Annex 7Recommendation 4 – household use of kerosene: assessment of the quality of the evidence and setting of the strength of the recommendation

Publication Details

A7.1. Assessment of the quality of the evidence

The areas of evidence assessed were:

  1. health risks from kerosene use (emissions)
  2. risks of burns, scalds and poisoning

A7.1.1. Health risks from kerosene use (emissions)

The systematic review on the health risks of kerosene use (see Review 9) compiled evidence on fuel grade and devices used for cooking, heating and lighting, emissions of health damaging pollutants, area concentrations of pollutants in homes, and that from epidemiological studies on a range of health outcomes. Assessment of the quality of the available evidence took account of all this information to assess consistency of the epidemiological findings with what is known about the type and levels of emissions from various kerosene-using devices in common usage.

A reasonable number of studies are available for heating and cooking with kerosene, but only two for lighting. These provide evidence that micro-environmental levels of PM2.5 and other health damaging pollutants can exceed WHO air quality guideline (AQG) levels with the combustion of kerosene. For simple wick devices, PM2.5 levels were in the range 20–400 μg/m3 when kerosene was used for lighting and 340 μg/m3 to more than 1000 μg/m3 when used for cooking. Such levels could lead to a substantially increased risk of multiple adverse health outcomes.

A total of 24 epidemiological studies reporting on the risk of kerosene use, mainly for cooking, with a few related to heating and lighting, were identified. Disease outcomes included lung and salivary gland cancer, respiratory symptoms/spirometry, asthma and allergic conditions, acute lower respiratory infection (ALRI), tuberculosis (TB) and cataract. Due to considerable heterogeneity in study methods, quality and findings, as well as small numbers of studies for some of the outcomes, meta-analysis was not attempted, and grading of evidence for public health interventions (GEPHI) assessment not applied. Grading of evidence for public health interventions (GRADE) domains have been used as a guide for assessing quality.

Study designs

The studies were observational, most cross-sectional, with the remainder using case-control designs. The majority were carried out in developing countries, with a few in more developed countries. Exposure comparisons were described according to fuel type, comparing kerosene with a range of other fuels, which included wood, other biomass and coal in some studies. A comparison was not specified in five of the studies. Outcome measurement was variable, ranging from clinical diagnoses and spirometry, to reported symptoms.

Risk of bias

Ten of the studies did not adjust for confounding factors during their analysis. This, combined with the potential for exposure misclassification due to comparison of kerosene use with other polluting fuels (or unspecified comparison), suggests a potentially high risk of bias in a substantial number of the studies. These sources of possible bias apply across sets of studies reporting on most of the study outcomes. For example, none of the three studies of lung cancer report adjusted odds ratios.

Indirectness

The review combined indirect evidence (on pollutant emissions, micro-environmental (area) concentrations and human exposures) with direct evidence on risks for a range of adverse health outcomes. Thus, direct evidence (albeit of low quality) is available, and there is consistency between emissions of, and exposure to, health damaging pollutants and the risk of disease.

Precision

Most studies had sufficient numbers of cases (case-control) and subjects (cross-sectional) for reasonable precision to be available for all of the outcomes. Pooled estimates were not available as meta-analysis was not conducted.

Heterogeneity

The heterogeneity of key aspects of study design has been noted above. No formal assessment of statistical heterogeneity was carried out, but considerable variation in results for exposure to kerosene use was noted, both within and between studies, often with non-overlapping 95% confidence intervals (CIs).

Publication bias

No formal assessment of publication bias was conducted as outcomes were heterogeneous for some outcomes, and numbers of studies were too few for other outcomes. Unpublished studies were not included, but the search did include Chinese language publications (although none was eligible).

Summary

There was extensive evidence that emissions from kerosene use for cooking, heating and lighting lead to levels of health-damaging pollutants which exceed WHO AQGs, considerably so for use of wick-type devices. Available data for area concentrations of emissions from kerosene burned for lighting were more limited than those for other uses. The epidemiological evidence appeared vulnerable to bias and demonstrated considerable heterogeneity in findings for several outcomes, so was assessed as being of low quality. Overall, however, this evaluation found that the high levels of emissions of health damaging pollutants would be consistent with studies reporting elevated disease risks. Thus, further research using study designs to overcome the limitations of many of the existing studies, should be conducted. It was also noted that four studies published after completion of the systematic review found significantly increased risks of several adverse health outcomes.

A7.1.2. Risks of burns, scalds and poisoning

The systematic review of burns and poisoning (see Review 10) was carried out to assess the levels of risk associated with the use of various household energy devices and fuels. The review included two types of evidence:

  • descriptive studies of risk factors for burns and poisoning, including the devices and fuels used in the home, that might provide evidence relevant to the recommendations made to improve air quality, and;
  • experimental studies measuring impacts of behavioural and technology interventions on burns and poisoning risks.

