Systematic review and meta-analysis of the associations between indoor air pollution and tuberculosis

Authors


Colin Sumpter, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. E-mail: colin.sumpter@lshtm.ac.uk

Abstract

Objective

Half the world's population uses biomass fuel for their daily needs but the resultant emissions and indoor air pollution (IAP) are harmful to health. So far, evidence for a link between IAP and tuberculosis (TB) was insufficient. We report an updated systematic review due to recent increase in the evidence and growing interest in testing interventions.

Methods

Systematic search of PubMed (including Medline), CAB abstracts (through Ovid SP) and Web of Knowledge using the following search terms: ‘IAP or biomass or cooking smoke’ and ‘TB’. 452 abstracts were reviewed, and only 12 articles were deemed to be reporting the effects of IAP on TB and were taken forward to full review, and one study was added through hand search of references. Data on measures of effect of IAP on TB were extracted, and meta-analysis was carried out to estimate pooled measures of effect.

Results

Thirteen studies have reported investigating association between IAP and TB since 1996. TB cases are more likely to be exposed to IAP than healthy controls (pooled OR 1.30; 95% CI, 1.04–1.62; P = 0.02).

Conclusions

There is increasingly strong evidence for an association between IAP and TB. Further studies are needed to understand the burden of TB attributable to IAP. Interventions such as clean cook stoves to reduce the adverse effects of IAP merit rigorous evaluation, particularly in Africa and India where the prevalence of IAP and TB is high.

Abstract

Objectif

La moitié de la population mondiale utilise du combustible à base de biomasse pour leurs besoins quotidiens, mais les émissions qui en résultent et la pollution de l'air intérieur sont nocives pour la santé. Jusqu’à présent, la preuve d'un lien entre la pollution de l'air intérieur et la tuberculose était insuffisante. Nous rapportons une analyse systématique mise à jour suite à l'augmentation récente des preuves et de l'intérêt croissant des interventions pour la mesure de ce lien.

Méthodes

Recherche systématique dans PubMed (Medline y compris), résumés CAB (sur Ovide SP), et ‘Web of Knowledge’, en utilisant les termes de recherche suivants: «pollution de l'air intérieur ou biomasse ou fumée de cuisson» et «tuberculose ou TB». 452 résumés ont été examinés et seuls 12 articles considérés comme rapportant sur les effets de la pollution de l'air intérieur sur la tuberculose ont été utilisés pour une analyse complète. Une étude a été ajoutée suite à la recherche manuelle de références. Les données sur les mesures de l'effet de la pollution de l'air intérieur sur la tuberculose ont été extraites et la méta-analyse a été effectuée pour évaluer les mesures poolées de l'effet.

Résultats

Treize études rapportaient sur des études sur l'association entre la pollution de l'air intérieur et la tuberculose depuis 1996. Les cas de tuberculose sont plus susceptibles d’être exposés à la pollution de l'air intérieur que les témoins sains (OR poolés: 1,30; IC95%: 1,04 à 1,62; P = 0,02).

Conclusions

Il y a de plus en plus de preuves solides d'une association entre la pollution de l'air intérieur et la tuberculose. Des études supplémentaires sont nécessaires pour comprendre la charge de morbidité de la tuberculose attribuable à la pollution de l'air intérieur. Des interventions telles que celles utilisant des cuisinières propres pour réduire les effets néfastes de la pollution de l'air intérieur méritent une rigoureuse évaluation, en particulier en Afrique et en Inde où la prévalence de la pollution de l'air intérieur et de la tuberculose est élevée.

Abstract

Objetivo

La mitad de la población mundial utiliza la biomasa como combustible para cubrir sus necesidades diarias, pero las emisiones resultantes y la contaminación intradomiciliaria (CID) son dañinas para la salud. Hasta ahora, la evidencia que relacionaba la CID y la TB era insuficiente. Aquí reportamos una revisión sistemática y puesta al día, debido al aumento de evidencia y al creciente interés en realizar una evaluación de las intervenciones.

