INT J TUBERC LUNG DIS 18(5):613–621 Q 2014 The Union http://dx.doi.org/10.5588/ijtld.13.0765

Indoor air pollution from solid fuel and tuberculosis: a systematic review and meta-analysis H-H. Lin,* C-W. Suk,† H-L. Lo,‡ R-Y. Huang,‡ D. A. Enarson,§ C-Y. Chiang†§¶ *Institute of Epidemiology and Preventive Medicine, National Taiwan University, Taipei, †Division of Pulmonary Medicine, Department of Internal Medicine, Wan Fang Hospital, Taipei Medical University, Taipei, ‡Department of Community Health, Mennonite Christian Hospital, Hualien, Taiwan; §International Union Against Tuberculosis and Lung Disease, Paris, France; ¶Department of Internal Medicine, School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan SUMMARY O B J E C T I V E : To conduct an updated systematic review and meta-analysis on the association between indoor air pollution and tuberculosis (TB). D E S I G N : We searched for English or Chinese articles using PubMed and EMBASE up to 28 February 2013. We aimed to identify randomised controlled trials and observational epidemiological studies that reported the association between domestic use of solid fuel and TB. Two reviewers independently extracted the information from included studies and assessed the risk of bias of these studies using pre-defined criteria. The effect sizes of eligible studies were pooled using a random-effects model; the heterogeneity across studies was quantified using I2 statistics. R E S U LT S : We identified 15 studies on solid fuel use

and active TB and one on solid fuel use and latent tuberculous infection. The summary odds ratios from case-control and cross-sectional studies were respectively 1.17 (95%CI 0.83 - 1.65) and 1.62 (95%CI 0.89 2.93), with substantial between-study heterogeneity (I2 56.2% and 80.5%, respectively). Subgroup analysis and meta-regression analysis did not identify any study-level factors that could explain the heterogeneity observed. C O N C L U S I O N : The level of evidence for the association between domestic use of solid fuels and TB was very low. High-quality studies are badly needed to clarify this association and to estimate the magnitude of the problem. K E Y W O R D S : coal; biomass; cooking

ONE IN TWO HOUSEHOLDS in the world use solid fuels (biomass and coal) for cooking and/or heating.1 Exposure to toxic pollutants from burning solid fuels can be hazardous, especially when the stove is inefficient and ventilation is poor.1–4 Such exposure is associated with acute lower respiratory infection in children aged ,5 years, chronic obstructive pulmonary disease and lung cancer.1–4 More than 3 million deaths and 108 million disabilityadjusted life-years could be attributable to indoor smoke from solid fuels associated with these conditions in 2010.5 Evidence on the association of exposure to combustion of solid fuels with tuberculosis (TB) is limited.6–8 The results of epidemiological studies have been inconsistent, with some reporting a positive association and others reporting none. Definitively determining this association is important, because a very high proportion of households use solid fuel in developing countries where the burden of TB is high. We conducted an updated

systematic review and meta-analysis on the association of exposure to solid fuel combustion with TB.

METHODS Search strategy and study selection The study followed the standard guidelines for systematic reviews and meta-analyses (Appendix*).9 Randomised controlled trials and observational epidemiological studies reporting the association between solid fuel-related indoor air pollution and TB were included. Exposure to solid fuel combustion was defined as the use of coal/lignite, charcoal, wood, straw/shrubs/grass, animal dung or crop residues for cooking and/or heating. Non-solid fuels, including electricity, liquefied petroleum gas, natural gas, biogas and kerosene, were considered ‘clean *The Appendix is available in the online version of this article, at http://www.ingentaconnect.com/content/iuatld/ijtld/2014/ 00000018/00000005/art00021

Correspondence to: Chen-Yuan Chiang, 111 Hsin-Long Road, Section 3, Taipei, Taiwan. Tel: (þ886) 933 723 426. Fax: (þ886) 2-2557 5584. e-mail: [email protected] Article submitted 17 October 2013. Final version accepted 6 January 2014.

614

The International Journal of Tuberculosis and Lung Disease

fuels’ and were used as the reference group for comparison. We excluded non-original publications and abstracts of conferences for which study details could not be obtained through contact with the authors. Studies were also excluded if the point estimate and 95% confidence interval (CI) of the association of interest (odds ratio [OR], risk ratio, rate ratio or hazard ratio) were not reported in, and could not be estimated from, the publication and were not accessible after contact with original authors. We performed a systematic literature review using a combination of search terms in PubMed and EMBASE up to 28 February 2013. We searched for articles published in English or Chinese. We also searched a non-public electronic database on TB literature (courtesy of H L Rieder of the International Union Against Tuberculosis and Lung Disease), which included over 30 000 articles on TB. Key search words for the database included: ’biomass fuel’, ’biomass cooking’, ’solid fuels’ ’indoor pollution’, ’risk factors’, ’predictors’, ’indicators’ and ’epidemiology’. In addition, we scanned the reference lists of identified citations or reviews. Data extraction Three reviewers (LL, RYH and HHL) performed the literature search and selection. Two reviewers (LL and RYH) screened the titles and abstracts from the electronic database to exclude non-relevant papers. Full-text papers identified through screening were reviewed by two reviewers (LL and RYH) for consideration based on inclusion and exclusion criteria. Any disagreement between the two reviewers was resolved by discussion with a third reviewer (HHL). For every eligible study, two reviewers (HHL and CWS) independently extracted detailed information on important study characteristics and study results using a standard data collection form. The results of data collection were compared and any differences were resolved through discussion with a third reviewer (CYC). Assessment of risk of bias in included studies We used the risk of bias approach proposed by the Cochrane collaboration for quality assessment of included studies.10 We developed a six-item tool to assess risk of bias in case-control studies and a fiveitem tool for cross-sectional studies. We evaluated adequacy in the selection of study subjects, measurement of exposure (exposure to solid fuel combustion) and outcome (TB), and adjustment of potential confounders. For exposure assessment, we considered the following to be low risk of bias: 1) direct measurement of exposure to pollutants at individual level; 2) measurement of pollutant concentration in the environment; 3) reporting type of fuel use plus taking into account any of the following items in

