Household air pollution exposure and respiratory health outcomes: a narrative review update of the South African epidemiological evidence

One of the greatest threats to public health is personal exposure to air pollution from indoor sources. The impact of air pollution on mortality and morbidity globally and in South Africa is large and places a burden on healthcare systems for treatment and care of air pollution-related diseases. Household air pollution (HAP) exposure attributed to the burning of solid fuels for cooking and heating is associated with several adverse health impacts including impacts on the respiratory system. The researchers sought to update the South African evidence on HAP exposure and respiratory health outcomes from 2005. Our quasi-systematic review produced 27 eligible studies, however, only four of these studies considered measures of both HAP exposure and respiratory health outcomes. While all of the studies that were reviewed show evidence of the serious problem of HAP and possible association with negative health outcomes in South Africa, no studies provided critically important information for South Africa, namely, local estimates of relative risks that may be applied in burden of disease studies and concentration response functions for criteria pollutants. Almost all of the studies that were reviewed were cross-sectional, observational studies. To strengthen the evidence of HAP exposure-health outcome impacts on respiratory health, researchers need to pursue studies such as cohort, time-series and randomised intervention trials, among other study designs. South African and other researchers working in this field need to work together and take a leap towards a new era of epidemiological research that uses more sophisticated methods and analyses to provide the best possible evidence. This evidence may then be used with greater confidence to motivate for policy-making, contribute to international processes such as for guideline development, and ultimately strengthen the evidence for design of interventions that will reduce HAP and the burden of disease associated with exposure to HAP in South Africa.

HAP is derived from multiple indoor sources (Colbeck and Nasir 2010) varying from one building to another and depending on fuels used for heating and cooking, smoking habits and use of a wide variety of consumer products and building materials (Shezi et al. 2017;Jafta et al. 2017;Tanaka et al. 2012;Verma et al. 2010). This is also influenced by time activity patterns i.e. cooking period, number of meals cooked, single or multiple fuel use and number of cigarettes smoked and further compounded by confounding variables such as outdoor air pollution sources near the homes (Shezi et al. 2017;Colbeck and Nazir 2010). In developing countries, the most significant indoor air quality issue is exposure to pollutants released during combustion of solid fuels used for cooking and heating in the home (Wylie et al. 2017;Pope et al. 2010).
The measured mean concentrations of the HAP vary depending on the sources, for example, average particulate matter with an aerodynamic diameter of 2.5 (PM 2.5 ) concentration in households using solid fuels has been reported to range from 133.5 µg/m 3 to 670 µg/m 3 (Balakrishnan et al., 2015;Balakrishnan et al., 2013;Clark et al., 2010), while the carbon monoxide (CO) average concentrations have been reported to range from 2.7 ppm to 14.3 ppm. The average PM 2.5 concentrations in households using cleaner fuels have been reported to range from 10 µg/m 3 to 38 µg/m 3 (Shezi et al. 2017;Tunno et al., 2015;Evans et al. 2000). Studies conducted in South Africa have also reported high mean concentrations in households using both clean and dirty fuels (Jafta et al. 2017;Shezi et al. 2017;Wernecke et al 2015;Rollin et al. 2004) which exceed the guidelines set out by the WHO for indoor air quality in households; South African indoor air quality guidelines pertaining to household air pollution are under review and have not yet been promulgated.
South Africa is a middle-income country burdened by poverty and inequality where many South Africans are exposed to biomass fuels used for cooking and heating indoors. Many Research article: Household air pollution exposure and respiratory health outcomes Page 2 of 14 families live in close proximity to industrial areas and major roads (Albers et al. 2015;StatsSA 2012;Norman 2007) where outdoor pollutants may also be transported indoors (Colbeck and Nasir 2010). Vulnerable groups such as people who are elderly, women, young children, people with pre-existing diseases and people living in poverty are more susceptible to air pollution health impacts (Barnes 2014).
The WHO comparative risk assessment determined the mortality burden and Disability Adjusted Life Year (DALYs) due to HAP. HAP associated with household solid fuel use for cooking and heating caused 0.5% of all deaths in South Africa in 2000 (uncertainty 0.3% -0.6%) (Norman et al., 2010). More than 10 years ago, Wichmann and Voyi (2005) published a review of the air pollution and health epidemiological studies in South Africa, focussing on methodological issues and the need for quantitative intervention studies. In 2014, Barnes (2014) described behavioural change studies for reduction of HAP in developing countries, calling for more rigorous study design and interventions grounded in behavioural change theory. In recent years, anecdotal evidence suggests that studies on the impact of HAP and associated respiratory health outcomes has been growing in South Africa (Barnes et al., 2009). In this narrative review, evidence from the recently published South African studies that considered HAP exposure and associated respiratory health outcomes to augment our current knowledge is collated. An attempt to identify research gaps and suggest directions for further studies is made.