Métodos

Búsqueda sistemática en PubMed (incluido Medline); abstracts CAB (a través de Ovid SP); y Web of Knowledge, utilizando los siguientes términos de búsqueda: ‘contaminación aire intradomiciliario o biomasa o humo de cocinas’ 12 artículos y ‘tuberculosis o TB’. Se revisaron 452 resúmenes y de los cuales se consideró que solo reportaban los efectos de la CID sobre la TB y se les realizó una revisión completa. Un otro estudio se añadió tras una búsqueda manual de las referencias. Se extrajeron datos sobre medidas del efecto de la PID sobre la TB y se realizó un meta-análisis para calcular una medida unificada del efecto.

Resultados

Trece estudios han reportado investigar la asociación entre el CID y la TB desde 1996. Los casos de TB tienen mayor probabilidad de exposición a la CID que los controles sanos (OR unificado 1.30; IC 95%: 1.04–1.62, P = 0.02).

Conclusión

Existe una evidencia cada vez mayor de la asociación entre la contaminación del aire intradomiciliario y la TB. Se requieren más estudios para entender la carga de TB atribuible a la CID. Intervenciones, tales como las cocinas limpias, que reduzcan los efectos adversos de la CID, merecen una evaluación rigurosa, particularmente en África e India en donde la prevalencia de la CID y la TB son altas.

Introduction

Half the world's population, primarily living in low income countries, relies on biomass fuel for heating and cooking. Around 3.4 billion people are burning biomass on open fires for daily cooking. It is also estimated that 1.5 m premature deaths primarily among women and children, and 40.8 million disability adjusted life years (DALYs) are lost each year due to exposure to indoor air pollution (IAP). These estimates are based on IAP effects on pneumonia and other acute lower respiratory infections (ALRI) among children under 5 years of age and chronic obstructive pulmonary disease (COPD) and lung cancer (related to coal use) among adults (WHO 2007). Tuberculosis (TB) is not included in this regularly quoted burden of IAP due to a lack of epidemiological evidence at the time of calculation (Desai et al. 2004).

Previous systematic reviews have shown no effect of IAP on TB and have highlighted the limited quality of the available evidence (Lin et al. 2007; Slama et al. 2010). A review of studies up to 2008, which comprised three case–control and three cross-sectional studies, found no association between IAP and TB (Slama et al. 2010). Since 2008, another eight studies of the effects IAP on TB have been published, and thus, it is timely to reconsider the evidence.

Methods

Following the guiding principles laid out in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) document, a systematic search for original research of the effects of IAP on TB was conducted. Studies were considered eligible where they were published in a peer reviewed journal, and written or translated into English. Both observational and experimental studies were eligible as long as they included a TB outcome and provided a measure of effect of IAP. Reports of previous systematic reviews, policy papers, abstracts to conferences, letters to journals and economic evaluations were excluded.

Two key search concepts were used (‘IAP from biomass’ and ‘TB’) along with associated synonyms to account for differences in terminology. Synonyms used for IAP were biomass and cooking smoke; for tuberculosis, TB. These search terms are consistent with previous review approaches (Desai et al. 2004; Slama et al. 2010). The search was conducted in April 2012 in three online databases: PubMed (including Medline), CAB abstracts (through Ovid SP) and Web of Knowledge. Each previously published systematic review and original study was hand-searched for additional references.

All articles identified were entered into EndNote X3 referencing software (Thompson-Reuters 2009), and duplicates were removed. A title and abstract review was conducted to screen out ineligible articles. Articles which apparently fulfilled the criteria were reviewed in full. After this detailed review a number of papers were excluded. Figure 1 shows the PRISMA chart for this process.

Figure 1.

Flow diagram of inclusion/exclusion.

Studies were reviewed for potential bias, methodological rigour and overall quality of the approach. Relevant data were extracted, odds ratios of IAP among TB cases compared with non-TB controls were plotted using forest plots and pooled odds ratio (OR) was calculated. This was repeated for three subgroups of studies – case–control studies only, case–control studies focusing solely on women and cross-sectional studies.