exposure assessment and data analysis: duration of exposure, cooking site, type of stove use or ventilation of cooking place. All else was considered high risk of bias. Age, sex and socio-economic status (SES) were considered as the most important confounders associated with the use of solid fuel for cooking and/ or heating. We did not sum across the items to obtain a summary score, but reported the items separately for each study included.10 Statistical analysis Separate meta-analyses were conducted for casecontrol and cross-sectional studies, as the two study designs were considered incomparable. We used OR as the measure of association. We assessed the heterogeneity of included studies visually using forest plot and statistically using the I2 statistic and the v2 test of the Q statistic. We used Dersimonian and Laird random-effects models to pool the ORs across studies with each study design, as the true effect sizes are likely to vary across observational studies due to various sources of bias. We performed subgroup analysis and meta-regression analysis to examine whether the summary effect size varied substantially by important domains of sources of bias. We also examined whether the observed heterogeneity between studies could be explained by these important study-level factors. We used funnel plot and contourenhanced funnel plot to visually assess small-study bias. Due to the small number of studies in each study design we did not apply statistical tests to detect small-study bias.11 All statistical analyses were conducted using Stata 11.1 (StataCorp, College Station, TX, USA).

RESULTS We identified 937 potentially relevant records from public databases and 6433 additional records from non-public databases. Of the 6083 records obtained after removing duplicates, 6002 were excluded by screening the title and abstract. The full text of the remaining 81 articles was obtained for further assessment; 16 of these met the study inclusion criteria and were included in the review (Figure 1). The 16 studies included 269 923 participants. Sample size ranged from 126 to 260 162 participants. Fifteen of the studies focused on TB disease and one on latent tuberculous infection (LTBI) (Table 1).12–27 Among the 15 studies on TB disease, 10 were case-control studies and five were cross-sectional. The study on LTBI was a cross-sectional study. Six studies included only females, and 10 studies included both sexes. Two studies separately reported the association between cooking and heating;12,20 for these studies we used the OR of cooking in the statistical analysis (e.g., pooling). Two thirds of the studies were published after 2009, and all studies

Indoor air pollution and TB

Figure 1

615

Flow diagram of study selection.

were conducted in low- and middle-income countries. In the assessment of risk of bias, eight of the 10 case-control studies were deemed to have a high risk of bias for exposure assessment because only the type of fuels used was employed as a proxy to classify whether individuals were exposed to combustion of solid fuels (Appendix Figure, A).12–14,16,18-21 A few studies assessed type of stove use, cooking site and ventilation of the cooking place, but did not take these factors into consideration in the classification of exposure and were thus also considered to have a high risk of bias for exposure assessment. One study was classified as having an uncertain risk of bias for exposure assessment because those who had more than 20 years of exposure to smoke from biomass fuels were compared with those who had not; however, the reason for using 20 years to dichotomise exposure was not clearly justified.15 Concerning the selection of study population, six of the 10 casecontrol studies had a high risk of bias in control selection as non-population-based controls were used.12–14,16,18,21 Lastly, two studies did not adjust for age,14,21 and four studies did not adjust for any measure of SES in the analysis (Appendix Figure, A).17,19–21 The quality of the six cross-sectional studies in this review was also low (Appendix Figure, B). All six studies had a high risk of bias for exposure assessment. Four of the six studies had high risk of

bias for outcome assessment because the diagnosis of TB was not based on bacteriological results or standardised criteria.23,24,26,27 Only two of the six studies simultaneously adjusted for age, sex and any measure of SES.22,23 One study used a non-random sample from the source population.24 In the 10 case-control studies on solid fuel use and TB disease, the reported OR ranged from 0.60 to 3.30, with substantial heterogeneity across studies (I2 56.2%, P value from heterogeneity test 0.032). The summary OR (1.17, 95%CI 0.83-1.65) from the random-effects model, which described the mean of distribution of true effects, revealed a modest but statistically insignificant association (Figure 2). The estimated predictive interval (0.43-3.19), which described the distribution of true effect sizes in the random effects model, further indicated the uncertainty of the association (Figure 2). Subgroup analysis and meta-regression analysis did not reveal significant effect size modification by any of the study characteristics considered in the review (Table 2). The OR from studies restricted to females (1.63, 95%CI 0.74-3.57) was slightly higher than OR from studies including both sexes (1.02, 95%CI 0.72, 1.45), although the difference was not significant (P ¼ 0.30). Visual inspection of the funnel plot and contour-enhanced funnel plot did not reveal any significant small-study bias (Figure 3A and B). For cross-sectional studies on solid fuel use and TB disease, the reported OR ranged from 0.17 to 2.58,

616

The International Journal of Tuberculosis and Lung Disease

Table 1

Study characteristics of included studies

Author, country, year, reference Case-control studies Pokhrel, Nepal, 201012

Population/setting Age 20 ~ 65 years, 100% female 125 new TB cases 250 hospital-based controls

Exposure*

Outcome assessment

Adjusted variables

Findings

Fuel type, stove Sputum-positive type, location of PTB kitchen, ventilation in the kitchen

Age, religion, income, Fuel stove (gas as residence locality, reference): residence district, literacy, Biomass aOR 1.21, type of present house 95%CI 0.48-3.05 construction, always lived Kerosene aOR in the present house, 3.36, 95%CI 1.01– pack-years of smoking, 11.22 number of family Heating (no members who smoked heating fuel or indoors, alcohol electricity as consumption, taking reference): vitamin supplements, Biomass/coal/ family history of TB and kerosene aOR ventilation in the kitchen 3.45, 95%CI 1.448.27 Use of biomass Newly diagnosed Education, type of kitchen, Reference: liquid fuel: cow dung, sputumsmoking tobacco and TB in petroleum gas wood grass, positive PTB family members, age and Biomass fuel aOR crop residue or sex 3.14, 95%CI 1.15coal 8.56, P ¼ 0.025 Type of fuel: Sputum smearStatus of kitchen in house, Biomass aOR 0.6, liquefied positive PTB adequacy of ventilation, 95%CI 0.2-1.6 petroleum gas, smoke in room while Kerosene aOR 0.1, biomass, cooking, respiratory 95%CI 0.0–1.3 kerosene, symptoms while cooking, Mixed aOR 0.5, mixed fuel fuel type for cooking 95%CI 0.2-1.1

