Using breath carbon monoxide to validate self-reported tobacco smoking in remote Australian Indigenous communities

Over the past two decades, smoking rates in Australia have halved in the general population from 35% to 18% with predictions of 14% by 2020[1]. However, for Indigenous Australians, smoking rates appear to have remained unchanged. Nationally, 51% of Indigenous men and 47% of Indigenous women report to be regular smokers[2].Smoking rates vary across Indigenous Australian populations. For example, much higher rates of between 59% to 83%[3‐8] have been documented in some remote communities of the Northern Territory (NT), with up to 92% of people reporting a history of tobacco use in one community[3].

Indigenous Australians experience a burden of disease 2.4 times that of non-Indigenous Australians[9], with the gap between Indigenous and non-Indigenous life expectancy at birth in 2009 estimated to be 11.5 years for men and 9.7 years for women[10]. About 12% of the burden of disease[9] and 20% of indigenous deaths[11] are attributable to tobacco use. In 2008, the Australian government made a commitment to "closing the gap" in Indigenous Australian life expectancy[12], including addressing smoking[13].

Documenting tobacco use in Indigenous communities has typically relied on self-report in surveys, with a few studies using biochemical markers to verify self-report. Urine cotinine has been used in both urban[14] and remote clinic-based studies[7]. However, this involves the complexities and cost of obtaining and testing a urine sample and does not provide an immediate assessment of smoking. Portable, hand-held Breath Carbon Monoxide (BCO) analyzers are tools used to immediately assess smoking status. They are suitable for both clinical and community-based studies[15‐23] and are being used in a small number of Indigenous Australian settings[24‐26]. The utility of BCO analyzers and optimum BCO cutoff to distinguish smokers and nonsmokers is being investigated in different populations around the world[27‐36] with different cutoff levels recommended in different populations dependent on the intended use of the BCO test. These include: assessing antenatal smoking[15, 16, 36]; clinical or community surveys[17, 22, 27‐29, 32, 33, 37]; validating smoking cessation[18, 20, 30]; assessing passive smoking[17] or environmental pollution[35]; or investigating sociocultural patterns of smoking[23]. There is, however, no guidance for the optimum BCO cutoff level to validate self-reported tobacco smoking in community-based surveys in Indigenous Australian populations.

This paper examines the sensitivity and specificity of the BCO test and the optimal cutoff level to distinguish between smokers and nonsmokers in three remote Indigenous populations.

Among the 400 people interviewed, 300 (75%) reported they smoked tobacco, and 100 (25%) reported they did not. Four of the 400 interviewed explicitly refused a BCO test, and 19 were in such poor health that a BCO test would have been unnecessarily intrusive. BCO was not tested in a further 57 people interviewed primarily because they had no time to take the BCO test (n = 40) or because a BCO analyzer was not available at the time of interview (n = 17). The remaining 320 people who provided both a BCO test and information about their tobacco use included 260 who self-reported they smoked tobacco and 60 people who self-reported they did not smoke tobacco. Eleven of the 260 were occasional tobacco smokers and reported they had not smoked within the preceding 24 hours. In accordance with inclusion criteria, these 11 occasional smokers were not included in the analysis. Therefore, BCO tests for 249 self-reported smokers and 60 self-reported nonsmokers were analyzed. The proportions of self-reported smokers (81% = 249/309) and nonsmokers (19% = 60/309) in this subsample were similar to proportions of self-reported smokers (75%) and nonsmokers (25%) in the sample overall (|z| = 1.90, P = 0.057).

These findings indicate that BCO can be effectively used to validate self-reported smoking in remote Australian Indigenous communities. The strong agreement between self-reported smoking and BCO indicates that self-reported smoking can be considered a reliable measure in this population.

Limitations of the study include that the sample was not randomly selected and so results cannot be generalized. However, participants were recruited to reflect each community's age and gender characteristics. Although BCO was not tested in all participants, the gender composition and proportions of self-reported smokers among those who provided a BCO test were similar to the sample overall. Confidence in the results is further reinforced by the high level of agreement between BCO and self-report, by the similarly high smoking rates found in other studies in the region, and because community members themselves informed researchers that results reflected their own family and community experience.