Review methods
A quasi-systematic (i.e. following the guidelines for systematic review but with slight differences in methods to accommodate all available evidence) review of the South African evidence on HAP (term group 1) and associated respiratory health outcomes (term group 2) was conducted using the PRISMA guidelines (Moher et al. 2009). PubMed, Web of Science, Science Direct and Google Scholar were searched for studies with full text in English, published between 2005 and 2017. The term groups listed in Table 1 were used for the separate searches and in various combinations. The reference lists of included papers were searched to ensure that no studies were omitted. To be eligible, studies had to have been carried out in South Africa. All epidemiological study designs including crosssectional studies, cohort studies, longitudinal studies, casecross-over studies, intervention studies etc. were eligible. All studies that provided estimates of HAP exposure concentrations (by indicator, proxy or actual measurements) as well as respiratory health outcomes found to be associated with HAP exposure were included. We noted whether the correlate and health outcome was measured at the level of the individual, household or community. The review was not restricted by defining a minimal study sample size and studies of all sample sizes were included.

Results
After removing all ineligible articles, mostly since they were studies unrelated to South Africa, our searches using both formal methods (described above) and informal means (such as personal communication with researchers) produced a total of 27 articles. Of this total, only 4 studies measured HAP exposure (mainly by proxy/indicators of air pollution exposure) and respiratory health outcomes. Ten studies assessed HAP but no respiratory health outcome(s) and the remaining 13 studies did not measure HAP, instead used ambient AQ monitoring station data (or other) for exposure assessment.  et al., 2014) with the other three studies having sample sizes of less than 1 000 individuals (no sample size calculations were provided). One study did not set out to measure HAP but rather focussed on ETS exposure.
Inter-comparison of the findings of the studies is difficult due to the heterogeneity in the study design, target population and health outcomes of interest. Both Albers et al. (2015) and Shirinde et al. (2014) found that among schoolchildren (albeit of different ages) there was an association between respiratory outcomes (most notably wheeze) and use of non-electrical heating sources. Elf et al. (2017) and du Preez et al. (2011) consider TB as a health endpoint and ETS and solid fuel use, and ETS, respectively, in relation to TB.
Tables 3 and 4 summarise the remaining studies that the researchers reviewed as part of this review exercise. These studies did not meet the study inclusion criteria; however, they do provide useful information on the levels of various air pollutants in the indoor and outdoor environments of different parts of South Africa. Since they did not mention respiratory health outcomes, the researchers do not discuss them in detail here, but included them as a reference for future research.
Of the twenty-three studies listed in Table 3 and 4, two studies were review articles, six studies used questionnaires and interviews which are methods of assessment that are based on self-report and recall. Either indoor and / or outdoor pollutants were measured in 15 studies and some of the studies measured multiple pollutants. Particulate matter with an aerodynamic diameter of 10 (PM 10 ) was measured in twelve studies while PM 2.5 and particulate matter with an aerodynamic diameter of 4 (PM 4 ) were measured in four and two studies, respectively. Two studies measured respiratory particulate matter (RPM). NO 2 was measured in five studies; NO x in one study and volatile organic compounds were also measured in one study. Three studies measured CO and seven studies measured sulphur dioxide (SO 2 ). Ozone (O 3 ) was measured in one study and total reduced sulphur was also measured in one study. Indoor undisturbed dust samples were undertaken in one study to analyse for dichlorodiphenyltrichloroethane (DDT), dichlorodiphenyldichloroethylene (DDE) and dichlorodiphenyl-dichloroethane (DDD), while settled dust and airborne fungal sampling was also conducted in one study.