Lakshmi, India, 201213

Adult female 126 incident TB cases 252 controls from same clinic

Behera, India, 201014

100% female adult Mean age: - cases: 34.3613.1 years - controls: 36.7612.8 years 95 TB cases 109 healthy females as control 100 % female Exposure to Smear-positive Age, body mass index, one Reference: nonMean age: biomass smoke and/or cultureroom household exposure to - cases: 42.9618.3 .20 years positive PTB crowding, years of formal biomass smoke years education and tobacco Crude OR 1.4, - controls: 39.9615 use 95%CI 0.6-3.1, P years ¼ 0.4 42 TB cases aOR 3.3, 95%CI 84 controls selected 1.06–10.30, P ¼ randomly from 0.03 neighbourhood Age 15 years, 58% Use of biomass Newly diagnosed Age, sex, education, income, Reference: gas/ men and 42% women fuels (wood, PTB by smear persons/room, separate electric 189 newly diagnosed coal, cow dung) or CXR kitchen, alcohol, smoking, Crude OR 1.8, PTB cases chronic disease 95%CI 1.10-2.90, Relatives P ¼ 0.02 accompanying non-TB aOR 0.9, 95%CI in-patients in hospital 0.46-1.76, P ¼ as controls 0.75 598 adult cases and 992 Cooking fire Sputum smearAge, sex, region, HIV Reference: never or community based (wood) or culturerare exposure controls exposure positive PTB Mild exposure: graded by aOR 0.7, 95%CI cooking indoors 0.3-1.5 or outdoors in Moderate dry and wet exposure: aOR 0.6, season 95%CI 0.3-1.3 Heavy exposure: aOR 0.6 95%CI 0.3-1.1, P ¼ 0.47 Age 20 years, 49% Use of biomass Sputum smearAge, sex, urban or rural Reference: nonmale in control group, stove or cultureresidence, crowding, level exposure to 56% male in TB group positive PTB of education, smoking and biomass smoke and 59% male in old income, state of birth, Crude OR 5.2, TB group INER socio-economic 95%CI 3.1–8.9 288 cases classification aOR 2.2, 95%CI - 277 smear-positive 1.1–4.2 - 154 culture-positive - 145 both positive 545 patients with ear, nose, throat ailments as controls

Garcia-Sancho, 2009, Mexico15

Shetty, 2006, India16

Crampin, 2004, Malawi17

Perez-Padilla, 2001, Mexico18

Indoor air pollution and TB

Table 1

617

(continued)

Author, country, year, reference Gninafon, 2011, Benin19

Kan, 2011, China20

Patra, 2012, India21

3 Population/setting

Exposure*

Gupta, 1997, India24

Du Preez, 2011, South Africa25

Adjusted variables

Findings

Age 15 years, 33% female 200 TB cases 400 neighbourhood controls matched by age and sex Age 15 years 202 cases (27.2% female) 404 neighbourhood controls (28.2% female) matched by sex and age

Use of solid fuel for cooking

New smearpositive PTB

Age, sex, ever smoking, alcohol, family history of TB

Reference: non-solid fuel for cooking aOR 1.4, 95%CI 0.7-2.7, P ¼ 0.37

Use of solid fuels

Smear-positive PTB

Age, sex, ever smoking, alcohol, family history of TB

Age: 0~14 years, 200 cases, 200 controls from among outpatients of public general hospital

Use of biomass fuel

Newly diagnosed None PTB and extrapulmonary TB

Reference: non-solid fuel for cooking and heating aOR for cooking: 1.08, 95%CI 0.62-1.87, P ¼ 0.78 For heating: aOR 1.04, 95%CI 0.542.02, P ¼ 0.90 Reference: nonbiomass fuel use Crude OR: 0.76, 95%CI 0.508-1.138, P ¼ 0.218

Cross-sectional studies Age 15 years, male/ Use of biomass Kolappan, 2009, female ratio: 6.5:1 fuel India22 255 cases 1275 controls selected in same village/unit Mishra, 1999, India23

Outcome assessment

Smear- or culturepositive TB

260 162 individuals Use of wood or History of PTB included from India’s dung vs National Family Health charcoal, coal/ Survey coke/lignite, kerosene, electricity, petroleum, gas, or biogas 707 individuals; age 16 Use of wood, coal, PTB diagnosed years; 30% female, cow dung cakes using CXR, 77% from rural areas symptoms, clinical and sputum examinations Age: 3 months - 15 years, Household TST induration 51.5% male biomass fuel 10 mm 196 cases exposure 196 neighbourhood controls

Mengersen, 2011, Lao People’s Democratic Republic26

100% female 388 women and 480 children

Saha, 2011, India27

442 TB cases Biomass fuel 442 non-TB subjects in wood, cattle an Indian village dung

Age, sex, smoking, alcohol, household standard of living index

Reference: nonbiomass fuel Crude OR: 2.9, 95%CI 1.8-4.7, P , 0.001 aOR: 1.7, 95%CI 1.0–2.9, P ¼ 0.04 Separate kitchen, housing Reference: cleaner type, crowding, age, sex, fuels urban or rural residence, Crude OR: 3.56, education, religion, caste/ 95%CI 2.82-4.50 tribe and geographic aOR: 2.58, 95%CI region 1.98-3.37 Age

Reference: coal/ kerosene aOR 2.54, 95%CI 1.07-6.04

Previous anti-tuberculosis treatment, SES, ethnicity, age and degree of M. tuberculosis exposure

Reference: nonbiomass fuel exposure OR: TST 10 mm aOR 1.20, 95%CI 0.39-3.72, P ¼ 0.75 Reference: ,1 h/day of time spent close to a fire OR Time spent close to a fire: 6 h/day vs. ,1 h/ day: 0.17, 95%CI 0.005-5.80 Reference: nonbiomass fuel use aOR 0.84, 95%CI 0.54-1.31

