Household air pollution, chronic respiratory disease and pneumonia in Malawian adults: A case-control study

Background: Four million people die each year from diseases caused by exposure to household air pollution. There is an association between exposure to household air pollution and pneumonia in children (half a million attributable deaths a year); however, whether this is true in adults is unknown. We conducted a case-control study in urban Malawi to examine the association between exposure to household air pollution and pneumonia in adults. Methods: Hospitalized patients with radiologically confirmed pneumonia (cases) and healthy community controls underwent 48 hours of ambulatory and household particulate matter (µg/m 3) and carbon monoxide (ppm) exposure monitoring. Multivariate logistic regression, stratified by HIV status, explored associations between these and other potential risk factors with pneumonia. Results: 145 (117 HIV-positive; 28 HIV-negative) cases and 253 (169 HIV-positive; 84 HIV-negative) controls completed follow up. We found no evidence of association between household air pollution exposure and pneumonia in HIV-positive (e.g. ambulatory particulate matter adjusted odds ratio [aOR] 1.00 [95% CI 1.00–1.01, p=0.141]) or HIV-negative (e.g. ambulatory particulate matter aOR 1.00 [95% CI 0.99–1.01, p=0.872]) participants. Chronic respiratory disease was associated with pneumonia in both HIV-positive (aOR 28.07 [95% CI 9.29–84.83, p<0.001]) and HIV-negative (aOR 104.27 [95% CI 12.86–852.35, p<0.001]) participants. Conclusions: We found no evidence that exposure to household air pollution is associated with pneumonia in Malawian adults. In contrast, chronic respiratory disease was strongly associated with pneumonia.


Introduction
Four million people die each year from diseases caused by exposure to household air pollution from the domestic burning of solid fuels 1 . Half a million of these deaths are due to acute lower respiratory infections (ALRI) in young children 2 . In adults, the majority of deaths are attributed to chronic obstructive lung disease, cardiovascular diseases, and lung cancer 3 . Although plausible, it is not known if household air pollution is associated with ALRI in adults as it is in children 4 .
In low-income areas such as Malawi, pneumonia is the commonest cause of admission to hospital for adults and has a high fatality rate [5][6][7] . HIV infection is a well-established risk factor for pneumonia; the extent to which other factors, such as household air pollution and other poverty-related exposures, affect the risk of pneumonia has not been adequately studied 4,8 . In low-income countries burdened with high rates of adult pneumonia, domestic use of solid fuel is widespread 1 . If an association between household air pollution and adult ALRI is found, the attributable risk is potentially high.
We conducted a case-control study, The Acute Infection of the Respiratory tract (AIR) study, to test the hypothesis that household air pollution and chronic respiratory disease (CRD) are associated with an increased risk of pneumonia in adults living in urban Malawi.

Setting
Malawi, population 16.7 million, is one of the world's poorest countries, and has a life expectancy of 59 years 9, 10 . Blantyre, Malawi's second city, has a HIV prevalence of 18.5% 11 . Queen Elizabeth Central Hospital (QECH) is a large government hospital providing free health care to a population of 1.3 million in greater Blantyre.

Participants
Cases were defined by the presence of radiologically-confirmed pneumonia requiring hospitalisation and controls were defined by the absence of pneumonia. Inclusion and exclusion criteria are presented in Box 1.
All adult medical admissions to QECH were screened for symptoms suggestive of pneumonia by study clinical officers to identify potential cases. For control recruitment, residential census enumeration areas were randomly selected from all enumeration areas within Blantyre city, with selection weighted by population size. Field workers followed randomly generated routes within these enumeration areas and screened all potential participants (including performing HIV tests) in each household along the route. A maximum of one individual was recruited per household, selected randomly. Screening continued until two controls had been recruited from that enumeration area. To supplement Box 1. Inclusion and exclusion criteria for cases and controls.

Inclusion criteria
Age 18 years or over Resident in Blantyre city Reported cough or chest pain or breathlessness or hemoptysis Reported fever or recorded fever (≥38°C) Crepitations or pleural rub or bronchial breathing Radiological changes judged to be new and consistent with pneumonia, without another obvious cause Requires hospitalisation Age 18 years or over Resident in Blantyre city

Exclusion criteria
Pre-admission diagnosis of terminal illness (e.g., metastatic malignancy, terminal AIDS) Current anti-tuberculosis treatment or evidence of current tuberculosis infection Prior hospitalisation within the last 4 weeks Prior participation in the study Lives in a residential institution (e.g., prison) Death prior to follow-up assessment Alternative diagnosis explaining their presentation Symptoms for 14 days or more Pre-admission diagnosis of terminal illness (e.g., metastatic malignancy, terminal AIDS) Current anti-tuberculosis treatment or evidence of current tuberculosis infection Hospitalisation for a pneumonia-like illness in the past 4 months or current pneumonia-like illness Prior participation in the study Lives in a residential institution (e.g., prison) Death prior to follow-up assessment Utilizes private health care facilities if has illness requiring hospitalisation