Results

Initially, 452 articles were identified (Figure 1) from Web of Knowledge (188) PubMed (158) and CAB Abstracts (106). After removal of duplicates (151), review of title and abstracts resulted in the exclusion of 281 papers. Reasons for exclusion were a focus on: agriculture or ecology identified through the essential search term ‘biomass’ (122); microbiology (33); outcomes other than TB (26); efficacy of interventions at reducing exposure rather than health outcomes (12); previous reviews related to IAP (19); policy or economics (19); and mathematical modelling (6). After reviewing full manuscripts, eight were excluded because the data reported were inadequate to calculate a measure of effect of IAP on TB.

Hand search of the bibliographies of reviewed papers identified one further article; thus, a total of 13 articles were included in the systematic review.

We identified 11 case–control studies and two cross-sectional studies investigating the link between TB and IAP. Following a more detailed review, one paper described as a case–control was reclassified as cross-sectional as the reported OR for TB was based on comparison of exposure within a TB-positive control group rather than any comparison between cases and controls. Key study features are summarised in Table 1 for the 10 case–control studies and in Table 2 for three cross-sectional studies. The data presented in these studies were deemed to be of high enough quality to be included in meta-analysis.

Table 1. Case–control studies: characteristics and reported estimates of association between IAP and TB
Study country (period)Study populationMethod of ascertainment of TBMethod of selection of controlsMethod of ascertainment of exposure to IAPSample size (case:control)Analytical methodOdds ratio (95% CI)
  1. IAP, indoor air pollution; TB, tuberculosis.

Urban/rural India (April 2006–March 2007) (Pokhrel et al. 2010)>20-year-old female onlySputum smear positive for AFB

Clinic recruited non-TB.

Matched on age and area of residence

Self-reported biomass fuel use378 (126:252)Multivariate logistic regression adjusted for education, kitchen, tobacco smoking and TB in family

3.14

(1.15–8.56)

Urban/rural Nepal (July 2005–April 2007) (Kolappan & Subramani 2009)>20-year-old female onlySputum smear positive for AFBClinic recruited non-TB. Matched on age and area of residence

Self-reported biomass fuel use

Validated by home visit

375 (125:250)Multivariate logistic regression adjusted for age, religion, income, residence, literacy, house-type, tobacco smoking, family tobacco smoking, alcohol, vitamin supplements, TB family history and ventilation

1.21

(0.48–3.05)

Urban/rural China (September 2008–unknown date) (Lakshmi et al. 2012)

>15-year-old

73% men/27% female

Sputum smear positive for AFBCommunity recruited non-TB. Matched on age, sex, areaSelf-reported biomass fuel use606 (202:404)

Unadjusted for confounders

No adjusted result quoted but stated to be ‘non-significant’

0.75

(0.44–1.27)

Urban Mexico(March 1995–April 2003) (Perez-Padilla et al. 1996)>15-year-old female onlySputum smear positive for AFB. Community case finding

Community recruited non-TB.

Matched on sex, neighbourhood

Self-reported fuel use and length of exposure.

Current stove use confirmedby home visit

126 (42:84)Multivariate analysis adjusted for age, BMI, over-crowding, education and tobacco smoking

3.3

(1.06–10.30)

Urban/rural India (July 2007–March 2008) (Perez-Padilla et al. 2001)>24-year-old female onlySputum smear positive for AFBClinic recruited non-smoking non-TB UnmatchedSelf-reported biomassfuel use204 (94:109)Multivariate logistic regression adjusted for place of residence, passive smoking, overcrowding, use of separate kitchen and ventilation

0.6

(0.2–1.6)

Urban India (October 2001–October 2003) (Kan et al. 2011)>15-year-old 58% men/42% femaleSputum smear positive for AFB (82%) and CXR confirmed (18%)

Clinic recruited non-TB.