Time spent close Self-reported PTB None to a fire was classified as .6 h, 4~6 h, 1~3 h, ,1 h

TB diagnosed Age, SES, smoking, other based on concomitant disease, clinical, residing in mud house radiological and bacteriological findings

*Solid fuels: coal/lignite, charcoal, wood, straw/shrubs/grass, animal dung and agricultural crop residue; non-solid fuels included electricity, liquefied petroleum gas, natural gas, biogas and kerosene; biomass fuels: wood, dung and crop residues; non-biomass fuels: kerosene, liquefied petroleum gas and electric stove. TB ¼ tuberculosis; PTB ¼ pulmonary TB; aOR ¼ adjusted odds ratio; CI ¼ confidence interval; CXR ¼ chest X-ray; INER ¼ Instituto Nacional de Enfermedades Respiratorias; TST ¼ tuberculin skin test; SES ¼ socio-economic status.

618

The International Journal of Tuberculosis and Lung Disease

Figure 2 Forest plot of included studies on indoor air pollution and TB disease. SE ¼ standard error; CI ¼ confidence interval; TB ¼ tuberculosis.

with much heterogeneity across studies (I2 80.5%, P value from heterogeneity test ,0.001). The pooled OR using a random-effects model was 1.62 (95%CI 0.89-2.93) (Figure 2). We were not able to find a significant effect size modification by any study characteristics considered, possibly because of the small number of studies (Table 3). We did not present a funnel plot because visual inspection of the plot

was considered to be non-informative for a small study number.28 There was only one study on solid fuel use and LTBI.25 In this cross-sectional survey of healthy South African children, which reported that exposure to biomass fuel was associated with a non-significant risk of LTBI. Using 5 mm, 10 mm and 15 mm as the cut-off for the tuberculin skin test, the ORs were

Table 2 Subgroup analysis and meta-regression analysis on solid fuel use and tuberculosis disease: case-control studies Study characteristics Total (n ¼ 10)

Summary estimate

95%CI

Heterogeneity v2

I2 %

s2

1.17

0.83–1.65

20.57

56.2

0.161

Control selection (P value for difference:* 0.99) High ROB (n ¼ 6) 1.19 Low ROB (n ¼ 4) 1.17

0.73–1.93 0.67–2.05

13.13 7.42

61.9 59.6

0.208

Exposure assessment (P value for difference: 0.21) High or unclear ROB (n ¼ 9) 1.27 ROB (n ¼ 1) 0.60

0.89–1.81 0.31–1.15

17.21 NA

53.5 NA

0.136

Adjustment for age and sex (P value for difference: 0.15) High ROB (n ¼ 2) 0.74 Low ROB (n ¼ 8) 1.35

0.51–1.07 0.92–1.99

0.17 14.97

0 53.2

0.126

Adjustment for SES (P value for difference: 0.15 ) High ROB (n ¼ 4) 0.89 Low ROB (n ¼ 6) 1.53

0.64–1.23 0.91–2.58

4.19 10.60

28.3 52.8

0.099

Sex composition of study population (P value for difference: 0.30) Female only (n ¼ 4) 1.63 Both sex (n ¼ 6) 1.02

0.74–3.57 0.72–1.45

7.06 10.57

57.5 52.7

0.145

*Obtained from meta-regression analysis. CI ¼ confidence interval; ROB ¼ risk of bias; NA ¼ not available; SES ¼ socio-economic status.

Indoor air pollution and TB

619

Figure 3 A) Funnel plot and B) contour-enhanced funnel plot for case-control studies of solid fuel use and TB disease. RR ¼ risk ratio.

respectively 1.36 (95%CI 0.43-4.31), 1.20 (95%CI 0.39-3.72) and 1.09 (95%CI 0.32-3.69).

DISCUSSION This updated systematic review included a substantially expanded group of published papers on indoor air pollution and TB; the additional studies have added more information on this important but previously inconclusive public health topic. After a comprehensive review of the currently available literature, we found that the quantitative results from available studies and the quality of these studies did not provide strong evidence for a positive

association between domestic use of solid fuels and TB. In a recent systematic review, the authors identified 10 case-control and three cross-sectional studies on the association between indoor air pollution and TB.8 The authors reported a summary OR of 1.30 (95%CI 1.04-1.62, P ¼ 0.029), and concluded that there was ‘increasingly strong evidence’ that indoor air pollution was associated with TB.8 However, despite the presence of substantial heterogeneity among case-control studies (I2 52%), the authors pooled across the studies using a fixed-effect model, which might not be appropriate. One study was classified as a case-control study in the previous

620

The International Journal of Tuberculosis and Lung Disease

Table 3

Subgroup analysis and meta-regression analysis on solid fuel use and tuberculosis disease: cross-sectional studies Summary estimate

95%CI

Heterogeneity v2

I2 %

s2

1.62

0.89–2.93

20.52

80.5

0.302

Sample selection (P value for difference:* 0.55) High ROB (n ¼ 1) 2.54 Low ROB (n ¼ 4) 1.44

1.07–6.04 0.71–2.93

NA 20.06

NA 85.0

0.305

0.69–3.52 1.00–2.90

20.33 NA

85.2 NA

0.385

0.43–3.24 1.52–3.30

5.99 1.89

66.6 47.0

0.189

Adjustment for SES (P value for difference: 0.88 ) High or unclear ROB (n ¼ 2) 1.15 Low ROB (n ¼ 3) 1.57

0.10–12.84 0.77–3.18

2.13 18.28

53.5 89.1

0.318

Sex composition of study population†(P value for difference: 0.31) Female only (n ¼ 1) 0.17 Both sex (n ¼ 4) 1.72

0.01–5.79 0.95–3.10

NA 18.72

NA 84.0

0.235

Study characteristics Total (n ¼ 5)

Outcome assessment (P value for difference: 0.93) High ROB (n ¼ 4) 1.55 Low ROB (n ¼ 1) 1.70 Adjustment for age and sex (P value for difference: 0.36) High or unclear ROB (n ¼ 3) 1.17 Low ROB (n ¼ 2) 2.24