Note
For pragmatic reasons relating to resource availability, individuals could be recruited as a 'provisional case' prior to having a chest x-ray. Individuals were subsequently excluded if there was no evidence of pneumonia on chest x-ray or if they later met exclusion criteria (e.g. were commenced on tuberculosis treatment or died). Individuals were designated as a 'case' only when they had completed follow-up.
door-to-door recruitment, HIV-positive individuals attending community antiretroviral clinics within Blantyre city were also screened. Recruitment was stratified by HIV status to enable the data to be analyzed as two separate case-control studies. Within these two subgroups, controls were frequency-matched to cases by age (18-34 years or ≥35 years) and gender.
Cases and controls were contemporaneously recruited and followed-up throughout the study period, to account for temporal changes in air pollution exposure related to season. Case recruitment was from July 2014 until January 2016, and follow up appointments took place between September 2014 and March 2016. Control recruitment and follow up appointments took place from August 2014 until February 2016.

Study procedures
Initial assessment of provisional cases included medical history and examination by study clinical officers, and diagnostic tests (Box 2). Pneumonia was confirmed by chest X-ray review by a study clinical officer and a study doctor.

Box 2. Hospital diagnostic tests for provisional cases.
• HIV test +/-CD4 count  13 , evaluated a range of potential risk factors and socioeconomic status. The primary exposures of interest were mean ambulatory PM 2.5 exposure and presence of CRD (defined using a composite questionnaire assessment (Box 3)). The full study protocol has been published elsewhere 14 .

Box 3. Composite definition of chronic respiratory disease
Answering affirmative to any of the following in the BOLD questionnaire: • Current usual cough • Current usual sputum production • Current breathlessness • Wheeze in past 12 months • Ever had diagnosis of emphysema • Current diagnosis of chronic bronchitis • Current diagnosis of asthma • Current long-term respiratory medication Note: Cases were asked to recall their status from 6 months previously, prior to their episode of pneumonia.

Sample size
Based on assumptions of α=0.05, β=0.2, and an estimated percentage of controls with CRD of at least 15%, the target sample size was 160 cases and 160 controls in the HIV-positive subgroup (ratio 1:1) to detect an odds ratio (OR) of 2.2 or greater. A smaller exploratory study was planned with 60 cases and 90 controls in the HIV-negative subgroup (ratio 1:1.5).

Statistical considerations
Data files were exported to Stata 13.1 (Statacorp, College Station, TX, USA) for analysis. Missing air pollution exposure data (for ambulatory PM 2.5 and CO levels and household CO levels) were imputed using spatial interpolation. Due to the large number of missing data, household PM 2.5 data were not imputed. Missing questionnaire data were imputed using multivariate multinomial models by exploiting their association with other observed variables. A socioeconomic status score was generated using principal components analysis, based on data regarding asset-based measures, education level, and household characteristics 15 .
Univariate logistic regression, including a priori potential confounders, was performed for each subgroup. For analysis of the HIV-positive subgroup, multivariate forward stepwise logistic regression was performed for each of the main exposures of interest (a priori potential confounders (as indicated in Table 2 and Supplementary File 2) were included in the model if their likelihood ratio test p-value was <0.2 (criteria for entry p<0.05 and removal p>0.1)). Adjustment in the HIV-negative subgroup was limited to frequency-matched factors (age and sex).
To test the hypothesis that pneumonia cases are spatially clustered, we used generalized additive models and smoothing latitude and longitude over the geographic reach of the study area 16 .

Ethical considerations
This study was approved by the College of Medicine Research Ethics Committee, University of Malawi (P.02/14/1518) and the Liverpool School of Tropical Medicine Research Ethics Committee (14.016). All participants gave written informed consent prior to participation in the study.

