The results of this study suggest that indoor exposure to smoke from biomass fuel combustion is a risk factor for TB. The association, however, appears to be mainly with use of biomass for heating, rather than cooking. The study also strongly suggests that exposure to smoke from kerosene fuel combustion, either in stoves or in lamps, is a risk factor for TB.
Religion, income, residence outside Kaski district, vitamin consumption, a family history of TB, and not always having lived in the present house also showed statistically significant associations with TB (). Pack-years of smoking (> 8 pack-years) showed an association with TB (p
= 0.06), which did not change appreciably after adjustment. Smoking is now an established risk factor for TB (Bates et al. 2007
; Chiang et al. 2007
; Leung et al. 2004
; Slama et al. 2007
; Yu et al. 1988
). The very elevated relative risk estimate for Buddhists relative to Hindus is striking. We considered the possibility that this may have been because some Buddhists who live around Pokhara are Tibetan and reside in refugee camps. Crowded conditions in those camps could facilitate TB transmission. However, only 8 of 40 Buddhists in the study (six cases, two controls) were Tibetan refugees—an insufficient number to explain the finding. Other studies have also shown differences in TB rates between racial and religious groups, including Tibetan Buddhists (Bhatia et al. 2002
; Hill et al. 2006
; Mishra et al. 1999
; Nelson et al. 2005
; Truong et al. 1997
Before concluding that statistical associations are causal, it is important to consider alternative explanations, particularly whether study results might be a result of selection bias, information bias, or confounding in the study design, data collection, or analysis. As with all case–control studies, selection bias in the recruitment of controls is a potential concern. In this study, a systematic procedure for recruitment of all controls from inpatient and outpatient departments of MTH was used, and only one potential control refused to participate. Because most cases were recruited from the RTC, and all controls from MTH, the catchment areas for MTH and RTC might have been different. RTC patients came from a broader area, because it is a referral center for the western development region of Nepal. A higher proportion of cases (28%) than controls (6%) were from five districts other than Kaski. The Kaski district includes Pokhara city, and in general, Kaski residents are more likely to live in urban areas and to be wealthier. This could simply mean that living outside of Kaski is associated with higher exposure to TB risk factors but, alternatively, could indicate some selection bias. We adjusted for area of residence (Kaski or other districts) in the final model, but this would not necessarily have eliminated such a bias.
Another possible source of selection bias arises because we did not exclude some other, non-TB respiratory disease cases from the control group. Unfortunately, control diagnoses were not collected at the time of the study and proved impossible to obtain in retrospect, because of the limited period for which the hospital retains patient records. Because absence of TB was confirmed in controls by X-rays, we can, however, be confident that no chronic obstructive pulmonary disease or pneumonia cases were among our controls. It is possible that inclusion of respiratory disease cases among the controls could have produced a bias toward the null, if risk factors for those cases were similar to risk factors for TB.
Information bias may take the form of outcome misclassification or exposure misclassification. Because all cases were newly diagnosed with active pulmonary TB on the basis of evidence from clinical tests, and controls were also confirmed by chest X-ray and on-the-spot sputum smear testing as not having active pulmonary TB, we consider that disease misclassification is unlikely to have occurred. We obtained all the exposure data by questionnaire. Case–control studies are often considered susceptible to recall bias, in that cases may be more likely than controls to remember past exposures. Because questions asked in this study were about common exposures, however, which both cases and controls experience on a day-to-day basis, we expect recall to have been accurate and any differential recall to have been minimal. We verified the high level of accuracy of reporting of two key exposure variables (stove type and ventilation) by visiting the homes of 28 study participants. Considering this, and that there is no prevailing belief that indoor smoke exposure from biomass-burning stoves or kerosene-burning stoves or lamps is related to TB occurrence, we believe exposure misclassification is likely to be minimal. One possible limitation, however, is that we only asked about the main cooking fuel used. This might have led to some misclassification of exposure status.
The third main area of potential bias is confounding. We collected data on a much more comprehensive range of exposures than did previous studies and investigated their potential to confound the associations with fuel use. Although confounding was present, adjustment with these variables did not eliminate the key associations. There may, of course, be some residual confounding due to misspecification of the variables, and there is no way to rule out the possibility of unknown confounding factors causing the associations found. One possibility is malnutrition, for which we obtained no data and which is a known risk factor for TB. However, family income, for which we did obtain data and which is an excellent indicator of a family’s ability to feed itself, was taken into account.