*Obtained from meta-regression analysis. † Mishra et al. reported odds ratio for both sexes and for females only.23 We used the odds ratio from both sexes in the analysis. CI ¼ confidence interval; ROB ¼ risk of bias; NA ¼ not available; SES ¼ socio-economic status.

review, but was considered by our group as a crosssectional study because it was based on a TB prevalence survey.22 To account for the betweenstudy variation, we used a random-effects model to pool across studies. The confidence interval of the summary OR was wider in our study (OR 1.17, 95%CI 0.83-1.65), and the P value of summary OR was non-significant (P ¼ 0.36). Our assessment of risk of bias revealed that exposure assessment was generally poor. In most of the studies, the exposure to indoor air pollution was based on self-report of the type of fuel used, while the exposure of interest was the actual exposure of the individuals to the pollution from solid fuel combustion. The high risk of bias for exposure assessment might not necessarily bias the result toward the null, as this misclassification of exposure, even if it was non-differential with respect to TB status, might result in an over- or under-estimation of the association.29 Moreover, all of the casecontrol studies in this review were retrospective studies; it was therefore still possible that TB cases might differentially recall the type of fuel used compared to controls. We were not able to empirically assess the impact of poor exposure assessment in this review because it was present in almost all studies. In the subgroup analysis, we noted that casecontrol studies that were restricted to females tended to report a higher OR than those from both sexes, although the difference was not statistically significant (P ¼ 0.30). We also found that more than half of the casecontrol studies used non-population-based controls. As solid fuel use may result in illnesses other than TB, we expect that exposure to solid fuel combustion would be overestimated in hospital-based controls, causing an underestimate of the association. In the

subgroup analysis and meta-regression, however, the summary OR from population-based and non-population-based case-control studies was not significantly different (P ¼ 0.99). In terms of confounding, several studies did not adjust for basic demographic factors or any SES measure. Poverty is a known risk factor for TB, and low-income families are more likely to use solid fuel for cooking or heating than high-income families. Confounding by SES would therefore result in a spurious positive association between indoor air pollution and TB. However, we did not find evidence of effect size modification by whether the study adjusted for demographic factors or SES. It should be noted that our subgroup analysis might not have sufficient statistical power to detect a small to moderate difference due to the small number of studies. Moreover, the subgroup analysis was in itself an observational analysis, and results can be confounded by other study-level factors. Clearly, further high-quality studies are needed. Randomised controlled trials and cohort studies could provide more definite results on this association.30 However, the outcome (TB) is sufficiently rare to require very large and prolonged studies whose feasibility would be challenging. If a casecontrol study is to be conducted, population-based controls would be preferred over non-populationbased controls. Inadequate assessment of exposure was a major issue identified. Concentration of pollutants from burning solid fuels depends on the type of stove used, the location of the burning and the ventilation of the site where the burning takes place. These factors can be used to quantify the relative concentration of the pollutants. Furthermore, duration of exposure should also be taken into account in exposure assessment and statistical analysis. If resources permit, direct measurement of

Indoor air pollution and TB

exposure to pollutants at the individual level or measurement of concentration of pollutants in the environment would allow a clearer quantification of intensity of exposure. Lastly, in an observational study (cohort/case-control/cross-sectional), confounding by SES should be properly addressed in the study design and analysis. The lack of high-quality evidence in this review does not provide evidence for a null association; neither does it refute the public health implications of this association. Given the high prevalence of domestic use of solid fuel in developing countries where TB is most concentrated, high-quality studies are badly needed to clarify the association between indoor air pollution and TB and to estimate the impact of the problem.1

13

14

15

16

17 Conflict of interest: none declared. 18

References 1 Torres-Duque C, Maldonado D, Perez-Padilla R, Ezzati M, Viegi G. Biomass fuels and respiratory diseases: a review of the evidence. Proc Am Thorac Soc 2008; 5: 577–590. 2 Kurmi O P, Arya P H, Lam K B, Sorahan T, Ayres J G. Lung cancer risk and solid fuel smoke exposure: a systematic review and meta-analysis. Eur Respir J 2012; 40: 1228–1237. 3 Kurmi O P, Semple S, Simkhada P, Smith W C, Ayres J G. COPD and chronic bronchitis risk of indoor air pollution from solid fuel: a systematic review and meta-analysis. Thorax 2010; 65: 221–228. 4 Dherani M, Pope D, Mascarenhas M, Smith K R, Weber M, Bruce N. Indoor air pollution from unprocessed solid fuel use and pneumonia risk in children aged under five years: a systematic review and meta-analysis. Bull World Health Organ 2008; 86: 390-398C. 5 Lim S S, Vos T, Flaxman A D, et al. A comparative risk assessment of burden of disease and injury attibutable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 2012; 380: 2224–2260. 6 Slama K, Chiang C-Y, Hinderaker S G, Bruce N, Vedal S, Enarson D A. Indoor solid fuel combustion and tuberculosis: is there an association? Int J Tuberc Lung Dis 2010; 14: 6–14. 7 Lin H H, Ezzati M, Murray M. Tobacco smoke, indoor air pollution and tuberculosis: a systematic review and metaanalysis. PLOS MED 2007; 4: e20. 8 Sumpter C, Chandramohan D. Systematic review and metaanalysis of the associations between indoor air pollution and tuberculosis. Trop Med Int Health 2013; 18: 101–108. 9 Liberati A, Altman D G, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. PLOS MED 2009; 6: e1000100. 10 Higgins J P T, Altman D G. Assessing risk of bias in included studies. In: Higgins J P T, Green S, eds. Cochrane handbook for systematic reviews of interventions. Chichester, UK: Cochrane Collaboration and John Wiley & Sons Ltd, 2008. 11 Sterne J A, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature. J Clin Epidemiol 2000; 53: 1119–1129. 12 Pokhrel A K, Bates M N, Verma S C, Joshi H S, Sreeramareddy