Discussion
Few published studies conducted in South Africa in recent years have sought to show associations between HAP and respiratory health outcomes. All of the studies that did so used cross-sectional study design. This is unfortunate since most international systematic reviews for HAP exposure and health outcome association only use data from cohort, time series, long-term panel (longitudinal) and case-crossover studies. Our studies are therefore not contributing to international evidence on this important topic, nor are they providing reliable evidence that is generalizable to others parts of South Africa. While crosssectional studies are typically less expensive and easier to implement compared to other epidemiological study designs, the data produced from cross-sectional studies is not as useful and the lack of randomisation, among other shortfalls, prohibits generalisation. Researchers need to work towards larger, more complex epidemiological studies and also use of existing, high quality (where possible) data in South Africa that will help to address the problem of HAP and respiratory health. While it will likely cost more, other study designs beside cross-sectional studies, will provide opportunities for HAP monitoring over substantial periods of time to provide exposure information that can then be associated with health outcomes, preferably diagnosed by a health professional (and not self-reported or parent-reported) for more precise results, in a more meaningful way.
It is also important to account for potential biases and critical confounders, including age, sex, and individual socio-economic status, among others, when planning a HAP and respiratory health study, and also when reporting study results as failure to control for confounding variables can lead to erroneous associations between HAP and respiratory outcomes.
While the reviewed studies predominantly report on combustion generated variables (type of fuel used for cooking or heating and active or passive smoking) with PM, CO and NO 2 , being products of incomplete combustions, other sources of indoor air pollutants not necessarily emitted by incomplete combustion such as building materials, ventilation characteristics and cleaning agents may not be ignored.
While, exposure data misclassification and validity of exposure are often overlooked or underestimated and not critically discussed (Wichmann and Voyi, 2005), the lack of direct exposure measurement such as the use of home monitors and personal monitors present results that are in some part questionable.
None of the studies reviewed provided concentration response functions for the criteria air pollutants in South Africa (DEA, 2009). This point was made by Wichmann and Voyi (2005) and it still holds true in 2017. This is still a major gap in our knowledge in both South Africa and on the continent. South African researcher continue to use the international literature and WHO evidence, without making contribution to this important body of knowledge.
Another shortfall is that only one study reviewed in this exercise was an intervention study (under real life conditions). The WHO calls for the support of research that is driven by interventions and searching actively for solutions, in particular for urban settings, in relation to air pollution and health (WHO 2014). In South Africa, research partnerships and consortia may assist to create large research teams to lead intervention studies to address HAP and adverse health impacts in areas of greatest concern.
Our study had some limitations. There were very few studies to critically review hence the researchers opted to describe them descriptively instead. Wichmann and Voyi (2005) provided a critical synthesis of the evidence in this field up to 2005. The authors did not find a substantial number of studies to add to that body of literature beyond 2005. The authors set out to find studies that had measured HAP and simultaneously measured respiratory health outcomes. The authors may have not identified all studies relevant to the review inclusion criteria, although they tried to avoid this by searching the literature regularly and speaking with South African researchers in the air pollution and health fields. The authors did not consider mortality as an end-point. Wichmann and Voyi (2006) found that exposure to cooking and heating smoke from polluting fuels was significantly associated with 1-59-month mortality, after controlling for mother's age at birth, water source, asset index and household crowdedness (RR=1.95; 95% CI=1.04, 3.68).