Matched on age and sex

Self-reported biomass fuel use378(189:189)Multivariate logistic regression adjusted for education, marital status, religion, household income, overcrowding, separate kitchen, alcohol consumption, smoking and chronic disease0.9(0.2–1.6)
Urban/rural India (2001–2003) (WHO 2011)>15-year-old 87% men/13% female

Sputum smear positive for AFB.

Community case finding

Community recruited non-TB.

Matched on age and sex

Self-reported biomass fuel use1530 (255:1275)Adjusted for smoking, alcohol consumption, and socio-economic status1.7 (1.0–2.9)
Urban Benin (2008) (Gupta et al. 1997)>22-year-old 65% men/35% female

Sputum smear positive for AFB

Community case finding

Community recruited non-TB.

Matched on age and sex

Self-reported biomass fuel use600 (200:400)Multivariate logistic regression adjusted for smoking, sex, alcohol use, family history of TB1.5 (0.9–2.3)
Urban Mexico (January 1998–April1999) (Garcia-Sancho et al. 2009)>15-year-old female onlySputum smear positive for AFB

Clinic recruited non-TB.

Unmatched

Self-reported current use of a woodstove for cooking833 (288:545)Multivariate logistic regression adjusted for age, sex, urban or rural residence, crowding, education, smoking, income2.2 (1.1–4.2)
Rural Malawi (November 1996–September 2001) (Behera & Aggarwal 2010)>15-year-old female onlySputum smear positive for AFB

Community recruited non-TB.

Unmatched

Self-reported biomass exposure stratified on placement of fire in wet and dry season768 (211:557)Only females included in analysis for cooking smoke exposure Multivariate logistic regression adjusted for age, sex, area and HIV0.7 (0.3–1.3)
Table 2. Cross-sectional studies: characteristics and reported estimates of association between IAP and TB
Study country (period)Study populationMethod of ascertainment of TBMethod of ascertainment of exposure to IAPSample sizeAnalytical methodOdds ratio (95% CI)
  1. IAP, indoor air pollution; TB, tuberculosis.

Urban Mexico (1996) (Gninafon et al. 2011)>40-years-old female onlySputum smear positive for AFBSelf-reported biomass exposure stratified on hours exposed (3 levels)83

Analysis looked at odds of exposureto IAP amongst controls selected as having TB (n = 83) only.

Age-adjusted only

2.27 (1.22–4.22)
Rural/urban India (1997) (Gupta et al. 1997)>15-years-old man and femaleSputum smear positive for AFB and CXR confirmedSelf-reported use of wood or cow dung as fuel in the home707

Multiple logistic regression

Age-adjusted only

2.54 (0.5–3.3)
Rural/urban India (1993) (Mishra et al. 1999)>20-years-old man and femaleSelf-reported active tuberculosis, no confirmationUse of unprocessed biomass fuel as the primary fuel for cooking in household260, 162

Multiple logistic regression

Adjusted for kitchen type, house type, crowding, age, gender, urban/rural, education, religion, caste and region

2.58 (1.98–3.37)

The adjusted OR of exposure to IAP among TB cases varied between the studies (Figure 2). Four studies showed that odds of exposure to IAP were slightly lower among TB cases, but none of these observations were statistically significant (Crampin et al. 2004; Shetty et al. 2006; Behera & Aggarwal 2010; Kan et al. 2011). Six studies showed higher exposure to IAP among TB cases, but only three studies showed a statistically significant association (Perez-Padilla et al. 2001; Garcia-Sancho et al. 2009; Lakshmi et al. 2012). The overall fixed effect OR for exposure to biomass amongst those with TB compared with those without TB is 1.30 (95% CI: 1.04–1.62, P = 0.019).

Figure 2.

Forest plot of ORs of exposure to biomass among tuberculosis (TB) case vs. non-TB controls.

The odds of exposure to IAP among female TB cases were higher (pooled OR 1.7; 95% CI 1.00–2.89) when the studies including men were excluded from the analysis (Figure 3). The pooled estimate from the two cross-sectional studies showed a stronger association between the IAP and TB (pooled OR 2.53; 95% CI 1.99–3.22; P < 0.001) though the observation from one study was not statistically significant (Figure 4).

