INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: allergy; asthma; rhinitis; eczema; kerosene

There is substantial worldwide variation in the prevalence of asthma and other allergic diseases, which are generally more common in economically developed countries (1), and within the developing world are more prevalent in urban than in rural areas (2). It has therefore been suggested that factors associated with the adoption of a more affluent or urbanized lifestyle play an important role in the etiology of allergic disease.

One life-style factor associated with urbanization is the use of cleaner fuels for domestic cooking and heating. Although rural communities in poor countries still rely heavily or exclusively on biomass fuels such as wood, charcoal, leaves, or animal dung, families in urban areas are increasingly likely to use kerosene, gas, electricity, or other modern fuels. Little is known about the association between fuel use and asthma and allergy in this context. Evidence from a limited number of studies in children suggests that the use of biomass fuels may be associated with a lower risk of atopy (5), exercise-induced bronchospasm (6), and respiratory symptoms (7) compared with other fuels, but confounding by other factors associated with a traditional life-style cannot be ruled out. Several studies in developed countries have shown an adverse effect of the use of gas appliances on respiratory illness in women (8, 9) and children (7, 10), but these have tended to involve a comparison of gas with other modern fuels. Studies of the effects of other refined fossil fuels such as kerosene on asthma and allergic disease are scarce.

Jimma is a town in southwest Ethiopia where asthma has emerged as a clinical problem within the past two decades (2), suggesting that a local environmental exposure is responsible. We have investigated the effect of exposure to nontraditional, modern fuels on the risk of allergic skin sensitization and disease symptoms in a random sample of inhabitants of Jimma.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In 1996 we carried out a cross-sectional systematic survey of households in Jimma town and in three remote rural communities in the Jimma district. An interviewer-administered questionnaire completed by all members of households in a systematic sample from a random starting point provided data on demography, respiratory and allergic disease symptoms, housing, and various life-style variables. Parents responded on behalf of children younger than 7 yr. Allergen skin sensitization to D. pteronyssinus, mixed threshings, and aspergillus was measured in consenting individuals living in every fourth household. Full details of the study methods have been reported elsewhere (2). Data on cooking and heating fuels used in the home were collected in this survey, but were not entered into the database analyzed previously (2). Subjects were asked about all the fuels used in the house for cooking and heating, though specific uses of fuels were not explored. Approval for the study was granted by the Jimma Institute of Health Sciences and Nottingham City Hospital ethics committees. All participants gave verbal consent.

Because those living in the rural areas used only biomass fuels, this analysis has been restricted to the urban population in which a mixture of traditional and modern fuels was used. Data were analyzed using STATA (version 6.0; Stata Corporation, College Station, TX). The primary outcome variable was allergen skin sensitization, defined as a mean skin wheal response of 3 mm or more to D. pteronyssinus or mixed threshings (positivity to aspergillus was very rare), after subtraction of the saline control response. Secondary outcomes were questionnaire-reported wheeze in the past year, rhinitis symptoms, and eczema symptoms ever and in the past year. Modern fuel use was analyzed initially as a binary exposure variable representing reported use of any nonbiomass fuels (gas, kerosene, or electricity). Logistic regression was used to compute odds ratios, with corresponding 95% confidence intervals derived from standard errors adjusted for clustering by household, before and after control for age, sex, and socioeconomic status (based on family occupation). The potential confounding effects of other previously identified risk factors, comprising vaccinations, material of mattress (cotton, synthetic foam, grass, kapoake, mat, or other), number of rooms in house, construction style of house (material of walls, floor, and roof), crowding, insecticide use, smoking, and ethnicity (2) were also explored. Any potential confounding variable that altered the age, sex, and socioeconomic adjusted odds ratio by 0.1 or more was retained in the logistic regression model. An alternative socioeconomic status variable based on education level was also considered. Where an effect of modern fuel use was identified, the individual effects of each fuel (gas, kerosene, and electricity) were investigated using the same methods.

Sample size estimates were originally calculated to provide sufficient power to detect differences in prevalence between urban and rural areas (2). However the available sample provided 80% power to detect odds ratios in relation to nonbiomass fuel use of 1.56 for wheeze, 1.61 for rhinitis, 1.94 for eczema, ever, 2.36 for recent eczema, and 1.99 for allergen sensitization.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Questionnaire data were collected on 9844 people in the urban population, estimated to be at least 95% of those eligible, although the exact response rate was uncertain due to lack of reliable population census data. Allergen skin sensitization was measured in 2372 (approximately 96%) of the one in four subsample, including 178 aged under 5 yr who were excluded from our prevalence analysis (2) because of lower participation in this age group (approximately 85%). Characteristics of subjects are given in Table 1. Current smoking was reported by 3% of the population.

Almost all individuals surveyed (9793 of 9844, 99%) used biomass fuels, the most common being wood (93%) and charcoal (78%, Table 2). Smaller numbers used leaves or dung but always in combination with wood and/or charcoal. Modern fuels were used by 959 people (10%), again usually in combination with biomass fuels, with only 34 individuals reporting exclusive use of modern fuels. Kerosene was the most commonly used modern fuel (6%); gas and electricity were both used by 3%.