Conclusions
The South African studies on HAP and respiratory health outcomes do provide some evidence of the serious impacts Research article: Household air pollution exposure and respiratory health outcomes Page 4 of 14 of HAP and especially the use of solid fuels in the home on respiratory health in the country, but the studies are few and limited. South African and other researchers working in this field need to work together and take a leap towards a new era of epidemiological research that uses sophisticated methods and analyses to provide the best possible evidence. This evidence may then be used with greater confidence to, for example, motivate for policy-making, contribute to international systematic reviews for guideline development and other purposes, and ultimately strengthen interventions that will reduce HAP and the burden of disease associated with exposure to HAP in South Africa.   A high prevalence of air pollution from second hand smoke, solid fuels, and kerosene among individuals in homes with a case of prevalent active tuberculosis disease was observed. Adults in 40% of homes reported a daily smoker in the home, and 70% of homes had detectable air nicotine.
In homes with a history of previous TB (prior to but not including the index case) as compared to those without previous tuberculosis, both second hand smoke (83% vs. 65%, respectively) and solid / kerosene fuel use for more than 1 h/day (27% vs. 21%, respectively) were more prevalent.
Albers et al (2015) Cross-sectional study Children 9-11 years old in grades 4 and 5 in six randomly selected primary schools in eMalahleni and Middelburg, Mpumalanga.

627
Type of energy sources and associated respiratory outcomes were collected using a structured questionnaire completed by parents/guardians of the children.  196 Questionnaire was used to collect information on ETS exposure.
Shezi et al. (2017) Cross-sectional study Households of pregnant females from the north and the south of Durban participating in the mother and child in the environment (the MACE cohort study).

households
Walk-through indoor assessment and postactivity questionnaire were used to collect information on the household building materials, occupant activities and outdoor sources such as industries and major roads in the vicinity of the homes. Indoor PM 2.5 levels were measured in 300 homes for a period of 24 hours.

Nkosi et al. (2017)
Cross-sectional study Children, including 10 asthmatics 13-14 years of age, a subset of the 2012 International Study of Asthma and Allergies in Children (ISAAC) from 10 schools in Gauteng and the North-West Province were included. Five schools were within 1-2 km from a mine dump in Gauteng or the North-West Province and 5 were 5 km or further from a mine dump in these provinces.

children
Personal air sampling was performed in the breathing zone of 10 asthmatics learners randomly selected from each school, outdoor SO 2 and PM 10 were measured for 8-h at each school.
Indoor respiratory dust in the classroom differed significantly between exposed (0. 17mg/m 3 ) and non-exposed (0. 01 mg/m 3 ) among children with asthma. Outdoor SO 2 levels were 0.002 ppb for exposed children and 0.01 ppb for unexposed children (p<0.001). Outdoor PM 10 was 16.42 mg/m -3 and 11.47 mg.m -3 for exposed and unexposed, respectively.  Cohort study Households of women from the VHEMBE study on indoor spraying for malaria and health effects.

households
Indoor undisturbed dust samples analysed for dichlorodiphenyltrichloroethane (DDT), and its degrading products, dichlorodiphenyldichloroethylene (DDE) and dichlorodiphenyldichloroethane (DDD) to determine dust loading levels and compared these levels to paired serum concentrations of p,p′-DDT-and p,p' DDE in women residents.
p,p′-DDT and p,p′-DDE had the highest detection frequencies in both dust (58% and 34% detection, respectively) and serum samples (98% and 100% detection, respectively). Significantly higher detection frequencies for o,p′-DDT, p,p′-DDE, and p,p′-DDD were observed in dust samples collected in buildings that had been previously sprayed for malaria control. Significant, positive association between dust loading and serum concentrations of p,p′-DDT and p,p′-DDE (Spearman's rho=0.68 and 0.54, respectively) was observed.

women
Indoor air monitoring of PM 10 , CO was conducted for a period of 24 hours, while indoor NO 2 and volatile organic compounds (VOC) were conducted for a period of 2 weeks.