Baseline characteristics of cases
Comparisons between cases who completed follow-up and provisional cases who did not complete follow-up (predominantly due to ineligibility, see Figure 1) were made to explore evidence of  ambulatory and household PM 2.5 and CO had no effect on pneumonia risk with unadjusted ORs of approximately one for all measures of exposure (Table 2). In the HIV-positive subgroup, there were no consistent findings related to frequency of cooking with solid fuels and no significant findings relating to fuel use, household ventilation, or other forms of pollution exposure (

Spirometry
Two independent reviewers deemed the spirometry data usable as per American Thoracic Society standards in 349 (87.7%) participants who completed follow-up. The two reviewers agreed on the spirometry interpretation for 99.1% of participants. Of the 91 (72.8%) cases that had abnormal spirometry at their initial follow-up appointment, it was only possible to repeat spirometry in 13 (14.3%) a minimum of 4 months after their pneumonia episode to determine their final spirometry status: spirometry remained abnormal in all these individuals. Pre-bronchodilator percentage of predicted forced expiratory volume in 1 second and forced vital capacity were lower in cases than in controls in both the HIV-positive and HIV-negative subgroups (

Multivariate analysis of potential risk factors
After adjustment for confounders, mean ambulatory and household PM 2.5 and CO exposures were not associated with pneumonia in the HIV-positive or HIV-negative subgroups ( Overall, 349 provisional cases (81.5%) were HIV-positive, with 33 (9.5%) of those newly diagnosed. One hundred and eighty-eight (78.0%) of those previously known to be HIV-positive were taking antiretroviral treatment prior to admission. One hundred and twenty-four (85.5%) provisional cases reported a previous diagnosis of emphysema and 34 (23.4%) reported previous tuberculosis. Forty-six (10.7%) provisional cases died prior to hospital discharge and a further 18 (5.2%) died following discharge, prior to followup. Note that post-discharge mortality was only known for those who had not already been excluded.
Univariate analysis of potential risk factors Findings were consistent for pollution assessment modalities in both HIV-positive and HIV-negative subgroups: exposure to

Discussion
We found no association between household air pollution exposure, measured using ambulatory and household monitoring of pollutants, and radiologically confirmed pneumonia in urban HIV-positive Malawian adults. This was consistent when measuring ambulatory and household PM 2.5 and CO, and selfreported exposures. Similar results were found in an exploratory study of HIV-negative individuals. In contrast, we found a strong association between CRD (defined by participant-reported symptoms and diagnoses), as well as spirometric abnormalities, and pneumonia in both HIV-positive and HIV-negative individuals in this setting.
The AIR study and our earlier BOLD study in Malawi both found a high prevalence of restrictive lung disease 17 . The underlying etiology, pathology, epidemiology and prognosis of this low FVC phenomenon requires further investigation, particularly since low FVC is associated with increased mortality in other settings 18,19 . The relationship identified by AIR between restriction and pneumonia is potentially relevant to our understanding of this increased mortality.
It seems likely that socioeconomic factors explain the unexpected findings of reduced risk of pneumonia in current smokers and consumers of alcohol, as in the context of urban Malawi, the poorest individuals cannot afford cigarettes and alcohol 20 . The low prevalence of smoking in this setting means that smoking is not a major driver of pneumonia risk in multivariate analysis, unlike in higher resourced countries 21 .
Reduced BMI was a strong predictor for pneumonia in both subgroups, even after adjustment for CD4 in the HIV-positive subgroup; malnutrition may play a role in pneumonia risk, as has been shown in children 22 . Further research into nutritional status in this population and possible interventions is warranted.
This is the only study of household air pollution and pneumonia in adults to have used multiple measurements of air pollution exposure with radiologically confirmed hospitalised pneumonia cases. While there is no gold standard method for air pollution monitoring, measuring ambulatory and household levels of two different major components of air pollution (PM 2.5 and CO) is likely to capture a representative picture of an individual's total exposure. Mean household PM 2.5 levels detected (all homes: median 149.5 µg/m 3 , IQR 85.0-289.0 µg/m 3 ) and mean household CO levels (all homes: median 6.4 ppm, IQR 2.9-12.6ppm) were comparable to those detected in a previous study of urban Malawian homes (mean 150 µg/m 3 , standard deviation 360 µg/m 3 and mean 6.14 ppm, respectively) 23 25 . This study from rural Kenya conducted household monitoring prospectively for 12 hours per day over a 2-year period, but did not use radiologically confirmed pneumonia, did not account for HIV status, and their cohort included children over the age of 5 years.
The lack of association between household air pollution and ALRI in adults identified in this study may be explained by the overwhelming effect of other major risk factors (such as CRD, HIV-associated factors and BMI) in this setting. This could explain why an association is evident in children but not adults, in whom lifelong exposure to other factors plays a more important role 2 . Alternatively, it is possible that we have not detected a true association between household air pollution and ALRI in adults owing to methodological limitations, in particular with exposure assessments.
Although the AIR study is the largest and most detailed study of household air pollution and ALRI in adults to date, it has a number of limitations. We were unable to evaluate HIV-positive and HIVnegative individuals together due to a lack of statistical power, but our findings were broadly consistent across both groups. The original sample size for both groups was not met because recruitment was slower than anticipated leading to lower power to detect an effect. However, in the HIV-positive group, we were able to detect an OR greater than 1.0003 per unit change for ambulatory PM 2.5 exposure (our primary exposure of interest) with 80% power. Findings in the HIV-negative group are exploratory only. Potential risk factors were assessed after the episode of pneumonia, and so questionnaire assessments may have been subject to recall bias. Our composite assessment of CRD is not validated and may have been vulnerable to recall bias, although our findings are corroborated by spirometric data. Objective measurements of air pollution exposures were made, but these may not be representative of pre-pneumonia exposures, although 138/142 (97.2%) cases reported that they had returned to normal levels of function. A sensitivity analysis, in which cases without reported full functional recovery were excluded, also found no effects of mean ambulatory PM 2.5 exposure (data not shown). In addition, our exposure monitoring does not account for differences in exposure over the life course. The ambulatory pollutant monitoring is unable to distinguish between outdoor and indoor exposures, although since our findings are consistent across ambulatory, household, and questionnaire assessments, we argue that our findings are reflective of the effects of household air pollution.
Pneumonia is a major health burden in sub-Saharan Africa 3 . Although there are compelling reasons for tackling household air pollution 26 , other issues need to be addressed to reduce the burden of pneumonia in adults. Evidence from this study can be used to establish global estimates for the contribution of household air pollution exposure to the burden of disease, to ensure the limited available resources for public health interventions are appropriately directed. Risk factors associated with pneumonia in this study, such as HIV and BMI, are typically associated with socioeconomic status, indicating that poverty is an important driver of pneumonia in urban African adults. To reduce the burden of pneumonia, further research into the effects of CRD and the underlying etiologies in this setting are required. Prenatal and childhood malnutrition may play a role. Targeted evidencebased strategies to reduce the high burden of CRD seen in young adults are needed and may help to tackle the high morbidity and mortality caused by pneumonia.