In univariate analyses, allergen skin sensitization was significantly more likely in those who used any modern fuel, with an estimated odds ratio (OR) of 1.98 (95% confidence interval [CI] 1.21 to 3.22, p = 0.01, Table 3). Following adjustment for age, sex, and socioeconomic status the odds ratio was slightly reduced to 1.78 (95% CI 1.06, 2.97) but remained significant (p = 0.03). There were also significant associations between modern fuel use and allergic disease symptoms, with adjusted odds ratios of 1.56 (95% CI 1.07 to 2.26) for recent wheeze, 2.06 (95% CI 1.46 to 2.91) for rhinitis, 2.64 (95% CI 1.70 to 4.11) for eczema ever, and 2.82 (95% CI 1.61 to 4.96) for recent eczema (Table 3). Further confounding effects by life-style variables were explored and the odds ratio for allergen skin sensitization was seen to be slightly reduced by adjusting for mattress type and material of floor, but was strengthened by adjustment for crowding and mumps vaccination. Simultaneous control for age, sex, socioeconomic status, and these four variables, in the 2122 individuals with complete data, resulted in little overall change to the magnitude of the association (OR = 1.70) though significance was lost (95% CI 0.93 to 3.11). For wheeze, control for mattress type slightly weakened the odds ratio, tuberculosis vaccination strengthened it, and addition of both to the model resulted in an odds ratio of 1.46 (95% CI 0.99 to 2.14). A number of life-style factors had small modifying effects on the odds ratios for rhinitis and eczema, but following simultaneous adjustment for those identified as important, the positive associations with modern fuel use persisted (adjusted OR = 1.71 95% CI 1.18 to 2.48 for rhinitis, 2.32 [1.44 to 3.75] for eczema ever and 2.08 [1.12 to 3.85] for recent eczema).

When relationships between specific modern fuels and allergen skin sensitization were explored in univariate analyses, significantly increased odds ratios were seen in relation to both kerosene (2.30, 95% CI 1.23 to 4.32, p = 0.01) and gas (3.63, 95% CI 2.32 to 5.68, p < 0.001) but not electricity (0.75, 95% CI 0.26 to 2.13, p = 0.60, Table 4). After controlling for age, sex, and socioeconomic status, the odds ratios for allergen skin sensitization in relation to kerosene and gas were 1.95 (95% CI 1.02 to 3.73) and 3.93 (95% CI 2.32 to 6.67), respectively; these were little changed by further controlling for identified confounders (OR = 2.06, 95% CI 1.01 to 4.20 and 3.78, 95% CI 2.05 to 6.98, respectively). Kerosene use was also seen to be significantly associated with all allergic symptom outcomes both before and after control for age, sex, and socioeconomic status (adjusted OR = 1.55 [1.01, 2.38] for recent wheeze, 2.57 [1.76, 3.75] for rhinitis, 2.99 [1.78, 5.04] for eczema ever, and 3.20 [1.62, 6.32] for recent eczema, Table 4). Further control for factors identified as important had little overall effect on the magnitude of the odds ratio for wheeze (1.50, 95% CI 0.97 to 2.31) or rhinitis (2.35, 95% CI 1.59 to 3.45); odds ratios of eczema were slightly weakened but remained significantly increased (OR for eczema ever = 2.47, 95% CI 1.46 to 4.20 and OR for recent eczema = 2.22, 95% CI 1.08 to 4.57). There were no significant associations between gas or electricity use and reported symptoms (Table 4), with the exception of eczema ever, which was more common in those who used electricity (OR adjusted for age, sex, and socioeconomic status = 2.51, 95% CI 1.29 to 4.90, p = 0.01). This relation weakened and was no longer significant following control for the identified confounders ethnicity, mattress type, mumps vaccination, material of walls, crowding, and insecticide use (adjusted OR = 1.93, 95% CI 0.92 to 4.05). Substituting education level for occupation as a measure of socioeconomic status made little difference to any of the odds ratios. There was no evidence of interaction with age in any of the associations identified.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have identified strong positive associations between allergic skin sensitization and allergic symptoms and the domestic use of modern fuels for cooking and heating, particularly kerosene and to a lesser degree gas, in an urban community of Ethiopia where the use of biomass fuels is still almost universal but a minority are also now using more modern fuels. Because there is no major industry creating outdoor air pollution in Jimma, and levels of road vehicle traffic are low, the majority of exposure to atmospheric pollution in this population is likely to occur indoors as a result of domestic fuel combustion. Most homes in Jimma are heated by open fires, burning biomass fuels in poorly ventilated rooms with no chimney and few windows. Most families cook in the main living area or in a separate but adjacent kitchen, using an open fire or clay biomass stove. Those using kerosene tend to use simple wick stoves, while the use of gas (methane) or electricity tends to involve more sophisticated appliances. Current smoking was not a major source of indoor pollution at the time of the study, being uncommon in the Jimma population and not a confounder in this study.