Risk factors

Many studies described risk factors for burns and poisoning, but few were population-based. Most studied cases in health facilities, thus providing relatively little information on levels and characteristics of risk in the community. Those cases reaching facilities are likely to be socioeconomically and geographically unrepresentative and possibly also not representative of the spectrum of injury. The finding that household fuel use (especially for cooking), particularly kerosene use, was among the most important causes of burn injuries, does provide an indication of the importance of the household setting, and of the role of kerosene in particular. The lack of data on burns from solid fuel stoves may be due to there being few community-based studies on these. Kerosene was responsible for most household fuel poisonings. This was judged to be a reliable finding as kerosene is the most widely used liquid fuel.

Interventions

The variety of interventions and outcomes made the experimental studies unsuitable for meta-analysis. There were only two studies investigating cooking-related burns. One was a randomized controlled trial (RCT) and the other a quasi-experimental study assessing the impacts of improved stoves and fuels on burns, and both were of high quality. Two other studies, also both well-designed and conducted, investigated the effects of awareness-raising on safety risk scores and behaviours. There were no experimental studies investigating prevention of residential heating burns in low- and middle-income countries (LMICs). Only one intervention study investigated lighting. The review authors concluded that the empirical evidence on specific preventive measures and the associated reduction in risk in LMICs is weak. For kerosene poisoning, one RCT and two quasi-experimental studies investigating a disparate range of interventions (educational materials, container proofing and home visits) and outcomes (knowledge and practice scores and incidence rates) were reported. No firm conclusions can be drawn due to this heterogeneity.

Summary

This assessment found that, while there is substantial evidence that household fuel use (and especially kerosene) is an important cause of burns and poisoning in LMICs, the relationship of solid fuels and other fuels (including liquefied petroleum gas (LPG)) to injuries is poorly described, primarily due to a lack of population-based studies. Given the specificity of the linkage between fuel use and injury from burns and poisoning, however, the evidence that household fuels present an important safety risk (a key aim of this systematic review) was assessed as being of moderate quality, with concern about kerosene noted. Although some high quality experimental studies have been reported, these are still few in number and too variable in respect of interventions and outcomes to be pooled. Evidence on the level of risk reduction that can be achieved using various preventive strategies was assessed as being of low quality.

A7.2. Determination of the strength of Recommendation 4: household use of kerosene

The available evidence indicated that levels of health damaging pollutants emitted during household kerosene use were sufficient to expect important health risks, and the safety concerns were also noted. The epidemiological evidence, however, was assessed as scarce and too inconsistent to allow firm conclusions about respiratory and other disease outcome risks. Therefore the priority should be to strengthen the evidence base, while discouraging kerosene use where cleaner and safer alternatives could be promoted.

When assessing the benefits and harms, the GDG expected that avoidance of kerosene use will lead to a reduction in the disease outcomes thought to be linked to its use, and (with greater certainty) a reduction in the risks of burns, fires and child poisoning. Kerosene is currently widely available, can be obtained in small quantities and is easily stored (albeit often not safely), hence the harm of removing an affordable, available fuel may be incurred if alternatives are not also affordable and easily available. Relatively inexpensive lighting alternatives such as solar lamps, for example, appear to offer comprehensive health and safety benefits (avoiding emissions, burns and poisoning risks), greater convenience, and potential medium-term cost savings.

When considering values and preferences, the GDG noted these may vary. It is expected that users will value switching to a cleaner and safer fuel if it is affordable and reliably available. The recent large-scale conversion from kerosene to LPG use for cooking in more than 40 million homes in Indonesia shows that this can be achieved at scale. However, inadequate safety regulation led to accidents with LPG and negative perceptions initially. The extent to which this experience can be generalized is unclear.

When assessing feasibility, the GDG noted that some investment will be required to replace kerosene with cleaner and safer alternatives but, since many countries subsidize kerosene, the balance sheet may be in favour of change. This was the prime motivation for the Indonesian programme, which reportedly reduced costs for both government and households due to LPG being more efficient in terms of energy per unit cost.

On the basis of this assessment, given the low quality of the epidemiological evidence, it was judged that the recommendation be Conditional while additional health research is conducted (Table A7.1).

Table A7.1. Decision table for strength of Recommendation 4: household use of kerosene.

Table A7.1

Decision table for strength of Recommendation 4: household use of kerosene.

Nevertheless, the concerns about emissions and safety led to the conclusion that household use of kerosene should be discouraged.