Data availability
The raw dataset for the AIR study is available on OSF Indoor air pollution and pneumonia is a well-recognised association in children. The authors of this study evaluated indoor air pollution and the risk of pneumonia in adults. They showed no increased risk of pneumonia attributable to indoor air pollution exposure. This is a well conducted and very intensive study documenting pneumonia accurately (radiologically) and exposure (environmental monitoring) objectively. As with all studies there are limitations such as the case control design -although significant numbers were recruited, the evaluation of air pollution exposure was post event (pneumonia), and the number of confounders that need to be adjusted for. What is striking is the absolute lack of effect OR 1 with 95% CI 0.99 -1.01.
It is unlikely that doing a prospective study with regular (pre event) air pollution monitoring, given the data presented would make any difference to the conclusion, and the short time between admission and subsequent air quality measurement precludes any real potential for significant differences in cooking behaviour before the event and after.
What this study does highlight is the difference between children and adults -and that we can't simply assume that the exposure and risks would be the same. Additionally the complex interactions of poverty, malnutrition, indoor air pollution, tobacco and alcohol require robust statistical methods. The univariate models showed effects of cooking with wood and odds of pneumonia to be 13. This was not evident in the HIV positive group -suggesting that HIV infection may play a far more significant role as a risk factor and the risk attributable to wood smoke does not significantly add to the high HIV risk. It is not possible by the nature of the study to evaluate the rate of pneumonia in the HIV negative vs. HIV positive groups for given exposure.
It would be of value if the statistical community could evaluate such complex interactions/ confounders/ effect modifiers in such a cohort with newer tools, so that when we see a result such as this, which although plausible is a little unexpected, that we are certain that the lack of statistical effect is not due to 'underpowered" statistical tools that can't handle such complex and multifaceted interactions.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound?

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Yes
No competing interests were disclosed.

Competing Interests:
I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard. 20  This case control study set in Malawi compares hospitalised pneumonia patients and controls, with the aim of identifying potential risk factors for pneumonia. In addition, the results are stratified by HIV status allowing analysis as two separate case control studies. One of the strengths of this study is the confirmation of pneumonia with radiology in the setting of a large hospital in Blantyre. Controls were selected from enumeration areas nearby, and not from the hospital but frequency matched on age and gender to the cases. While this might not be the ideal methodology for finding controls, it is a common method in this environment. Inclusion and exclusion criteria were fairly strict meaning that study ineligibility on screening was relatively high. Air pollution monitoring was not always available with a delay from recruitment to monitoring start, which may contribute to the lack of association found between air pollution and pneumonia. Ideally exposure measurement would precede the outcome. Unsurprisingly, and in keeping with developed countries, there was an association between chronic respiratory disease and pneumonia. Working in this setting is never straightforward and the authors have certainly tried to undertake as robust a study as possible.

Is the study design appropriate and is the work technically sound? Yes
Are sufficient details of methods and analysis provided to allow replication by others?