To our knowledge, this is the first study to compare quantitatively assessed cigarette smoke exposure with smoking history in critically ill adults. The results indicate that smoking history markedly underestimates cigarette smoke exposure in the critically ill. Nearly all of the nonsmokers by history had evidence of either active smoking (21%) or SHS exposure (75%) by serum cotinine or urine NNAL levels or both. These findings have major implications for future studies on the effects of cigarette smoke exposure on critical illness, and for hospitalized patients more generally.
To date, few studies have been published on the effects of recent tobacco smoke exposure on critical illness, largely because of difficulty obtaining an accurate smoking history. In epidemiologic studies, even apparently minor degrees of misclassification of exposure can significantly bias the results (15
); thus, accurate assessment of true active smoking rates and SHS exposure is essential for future research on the effects of smoking on critical illness. Because of excellent sensitivity and specificity, the measurement of cigarette smoke exposure biomarkers has been recommended to obtain a more precise estimation of active cigarette and SHS exposure in outpatient studies (1
). Most studies to date in the critically ill have used smoking history (in many cases obtained from the chart) and only reported active cigarette smoke exposure (17
). The results of our study show that biomarkers of cigarette smoke exposure substantially increase the detection of active smoking compared to smoking history in this understudied population. Furthermore, we demonstrate for the first time to our knowledge that SHS exposure in the critically ill can be quantified by cigarette smoke biomarkers and is highly prevalent.
To study the effects of recent tobacco smoke exposure on the critically ill, it will be important to assess the intensity of exposure over time relative to the course of illness, because smoking behavior may change in the setting of acute illness. Once hospitalized and no longer exposed to cigarette smoke, short-lived biomarkers (i.e., serum cotinine) will decrease quickly relative to markers with longer half-lives, and light or intermittently active smokers may be misclassified as nonsmokers with SHS exposure. Urine NNAL will be particularly useful in this setting because it has a longer half-life, and classification of exposure by urine NNAL levels should remain accurate even if specimens are collected days after admission. As evidence of this point, five (17%) smokers by history were classified as active smokers by urine NNAL levels but had serum cotinine levels in the SHS exposed range. Overall, urine NNAL detected higher levels of exposure than did serum cotinine in 13 subjects (22%).
The prevalence of active smokers in this cohort of critically ill patients (57%) was higher than the prevalence of active smokers in a cohort of urban hospitalized patients (46%) (19
). It was also remarkably higher than the Tennessee average (23%) and the nationwide average (20%) (20
). Previous studies have measured higher rates of smoking in subjects with lower socioeconomic status and in subjects with substance abuse problems (19
). Compared to the national population, our cohort had fewer persons younger than age 65 yrs with private insurance (23% vs. 68%) and a higher rate of illicit drug use (17% vs. 8.3%) (23
). Thus, the lower socioeconomic status and higher illicit drug use of our patient population may explain, at least in part, the high prevalence of active smoking. Another possible explanation for the high prevalence of active cigarette smoke exposure in this population is that recent smoking is itself a risk factor for the development of critical illness.
The prevalence of SHS exposure in nonsmokers by history in this cohort of critically ill patients (75%) is also markedly higher than the estimated nationwide prevalence of 43% generated from National Health and Nutrition Examination Surveys data, a population-based study of nearly 30,000 subjects that used serum cotinine to identify SHS exposure (25
). Despite the higher serum cotinine cut-point used in the National Health and Nutrition Examination Surveys study (10 ng/mL vs. 3.1 ng/mL), the prevalence of SHS exposure was still lower than that of our critically ill cohort. Also, the limit of quantitation for serum cotinine in the National Health and Nutrition Examination Surveys study was 0.05 ng/mL for samples analyzed from 1988 to 2000 and 0.015 ng/mL for samples analyzed from 2001 to 2002, as compared to 0.02 ng/mL in our study. For comparison, we analyzed our samples using a limit of quantitation of 0.05 ng/mL and found no difference in means and percentiles. Thus, our use of a more sensitive serum cotinine assay was unlikely to have contributed to the high prevalence of SHS exposure in this population. The prevalence of smoking among adults in Tennessee is similar to the nationwide prevalence (23% vs. 20%) (20
), and a public smoking ban in Tennessee was instituted for 4 of the 5 months in which our study was conducted. Therefore, it is unlikely that geographic differences contribute to the higher prevalence of SHS exposure in this study population. Another possible explanation for the high prevalence of SHS exposure is the inclusion of occasional or “social” smokers in the nonsmokers by history group. Assessing for occasional smoking requires a detailed interview with the primary subject, which is not feasible in most critically ill patients. Last, cotinine cutoff values have been shown to vary by ethnicity. Although 93% of our cohort was white, sensitivity analysis showed no difference in classification of cigarette smoke exposure with the nationally representative cutoff (3.1 ng/mL) used in this study compared to a cutoff specific for whites (4.85 ng/mL) (10
). The high prevalence of SHS exposure in this cohort may be explained in part by the addition of urine NNAL, which has a longer half-life and would detect SHS exposure long after cotinine levels declined to below the limit of quantitation. Also, SHS itself could be a risk factor for the development of critical illness, an important topic that merits further study.
It should be emphasized that no gold standard exists for measuring cigarette smoke exposure, although, historically, self-report has been used as the standard for comparison. Self-reported smoking status is associated with a number of limitations, including recall bias, social desirability bias, and poor accuracy regarding quantitative aspects of exposure (8
). In the critically ill, these problems are further compounded by altered mental status and respiratory failure such that most subjects are unable to provide a smoking history. Because biomarkers of cigarette smoke exposure are highly specific to tobacco products, previous epidemiologic studies have used biomarkers of cigarette smoke exposure to improve the accuracy of self-report (11
). The results of this study confirm that quantitative biomarker data provide significantly more detailed and objective information on tobacco smoke exposure compared to smoking history in critically ill subjects.
Our study has some limitations. First, this study’s subjects were predominantly white and were enrolled from one center. Although the use of such a homogeneous sample may limit the generalizability of these results, it also eliminates the confounding effect of racial differences on nicotine metabolism (28
). Second, both smokeless tobacco and nicotine replacement therapy can raise nicotine levels, potentially confounding interpretation of cigarette smoke exposure; however, nicotine replacement therapy does not affect NNAL levels (13
). One subject in our study had a history of smokeless tobacco use, and interpretation of that subject’s elevated serum and urine cotinine and NNAL levels is uncertain. Third, it is unclear if renal function or medications administered in the intensive care unit interfere with the detection of urine NNAL. The method used to measure urine NNAL in this study is the most sensitive method reported to date and optimizes separation from interfering substances in the sample by converting the analyte to a relatively nonpolar derivative (29
). In our 60 critically ill subjects, of whom 38% had acute or chronic renal failure, one subject with acute renal failure had interference. We adjusted all urine NNAL levels for urine creatinine concentration, and rank sum analysis showed that there was no significant difference between urine NNAL levels in patients with acute or chronic renal insufficiency compared with those with normal renal function in our study population.