A notable finding in our study was the association with biomass used as a heating fuel. This was unexpected because the study design focused on cooking-fuel use. Hence, the study population was limited to women, who generally do the cooking in Nepal. Although we collected data on history of stove and cooking-fuel use, we did not collect a comparable level of data for heating fuels and so are unable to examine heating-fuel use for evidence of an exposure–response relationship.
In hindsight, the findings with biomass as a heating and a cooking fuel make sense. Women may light a cooking fire, set the pot atop it, and leave the room, returning only periodically while cooking takes place. On the other hand, use of heating fuel involves minimization of ventilation and deliberate exposure, as the family sits around the fire. In tropical India and Africa, where several of the other TB and biomass studies have been carried out, use of heating fuel is less common than in the mid-hills of Nepal, where nighttime and winter temperatures are lower.
Our study also found the OR for TB to be high among both kerosene stove and lamp users, particularly the latter. Kerosene cooking fuel and kerosene lamp users were for the most part mutually exclusive groups. Only one of the 22 kerosene lamp users in the study used a kerosene stove. Kerosene stove users were more likely to use electricity for lighting. With one exception, as far as we are aware, no previous studies have examined a relationship between kerosene and TB (Padilla et al. 2001
). This one study, carried out in Mexico, obtained crude ORs for use of kerosene-burning stoves of 1.9 (95% CI, 0.8–4.5) for active TB and 4.4 (95% CI, 1.7–11.5) for past TB; no adjusted estimates were presented. We have been unable to find any studies where the relationship between kerosene lighting and TB has been investigated or even incidentally reported.
The question arises as to why kerosene as a cooking fuel could be a TB risk factor but not biomass cooking fuel. This could have something to do with the nature of the emissions. Biomass burning produces very obvious smoke, which may irritate the eyes and respiratory tract, encouraging avoidance behavior. Kerosene, on the other hand, has the appearance of burning more cleanly, even if it does produce substantial amounts of fine particulate matter and vapor-phase chemicals, and may not encourage the same avoidance behavior as biomass smoke. Cooks may be more likely to remain in the room while cooking with kerosene fuel. There are also likely to be differences in the toxic effects of the pollutant mixtures from the two fuels.
Kerosene is one of the main sources of cooking fuel in urban areas and lighting fuel in rural areas of developing countries, including Nepal. Therefore, if kerosene burning can be confirmed as a TB risk factor in other studies, the public health implications would be substantial. In rural areas not connected with electric power, kerosene wick lamps are burned at least 4–5 hr every day. Commonly, these lamps are homemade devices that are highly energy inefficient, with low luminosity. Simple wick kerosene lamps emit substantial amounts of smoke and particles (Schare and Smith 1995
). A study conducted in rural Malawi has shown a higher loading of particulates in alveolar macrophages in men from exposure to kerosene in lamps compared with candles, hurricane lamps, and electric lamps (Fullerton et al. 2009
). Other emissions from kerosene combustion include carbon monoxide, carbon dioxide, sulfur dioxide, nitrogen dioxide, formaldehyde, and various VOCs (volatile organic carbons) (Traynor et al. 1983
). An indoor air pollution study conducted in Bangladesh slums has shown significantly higher concentrations of benzene, toluene, xylene, hexane, and total VOCs emitted from kerosene stoves than from wood-burning stoves (Khalequzzaman et al. 2007
The use of kerosene fuel is associated with harmful effects that have been documented in a few studies. These effects include impairment of ventilatory function and a rise in blood carboxyhemoglobin in women exposed to kerosene fuel smoke (Behera et al. 1991
), and a higher incidence of acute lower respiratory infection in children in homes using KFS and BFS (Sharma et al. 1998
A causal relationship between exposure to biomass fuel smoke and TB is biologically plausible. The smoke could affect either risk of infection or risk of disease in infected people, or both, as has been shown to be the case with tobacco smoking (Bates et al. 2007
). Without knowledge of the time of infection, however, the present study cannot distinguish between the two possibilities. Inhalation of respirable particles and chemicals found in smoke from these sources generates an inflammatory response and impairs the normal clearance of secretions on the tracheobronchial mucosal surface, and may allow TB bacteria to escape the first level of host defenses, which prevent bacilli from reaching the alveoli (Houtmeyers et al. 1999
). Smoke also impairs the function of pulmonary alveolar macrophages, an important early defense mechanism against bacteria (Health Effects Institute 2002
). Alveolar macrophages isolated from the lungs of smokers have reduced phagocytic ability compared with macrophages from nonsmokers and secrete a lower level of proinflammatory cytokines (Sopori 2002
). Exposure to wood smoke in rabbits has been shown to negatively affect antibacterial properties of alveolar macrophages, such as their ability to phagocytize bacteria (Fick et al. 1984