blanks that we have reported. This is particularly evident for the Malawi samples, which were stored for the longest duration. We report our data separately by size fraction, sampling duration, country, and fuel type and used appropriate methods for blank correction to overcome these weaknesses where possible.
Further work should use a standard proto col for endotoxin measurement and should seek to standardize durations of sample collection. Optimally, personal exposure measurements should be considered, especially in the context of healthrelated exposure measurement. Our study design collected only two samples from homes burning crop residues, and any future study should seek to address this data gap.
Controlling and reducing exposure to bio mass fuel smoke in homes in the developing world are complex and difficult areas with such options as modifications of behavior, introduc tion of better and more efficient stoves, and improved household ventilation (Zhang and Smith 2007). Methods of reducing airborne endotoxin concentrations will be broadly similar, but there may also be opportunities to reduce bacterial and endotoxin content of the source fuel via harvesting and/or production methods and changes to how fuel is stored. Higher cook ing temperatures are likely to degrade endotoxin, and more efficient cooking using improved stove technologies can also reduce the generation of PMbound endotoxin. A recent study has also suggested that outdoor storage of wood chips increased endotoxin content (Sebastian et al. 2006), so dry, indoor storage areas for fuel may reduce the airborne endotoxin levels when burn ing eventually takes place.
Our study raises the possibility of an important new risk factor, and preventive strat egies, for respiratory morbidity and mortal ity in the developing world. The mechanism for the association between biomass smoke exposure and infections of the lower respira tory tract in children remains unclear but is likely to be multifactorial and influenced by housing conditions, nutritional status, and other coexposures. It is possible that inhaled endotoxin, being proinflammatory, may be one contributory factor in this mechanistic pathway. Pneumonia remains one of the larg est contributors to underfive mortality, and exposure to high concentrations of airborne endotoxin may be an important risk factor for the severity of illness (Dales et al. 2006). From a public health perspective, interventions to reduce PM and endotoxin exposures gener ated from household combustion of solid fuels should be implemented as a matter of urgency.

Conclusions
This study has shown that airborne endotoxin concentrations in homes burning biomass fuels are considerably higher than those found in homes in the developed world and at levels comparable to agriculturalrelated occupations. Some homes recorded cooking period concen trations > 1,000 EU/m 3 , more than 20 times the healthbased occupational guidance limit suggested in the Netherlands. There is a need for a larger study using a standard protocol that allows further identification of the determinants of exposure in these homes. This would increase our understanding of which fuels produce the high levels. Methods to separate the influence of endotoxin concentrations from those of high airborne PM levels are also required, as are epi demiologic and intervention studies to deter mine the health effects of reducing exposure to these high endotoxin levels.