19

20

21

22

23

24

25

26

27

28

29

30

621

C T, Smith K R. Tuberculosis and indoor biomass and kerosene use in Nepal: a case-control study. Environ Health Perspect 2010; 118: 558–564. Lakshmi P V, Virdi N K, Thakur J S, Smith K R, Bates M N, Kumar R. Biomass fuel and risk of tuberculosis: a case-control study from Northern India. J Epidemiol Community Health 2012; 66: 457–461. Behera D, Aggarwal G. Domestic cooking fuel exposure and tuberculosis in Indian women. Indian J Chest Dis Allied Sci 2010; 52: 139–143. Garcia-Sancho M C, Garcia-Garcia L, Baez-Saldana R, et al. Indoor pollution as an occupational risk factor for tuberculosis among women: a population-based, gender oriented, casecontrol study in Southern Mexico. Rev Invest Clin 2009; 61: 392–398. Shetty N, Shemko M, Vaz M, D’Souza G. An epidemiological evaluation of risk factors for tuberculosis in South India: a matched case control study. Int J Tuberc Lung Dis 2006; 10: 80–86. Crampin A C, Glynn J R, Floyd S, et al. Tuberculosis and gender: exploring the patterns in a case control study in Malawi. Int J Tuberc Lung Dis 2004; 8: 194–203. Perez-Padilla R, Perez-Guzman C, Baez-Saldana R, TorresCruz A. Cooking with biomass stoves and tuberculosis: a case control study. Int J Tuberc Lung Dis 2001; 5: 441–447. Gninafon M, Ade G, A¨ıt-Khaled N, Enarson D A, Chiang C Y. Exposure to combustion of solid fuel and tuberculosis: a matched case-control study. Eur Respir J 2011; 38: 132–138. Kan X, Chiang C Y, Enarson D A, Chen W, Yang J, Chen G. Indoor solid fuel use and tuberculosis in China: a matched case-control study. BMC Public Health 2011; 11: 498. Patra S, Sharma S, Behera D. Passive smoking, indoor air pollution and childhood tuberculosis: a case control study. Indian J Tuberc 2012; 59: 151–155. Kolappan C, Subramani R. Association between biomass fuel and pulmonary tuberculosis: a nested case-control study. Thorax 2009; 64: 705–708. Mishra V K, Retherford R D, Smith K R. Biomass cooking fuels and prevalence of tuberculosis in India. Int J Infect Dis 1999; 3: 119–129. Gupta B N, Mathur N, Mahendra P N, Srivastava A K, Swaroop V, Agnihotri M S. A study of household environmental risk factors pertaining to respiratory diseases. Energy Environ Monitor 1997; 13: 61–67. du Preez K, Mandalakas A M, Kirchner H L, et al. Environmental tobacco smoke exposure increases Mycobacterium tuberculosis infection risk in children. Int J Tuberc Lung Dis 2011; 15: 1490–1496, i. Mengersen K, Morawska L, Wang H, et al. The effect of housing characteristics and occupant activities on the respiratory health of women and children in Lao PDR. Sci Total Environ 2011; 409: 1378–1384. Saha A, Sharma Y K, Kulkarni P K, Saiyed H N. Risk of tuberculosis and fuel use: a population study. Occup Environ Med 2011; 68: 934. Terrin N, Schmid C H, Lau J. In an empirical evaluation of the funnel plot, researchers could not visually identify publication bias. J Clin Epidemiol 2005; 58: 894–901. Dosemeci M, Wacholder S, Lubin J H. Does nondifferential misclassification of exposure always bias a true effect toward the null value? Am J Epidemiol 1990; 132: 746–748. Smith K R, McCracken J P, Weber M W, et al. Effect of reduction in household air pollution on childhood pneumonia in Guatemala (RESPIRE): a randomised controlled trial. Lancet 2011; 378: 1717–1726.

Indoor air pollution and TB

i

APPENDIX

Figure Risk of bias in included studies. SES ¼ socio-economic status.

Study protocol Approximately half of the world’s households use solid fuels such as dung, crop residues and coal for the purpose of cooking and heating. Exposure to toxic pollutants from these sources can be hazardous, especially when the ventilation conditions of the household are poor. The use of solid fuels has been associated with acute lower respiratory infection in children aged 5 years, chronic obstructive pulmonary disease and lung cancer (only for use of coal). Using these outcomes, it was estimated that more than 1.6 million deaths and 38.5 million disability-adjusted life-years, or about 3% of the global burden of disease, could be attributable to indoor smoke from solid fuels in 2000.1 Two systematic reviews on exposure to combustion of solid fuel and tuberculosis (TB) have been published previously.1,2 Although some articles reported a positive association between TB and indoor solid-fuel combustion, the researchers concluded that evidence supporting the relationship between TB and indoor solid-fuel combustion is still

inconclusive, partly due to the small number of studies and concerns about the quality of the studies identified. Several studies on the association between indoor solid-fuel combustion and TB have been published since the publication of the previous systematic review. We therefore plan to carry out an updated systematic review and meta-analysis on the effect of exposure to solid-fuel combustion on TB. Objective The aim of the study was to conduct an updated systematic review and meta-analysis on the effect of household exposure to combustion of solid fuel for TB. Methods

Criteria for considering studies in the review TYPES OF STUDIES We will include randomised controlled trials and observational epidemiological studies, including cohort studies, case-control studies and cross-sectional studies that reported on the specific exposures and outcomes of interest (see below).

ii

The International Journal of Tuberculosis and Lung Disease

TYPES OF EXPOSURES Exposure to household solidfuel combustion comprise the use of 1) coal/lignite, 2) biomass fuels including wood, straw/shrubs/grass, animal dung, crop residues and charcoal for heating or cooking. Non-solid fuels included electricity, liquefied petroleum gas, natural gas, biogas and kerosene, which will be considered as ‘clean fuels’ and used as the reference group of comparison. Articles that primarily measured the association between TB and tobacco smoking, exposure to secondhand smoke, moulds and occupational exposure unrelated to solid-fuel combustion will be excluded. TYPES OF OUTCOME MEASURES Relevant outcome measures of TB include latent tuberculous infection (LTBI) diagnosed by tuberculin skin test or interferon-gamma release assay (IGRA), active TB disease and TB mortality. EXCLUSION CRITERIA We will exclude non-original publications and abstracts of conferences for which the details of the study could not be obtained by contacting the authors. Studies will be excluded if the point estimate and its 95% confidence interval of the association of interest (odds ratio [OR], risk ratio, rate ratio or hazard ratio [HR]) are not reported in and cannot be estimated from the publication and are not accessible after contact with original authors.