Figure 3.

Forest plot of ORs of exposure to biomass among female tuberculosis (TB) case vs. non-TB controls.

Figure 4.

Forest plot of ORs of tuberculosis in those exposed to biomass vs. unexposed (cross-sectional).

Discussion

The pooled OR from the cross-sectional studies showed a strong effect of IAP. However, adjustment for confounding was limited to age in only two of the three studies, and thus, these observations should be interpreted with caution. At best these cross-sectional studies show that the association is a possibility; this is reflected in the recent shift towards case–control studies.

The quality of case–control studies included in this review was debatable. The ascertainment of the outcome TB was robust as it was sputum-positive TB confirmed by health professionals. However, the ascertainment of exposure to IAP was open to recall bias and misclassification. Two studies verified exposure by home visit (Garcia-Sancho et al. 2009; Pokhrel et al. 2010) but none measured exposure objectively. It is possible that respondents misclassified themselves or recalled wrongly. Many families use multiple energy sources in their homes, which limits binary classification. Only one study used any gradation (ordered categorical) measure of exposure (Crampin et al. 2004), and one study presented a more complex composite measure made up of ventilation, kitchen type, mixed-fuel use and length of exposure, although ultimately this was not used for analysis in this review to ensure consistency with other studies (Behera & Aggarwal 2010). This is a major limitation of the studies included in this meta-analysis. Ideally, the level of IAP needs to be measured using particulate matter readings in dwellings or stove monitors, a costly but possible measure of exposure, after case identification.

All studies made some attempt to adjust for potential confounders. In particular smoking, socioeconomic status and over-crowding are associated with both biomass use and TB. It is possible that these co-linear effects can never be truly removed from observational studies by statistical analysis. Two studies failed to adjust for tobacco smoking but one cited low or zero prevalence of smoking as the rationale for this (Perez-Padilla et al. 1996).

Five of the 10 case–control studies employed hospital based controls. This may have led to some selection bias by over-sampling those who seek health service, and it can be seen from sample breakdowns that groups in these studies differ, sometimes significantly. The challenge of selecting controls is particularly pronounced, as biomass appears to be pervasive in some areas, and the minority who use cleaner fuels do so apparently because they enjoy a higher socio-economic status. This may lead to overestimates of the effects observed. With three notable exceptions (Garcia-Sancho et al. 2009; Kolappan & Subramani 2009; Gninafon et al. 2011) studies also used hospital based cases (7/10 studies), which may have led to some selection bias and an underestimate of the effect of the exposure by failing to recruit those at greatest risk of exposure to IAP, people who do not seek treatment for health conditions due to lower socio-economic status.

The study populations are primarily adult women aged >15. However, the four studies that included both men and women tended to have an over representation of men in their sample. These studies are likely to have underestimated the levels of exposure in the population, as many men will be misclassified as exposed due to household ownership of a biomass stove but in reality have little or nothing to do with cooking, where the likely exposure occurs.

Although no studies were excluded solely for being in a language other than English, there may be a small risk of English language bias. China has a high burden of IAP and would be a likely source of further detailed research into the health effects, although the focus in Chinese literature is on coal rather than biomass fuel use. There is a noticeable bias in the geographical location for studies into IAP health effects and interventions: The African continent suffers a greater burden than any other region with an average across all African countries of 81% reliant on solid fuel (WHO 2007), but only two of 13 studies were conducted there.

Recently, the body of evidence on associations between IAP and TB has grown significantly. This review found several methodologically robust studies, not included in previous systematic reviews, which, when pooled, provide stronger evidence for an association between TB and IAP. Although there are some limitations in the individual studies, the pooled estimates suggest that there is an association between IAP and TB. The potential impact of any causal relationship between TB and IAP is very large, particularly in Africa and South East Asia, where the prevalence of both IAP exposure and TB is high (WHO 2007, 2011). Further research into this topic should be a high priority.

Ancillary