The community-based survey method recorded self-reported smoking status and immediately returned BCO test results to study participants. This allowed discrepancies between self-report and BCO to be investigated at the time of interview with further questions about smoking history or pattern of use, which assisted in refining the data.

Ten (4%) self-reported smokers had BCO below the ≥7 ppm cutoff. In nine of these, the self-reported time since last cigarette was between 2 and 12 hours prior to BCO test. Three of these nine had not smoked since the previous evening. Given that BCO has a half-life of between 3 and 4 hours and can decline by 2.1 to 7.5 ppm per hour, depending on the initial BCO level[44], there was sufficient time for BCO to decline below the cutoff. Smokers consistently reported periods of heavy and light smoking with a greater amount of tobacco smoked in the first few days after fortnightly paydays. Similar to smoking patterns documented in other Indigenous communities[42], the majority of participants reported regularly running out and frequently requesting tobacco from family and friends. This provides a plausible explanation for self-reported tobacco smokers who only smoke occasionally and/or have low BCO. Although those who smoke few cigarettes per day can also have normal BCO[21, 23, 44], lapsed time since last cigarette, independent of the number of cigarettes smoked, accounted for most self-reported smokers with low BCO in this study.

The study also documented four (7%) self-reported nonsmokers with BCO ≥7 ppm. While such discrepancies may allude to false self-report or exposure to secondhand smoke, two of these four results could be accounted for by cannabis use. It is possible that exposure to secondhand smoke contributed to the small number of false positives in this study, however cannabis use in this population is likely to be an important factor. Studies of cannabis use in similar communities in the same region indicate that most cannabis users (94%) blend cannabis with tobacco and smoke the mixture in handmade "bucket bongs," with tobacco smokers about 19 times more likely than nonsmokers to also smoke cannabis[45]. Given that up to two-thirds of males and half of females regularly use cannabis in the region's communities[45, 46], cannabis use should be considered where self-reported nonsmokers show high BCO in further self-reported smoking validation studies. The present study did not systematically collect data about cannabis use at interview. A study investigating both tobacco and cannabis smoking requires a different approach, including suitable protocols designed to minimize the ethical and legal risks of studying illegal behaviors in these disadvantaged and disempowered populations in Australia[45].

Only a few studies in Indigenous Australian populations have used biomarkers to verify self-reported smoking. A community-based study in an urban population recorded 10% of self-reported nonsmokers with BCO above cutoff of 9 ppm[26], while a clinic-based study in a remote population tested BCO but did not analyze discrepancies[25]. Two clinic-based studies using urine cotinine, one remote and one urban, both found higher discrepancies. The remote study found 15% of self-reported nonsmokers produced levels above cutoff (539 nmom/l)[7], while the urban study found 17% of self-reported nonsmoking pregnant women produced levels above cutoff (250 ng/ml)[14]. In the study reported here, 7% of self-reported nonsmokers had BCO above cutoff (≥7 ppm). Caution is, however, required before making direct comparison between these studies given the different population groups, recruitment methods, sample size, biomarkers, and study context.

A range of BCO cutoff levels has been used to validate self-reported smoking in different population groups around the world. Cutoffs as high as 10 ppm have been used[20, 47, 48]. Others have used 9 ppm[22, 26, 37, 44]; 8 ppm[19]; 7 ppm[28]; 6.5 ppm[17]; or 6 ppm[21]. Several studies recommend cutoffs as low as 2 to 3 ppm[29, 31, 36]. However, self-reported smoking status and BCO level can vary between ethnic groups in the same location[23]. This suggests possible cultural or communication differences in responses to questions about smoking, a challenge well-known in remote Aboriginal communities[7, 39]. Different sociocultural patterns of smoking may mean different BCO cutoffs are required in different populations. In this study population, reducing the BCO cutoff from ≥7 ppm to ≥5 ppm would increase the self-reported smokers verified from 96.0% to 99.6%, while self-reported nonsmokers verified would remain unchanged at 93.3%. With better information in future studies about those who smoke cannabis and not tobacco in these communities, the proportion of self-reported nonsmokers of tobacco verified could be as high as 96.6%.