We used allergen skin sensitization as our primary outcome as this is an objective measure of allergy. However because skin sensitization data were available only in a subgroup of the sample population, we also investigated effects on reported symptoms of allergic disease in the whole sample. Some misclassification of symptoms attributable to other diseases and conditions prevalent in this population is inevitable, but because this is unlikely to be dependent on exposure, the result of such nondifferential misclassification is likely to be an underestimation in the magnitude of true effects on allergic outcomes.

Others have reported a lower risk of atopy or asthma in those using biomass fuels compared with other fuels (5), but have not generally explored the extent to which this might be due to confounding by some other factor associated with affluence. We controlled extensively for potential confounding effects and found that in general, the magnitude of the effects of kerosene and gas remained stable in all multivariate analyses. However the fact that the use of electricity was not associated with our outcomes indicates that confounding, particularly by high socioeconomic status, is unlikely to account for our findings for kerosene and gas, because if that were the case we would have expected all three individual modern fuels to be associated with an increased risk of disease.

One hypothesis arising from earlier reports of a reduced risk of allergy in those using biomass fuels is that exposure to these fuels protects against the development of allergy and allergic disease. There are reports of an increased risk of lower respiratory tract infections (LRIs) in early infancy associated with exposure to wood-burning stoves (11, 12) and other sources of particulates such as tobacco smoke (13), and LRIs in early life are hypothesized to reduce the risk of subsequent development of allergic disease through effects on the developing immune system (14). Our findings support an alternative explanation, because virtually all our study population were exposed to biomass fuels to some degree, and associations were not seen for each individual modern fuel as would be expected if a reduced exposure to biomass fuels among those using modern fuels were responsible. An adverse effect of exposure to refined fossil fuels such as kerosene and gas on allergy is therefore a more likely explanation, probably as a result of some component of the pollution arising from hydrocarbon combustion.

In this study kerosene was the fuel showing the most consistent effects, but only a minority of other studies have specifically examined the effect of kerosene use on allergic disease. One of these is a study of Kenyan schoolchildren that reported a small and nonsignificant increased risk of exercise-induced bronchospasm in relation to the use of kerosene for cooking (6). In contrast, another Kenyan study found no effect of kerosene use on childhood asthma, although this study involved small numbers and only the primary fuel used for cooking was considered (15). Other limited evidence that kerosene may induce and/or aggravate asthma comes from a case report of six adults and three children (16), and from a clinical evaluation study of 286 women with asthma (17). The burning of kerosene produces a number of pollutant gases such as nitrogen dioxide, carbon monoxide, and a range of volatile organic compounds (VOCs) including polyaromatic and aliphatic hydrocarbons (18), as is also true of other refined hydrocarbon fuels such as petroleum. Our findings are therefore generally consistent with those of epidemiological studies suggesting that exposure to road traffic may increase the risk of atopy (19), and human exposure studies of traffic-related pollutants showing that exposure enhances the reaction to inhaled allergen among those with established allergy (22).

We also identified a strong effect of gas on allergen skin sensitization, but not allergic disease symptoms in this study. Previous studies looking at effects of gas on allergy in developing countries are scarce, with just the previously mentioned Kenyan study by Mohamed and coworkers, which reported no increased risk of asthma in children living in homes where gas was the main cooking fuel (15). There is more extensive evidence on the effect of gas exposure in western countries, but these have all compared gas use with exposure to other modern fuels. Jarvis and coworkers reported an adverse effect of the use of domestic gas appliances compared with electricity on asthma-like symptoms in women in the United Kingdom, and positive but nonsignificant associations with hayfever and allergen sensitization (8). Others have studied the effect of gas use on respiratory illness but findings have been inconsistent (7, 9, 10, 26, 27). This may be because in developed countries personal exposure to the pollutants arising from gas is largely influenced by outdoor sources. Nitrogen dioxide is widely suspected to be the pollutant responsible for any adverse health effects of gas appliances, although hypotheses based on other by-products of gas combustion such as VOCs and carbon monoxide are equally feasible.

There are a number of possible mechanisms to explain the observed effects. Air pollutants such as NO₂ and diesel exhaust carbon particles that contain hydrocarbons have been shown to interact or bind to allergen particles increasing their allergenicity (28). Diesel exhaust particles and their associated hydrocarbons are also thought to initiate or elevate immunoglobulin E (IgE) production through a variety of mechanisms (31), and the hydrocarbon VOCs emitted from kerosene may be having a similar effect. Alternatively, pollutants released during fuel combustion may be having a direct effect on the epithelial cells; airway epithelial cells show increased permeability, damage, and inflammation when exposed in vitro and in vivo to NO₂ and diesel exhaust particles (34).

In conclusion, these findings provide evidence that the use of refined fossil fuels, particularly kerosene, for cooking and heating increases the risk of allergy and symptoms of allergic disease. Although in this population such fuels were used only by a minority and primarily as a supplementary fuel, their use is likely to become increasingly common in Ethiopia and other developing countries. We hypothesize that either direct contact with these fuels, or more likely exposure to combustion pollutants emitted during their use, is likely to be responsible, and if so may have played a major role in the emergence of allergic diseases in the developing and developed world. Further research into the mechanisms of these associations is therefore likely to be helpful in explaining and preventing the rising prevalence of asthma and allergy.