Search methods for the identification of studies ELECTRONIC SEARCHES We will perform a systematic literature review using a combination of search terms using PubMed and EMBASE up to 28 February 2013 (Table A.1). We will search for articles with no Table A.1 Search strategy and terms used to identify studies on indoor air pollution and tuberculosis in electronic databases For PubMed MeSH term search (‘Tuberculosis’[MeSH] AND ‘Air pollution, indoor’[MeSH]) OR (‘Tuberculosis’[MeSH] AND ‘Biomass’[MeSH]) OR (‘Tuberculosis’[MeSH] AND ‘Fuel oils’[MeSH]) Direct keyword search (‘tuberculosis’AND’indoor air pollution’) OR (‘tuberculosis’ AND ‘cooking fuel’) OR (‘tuberculosis’ AND ‘biomass’) OR (‘tuberculosis’ AND ‘solid fuel’) OR (‘tuberculosis’ AND ‘coal’) For EMBASE (‘tuberculosis’AND’indoor air pollution’) OR (‘tuberculosis’ AND ‘cooking fuel’) OR (‘tuberculosis’ AND ‘biomass’) OR (‘tuberculosis’ AND ‘solid fuel’) OR (‘tuberculosis’ AND ‘coal’)

limitation of publication type or language. We will also search a non-public electronic database on TB literature (courtesy of H L Rieder of the International Union Against Tuberculosis and Lung Disease, Paris, France), which includes over 29 000 articles on TB and is updated on a regular basis. Key words of search for the database include: ’biomass fuel’, ’biomass cooking’, ’solid fuels’ ’indoor pollution’, ’risk factors’, ’predictors’, ’indicators’ and ’epidemiology’. SEARCHING OTHER RESOURCES We will scan the reference lists of identified citations or reviews. To the knowledge of the investigators, researchers who had conducted studies of similar topic with unpublished data will be contacted for any additional studies.

Data collection and analysis SELECTION OF STUDIES Two reviewers will independently perform literature search and selection. The reviewers will screen the papers identified from the electronic database by title and abstract. Full texts of papers identified through screening will be reviewed by the two reviewers for consideration based on the inclusion and exclusion criteria (see Criteria for considering studies in the review). Disagreement between the two reviewers will be resolved by discussion with a third senior reviewer. If a study is reported in more than one paper, we will only include the paper with the more detailed study information. DATA EXTRACTION AND MANAGEMENT For every eligible study, two reviewers will independently collect detailed information on important study characteristics and results using a standard data collection form (see Table A.2). The results of data collection will be compared and any differences will be resolved through discussion with a third reviewer. ASSESSMENT OF RISK OF BIAS IN INCLUDED STUDIES Tools for the assessment of risk of bias (sometimes called quality assessment, although this terminology is less favoured in Cochrane collaboration) have been established for randomised controlled trials, but the corresponding criteria and tools have not been developed for observational studies. We have developed a six-item tool to assess risk of bias in casecontrol studies and a five-item tool for crosssectional studies, with focus on three major sources of biases in observational studies: selection bias, measurement bias and confounding bias (Tables A.3 and A.4). Per the approach of risk of bias used by the Cochrane collaboration, we will not sum across the items to form a summary score, but rather will report the items separately for each included study. MEASURES OF ASSOCIATION As TB disease and death are relatively rare events (,1%), we will assume that OR, risk ratio, rate ratio or HR are approximately

Indoor air pollution and TB

iii

Table A.2 Data collection form

Table A.4 Criteria for risk of bias in cross-sectional studies

Authors/publication year/country: Study design:

1. Selection bias 1.1 Selection of participants Random sample from the source population or tuberculosis prevalence survey: low risk Unclear methods of selecting participants: high risk

Population: Selection of study population: Age/sex: Case: Control: Matching: Response rate: Exposure measurement: Source of information: Type of exposure: Timing of exposure (before or after disease onset): Measurement: Possibility of recall bias: Outcome measurement: Outcome of interest: Source of information: Method of diagnosis: Statistical method/adjustment for confounding/conditioning on infection or contact: Effect measurement: Measurement of association: Test of effect modification: Dose response: Comments:

Table A.3 Criteria for risk of bias in case-control studies 1. Selection bias 1.1 Selection of controls Community-based: low risk Hospital-based and others: high risk 1.2 Selection of cases Consecutive recruitment or systemic sampling of incident cases: low risk Others: high risk 2. Measurement bias 2.1 Exposure assessment Objective measurement of indoor air-pollution: low risk Measurement of at least one of the following (and take into consideration in the analysis): the person who does the cooking, duration of cooking, stove type, ventilation status: low risk Others: high risk 2.2. Outcome assessment Smear or culture positive: low risk Chest X-ray consistent with tuberculosis þ AFB-negative x 3 sets þ no response to empirical antibiotics þ response to antituberculosis treatment: low risk Others: high risk 3. Confounding bias 3.1 Confounding adjustment I At least adjusted for age and sex: low risk No adjustment: high risk 3.2 Confounding adjustment II Adjust for socio-economic status, including one of the following: low risk Education Income House construction type Crowding (number of persons per room) Literacy Kitchen type No adjustment: high risk AFB ¼ acid-fast bacilli.