Cross-sectional study
Households of children (20 children from each of 7 schools, Grades 3-6) (mainly asthmatics) who took part in the South Durban Health Study.

households
Observation (walk through survey), sampling and analyses of settled dust and airborne fungal sampling.
Asp f1 allergen was detected in all homes, and Bla g1 allergen was detected in half of the homes. Detection frequencies varied from 51% for Bla g1 to 100% for Asp f1. House dust allergens, Der f1 and Der p1 exceeded concentrations associated with risk of sensitization and exacerbation of asthma in 3% and 13%, respectively, of the sampled homes, while Bla g1 exceeded guidance values in 13% of the homes. Although airborne fungal concentrations in sleep areas and indoors were lower than outdoor concentrations, they exceeded 1 000 colony forming units per cubic meter of air in 29% of the homes. 25.7% of students were exposed ETS at home, 34.2% outside of the home. Parental and close friend smoking status, allowing someone to smoke around you and perception that passive smoking was harmful were significant determinants of adolescent's exposure to both ETS at home and outside of the home.
Barnes et al. (2011) Intervention study The study took place in two poor rural villages in the North-West Province.
Intervention group (n=36) and control group (n=38) The study employed a quasiexperimental before and after study design with a control group. Baseline data was collected during winter in both the intervention and control group. Follow-up data was collected from both groups 12 months later. The intervention was implemented immediately after the baseline data collection in the intervention group only.
The following were promoted (burn outdoors when possible, if fires are burned indoors, open at least two sources of ventilation during peak emission times, reduce the amount of time that children spend in the burning room while fires are burning).
The likelihood of eczema ever was increased by exposure to ETS.
Eczema is not a respiratory health outcome.
Nkosi et al.
Crosssectional study Study on wheeze, asthma, and rhino conjunctivitis associated with community proximity to mine dumps. The study focussed on 13-14-year old pupil who attended schools located between 1-2 km from mine dumps and those living >5 km away from the mine dumps in the Gauteng and North-West province.
3 641 Self-administered questionnaire on asthma and rhinitis (among others) was completed by the adolescent living at a certain distance (within 1-2 km -exposed and >5 km -unexposed) from five pre-selected mine dumps.
No exposure measurement. Proximity to mine dump proxy for exposure.

397
Questionnaire was used to collect information on respiratory outcomes. Overall, 55.9% of all non-smokers reported exposure to second hand smoke from at least one source (i.e., in the home, workplace or at a hospitality venue).
No IAQ measurements. ETS as a proxy for exposure.
Research article: Household air pollution exposure and respiratory health outcomes Page13 of 14 The adjusted odds ratios (AORs) were elevated (p<0.05) for children in the south for 5 of the 13 outcomes investigated: doctordiagnosed chronic bronchitis (AOR 3.5, 95% confidence interval (CI) 1.6 -7.7), as well as bronchitis by symptom definitions; watery/ itchy eyes; wheezing with shortness of breath; and marked airway hyper reactivity (AHR). In addition, marked AHR was associated with SO 2 exposure. The prevalence of symptoms consistent with asthma of any severity was 32.1%. Covariate-adjusted prevalence were higher among children from schools in the south than among those from the north for persistent asthma (12.2% v. 9.6 %) and for marked airway hyper reactivity ( Association (excess mortality risk) between PM 10 , and respiratory disease, cardiovascular disease and cerebrovascular disease mortality of 1.1% (CI −1.1, 3.3), 1.7% (CI −0.1; 3.5 and 3.2% (CI 0.3; 6.2) following a 10 µg/m 3 increase (entire year data) respectively; association between NO 2 and respiratory disease mortality of 1.7% (CI −1.3; 4.7), following a 10 µg/m 3 increase (entire year data) between NO 2 , and cardiovascular disease and cerebrovascular disease mortality of 2.6% (CI 0.2; 5.0) and 6.6% (CI 2.4; 11.0) following a 10 µg/m 3 increase (entire year data), respectively.
Ambient air pollution monitoring data. No IAQ measurements.