Assessing recent smoking status by measuring exhaled carbon monoxide levels

Abstract

Background: Cigarette smoke causes both acute and chronic changes of the immune system. Excluding recent smoking is therefore important in clinical studies with chronic inflammation as primary focus. In this context, it is common to ask the study subjects to refrain from smoking within a certain time frame prior to sampling. The duration of the smoking cessation is typically from midnight the evening before, i.e. 8 hours from sampling. As it has been shown that a proportion of current smokers underestimates or denies smoking, objective assessment of recent smoking status is of great importance. Our aim was to extend the use of exhaled carbon monoxide (CO(breath)), a well-established method for separating smokers from non-smokers, to assessment of recent smoking status.

Methods and findings: The time course of CO(breath) decline was investigated by hourly measurements during one day on non-symptomatic smokers and non-smokers (6+7), as well as by measurements on three separate occasions on non-smokers (n = 29), smokers with normal lung function (n = 38) and smokers with chronic obstructive pulmonary disease (n = 19) participating in a clinical study. We used regression analysis to model the decay, and receiver operator characteristics analysis for evaluation of model performance. The decline was described as a mono-exponential decay (r(2) = 0.7) with a half-life of 4.5 hours. CO decline rate depends on initial CO levels, and by necessity a generic cut-off is therefore crude as initial CO(breath) varies a lot between individuals. However, a cut-off level of 12 ppm could classify recent smokers from smokers having refrained from smoking during the past 8 hours with a specificity of 94% and a sensitivity of 90%.

Conclusions: We hereby describe a method for classifying recent smokers from smokers having refrained from smoking for >8 hours that is easy to implement in a clinical setting.

Conflict of interest statement

Competing Interests: The authors have declared that no competing interests exist.

Figures

Figure 1. Baseline CObreath levels measured on non-smokers, smokers with normal lung function and smokers with COPD.

A. CObreath from smokers and non-smokers recruited for a time course study of CObreath decline (group 1). CObreath (ppm) was measured in the morning; smokers having refrained from smoking during the past >8 hours. ** indicates p<0.01. B. CObreath measured on smokers with normal lung function (“smokers”), smokers with COPD and non-smokers with normal lung function (group 2). CObreath (ppm) was measured in the morning; smokers having refrained from smoking during the past 8 hours. *** indicates p<0.001.

Figure 2. Time course of CObreath decline after smoking one cigarette (group 1).

A. After normalisation against each individual peak value, relative CObreath values were plotted as a function of time since smoking. By non-linear regression, a one phase exponential model was fitted to the decay (Y = (Y0-0.35)*e−0.36×+0.35, r2 = 0.77). B. The natural logarithm of CObreath, ln(CObreath) was plotted versus time since smoking, and the decay was described by linear regression (Y = Y0-0.15×, r2 = 0.70). From the slope, CObreath half-life during the day was estimated to 4.5 hours (ln(2)/slope). The 95% prediction limits are also showed in the figure (dashed lines).

Figure 3. Proposed model for CObreath decline (group 2), and receiver operator characteristics (ROC) analysis for evaluation of model classification performance on smoking subjects with and without COPD.

A. CObreath measurements (smokers) were recorded in the morning. After normalisation ln(CObreath) was plotted versus self-reported time since smoking. To allow comparisons with group 1, measurements performed >8 hours since smoking were omitted. CO decline was modelled by linear regression (solid line) on smokers with normal lung function. From the slope, CObreath half-life was estimated to 4.3 hours (ln(2)/slope). The 95% prediction limits are indicated in the graph as dashed lines. The upper prediction limit (Y = 3.9-0.16×) was evaluated as a cut-off. As test set, CObreath measured on smokers with COPD (n = 19) were included (indicated with triangles). B. CObreath values measured ≤7 hours were used to estimate specificity (classified as positive for recent smoking status), and values measured between 8–10 hours since smoking were used to estimate the sensitivity (classified as negative for recent smoking status). A cut-off of 12 ppm gave a sensitivity of 94% and a median specificity of 97%. Area under the curve (AUC) was 0.96. X-axis: False positive Rate, Y-axis: True positive Rate.