2. Measurement bias 2.1 Exposure assessment Objective measurement of indoor air-pollution: low risk Measurement of at least one of the following (and take into consideration in the analysis): the person who does the cooking, duration of cooking, stove type, ventilation status: low risk Others: high risk 2.2. Outcome assessment Smear or culture positive: low risk CXR consistent with TB þ AFB-negative x 3 sets þ no response to empirical antibiotics þ response to anti-tuberculosis treatment: low risk Others: high risk 3. Confounding bias 3.1 Confounding adjustment I At least adjust for age and sex: low risk No adjustment: high risk 3.2 Confounding adjustment II Adjust for socio-economic status, including one of the following: low risk Education Income House construction type Crowding (number of persons per room) Literacy Kitchen type No adjustment: high risk CXR ¼ chest X-ray; AFB ¼ acid-fast bacilli.

the same. For LTBI, we will distinguish between different measures of association, as LTBI prevalence can be very high (.5%) in endemic areas. ASSESSMENT OF HETEROGENEITY We will assess the heterogeneity of the included studies (by study design, such as cohort study, case-control study and cross-sectional study) using the I2 statistic and the v2 test of the Q statistic. ASSESSMENT OF REPORTING BIAS We will use funnel plot and contour-enhanced funnel plot to visually assess small-study bias, and apply statistical tests (Egger test and Begg test) to test potential statistically significant bias from missing small and insignificant studies. DATA SYNTHESIS Within each study design, we will assess the heterogeneity of the included studies using the I2 statistic. For observational studies, we will use the Dersimonian and Laird random-effects model to pool the associations, as the true effect sizes are likely to vary across observational studies due to various sources of biases. We will not pool across study designs (such as cohort study, case-control study and cross-sectional study), as different study designs might suffer from different types of biases and pooling across them is generally not recommended.

iv

The International Journal of Tuberculosis and Lung Disease

SUBGROUP ANALYSIS AND INVESTIGATION OF HETEROGEWe will perform subgroup analysis and metaregression based on important domains of sources of biases (see Assessment of risk of bias in included studies) using stratified analysis and meta-regression. We will examine whether the summary effect size varies substantially by important study characteristics, and whether the observed heterogeneity between studies can be explained by these important studylevel factors. NEITY

Anticipated results As the number of epidemiological studies on household exposure to combustion of solid fuel (indoor air pollution) and TB has increased in the past few years, our updated systematic review and meta-analysis will provide critical evidence base for this important association. Our research will also give an overview of study design and study quality on indoor air pollution and TB, and highlight important knowledge gaps that need to be addressed by future research.

Indoor air pollution and TB

v

RESUME O B J E C T I F : Faire une revue syst´ematique mise a` jour et une m´eta-analyse de l’association entre la pollution domestique et la tuberculose (TB). S C H E M A : Nous avons identifi´e des articles en anglais ou en chinois utilisant PubMed et EMBASE jusqu’au 28 f´evrier 2013. Nous avons inclus toutes les e´ tudes randomise´ es contr ol ˆ e´ es et les enqueˆ tes e´ pid´emiologiques d’observation qui rapportaient une association entre l’utilisation domestique de combustible solide et la TB. Deux personnes en ont extrait ind´ependamment les informations et e´ valu´e les risques de biais grace ˆ a` des crit`eres pr´ed´efinis. Les ampleurs de l’effet des e´ tudes retenues ont e´ t´e mises en commun gr ace ˆ au mod e` le d’effet al e´ atoire ; l’h´et´erog´en´eit´e des e´ tudes a e´ t´e quantifi´ee grace ˆ aux statistiques I2.

R E S U LT A T S : Nous avons identifi´e 15 e´ tudes relatives a` l’utilisation de combustible solide et a` la TB active et une e´ tude relative a` une infection tuberculeuse latente. Les OR r´esum´es des e´ tudes cas-t´emoins et des e´ tudes transversales e´ taient de 1,17 (IC95% 0,83–1,65) et de 1,62 (IC95%0,89–2,93) respectivement, avec une h´et´erog´en´eit´e importante entre les e´ tudes (I2 56,2% et 80,5% respectivement). L’analyse des sous-groupes et l’analyse en m´eta-r´egression n’ont mis en e´ vidence aucun facteur li´e a` l’´etude susceptible d’expliquer cette h´et´erog´en´eit´e. C O N C L U S I O N : Le niveau de preuve de l’association entre usage domestique de combustible solide et TB a e´ t´e tr`es faible. Il faut des e´ tudes de grande qualit´e pour clarifier cette association et estimer l’ampleur du probl`eme.

RESUMEN O B J E T I V O: Realizar un examen sistema´tico actualizado y un metana´ lisis sobre la asociacion ´ entre la contaminacion ´ del aire en interiores y la tuberculosis (TB). M E´ T O D O: Se llevo ´ a cabo una busqueda ´ de art´ıculos escritos en ingl´es u en chino en las bases de datos PubMed y EMBASE hasta el 28 de febrero del 2013. Se identificaron ensayos comparativos aleatorizados y estudios epidemiol ogicos ´ de observacion ´ que notificaran la asociacion ´ entre el uso dom´estico de combustibles solidos ´ y la TB. Dos examinadores extrajeron de manera independiente la informacion ´ de los estudios incluidos y evaluaron el riesgo de sesgo de estos estudios, mediante criterios predefinidos. Se agruparon los datos sobre la magnitud del efecto en los estudios que cumpl´ıan los requisitos de inclusion, ´ mediante un modelo de efectos aleatorios; se cuantifico´ la heterogeneidad entre los estudios mediante la estad´ıstica I2.

Se encontraron 15 estudios sobre el uso de combustibles solidos ´ y la TB activa y un estudio sobre el uso de combustibles solidos ´ y la infeccion ´ tuberculosa latente. El cociente de posibilidades acumulado en los ensayos comparativos fue 1,17 (IC95% 0,83 - 1,65) y en los estudios transversales fue 1,62 (IC95% 0,89 - 2,93), con gran heterogeneidad de los estudios (I2 56,2% y 80,5%, respectivamente). El ana´lisis de subgrupos y el metana´lisis de regresion ´ no pusieron en evidencia ningun ´ factor a escala de los estudios, que pudiese explicar la heterogeneidad observada. ´ N: La fuerza probatoria de la asociacion CONCLUSIO ´ entre el uso de los combustibles solidos ´ y la TB fue muy baja. Se precisan con urgencia estudios de buena calidad que permitan dilucidar esta asociacion ´ y calcular la magnitud del problema. R E S U LT A D O S: