Cigarette Smoke Exposure and the Acute Respiratory Distress Syndrome

Abstract

Objective

The association between cigarette smoke exposure and the acute respiratory distress syndrome (ARDS) in patients with the most common ARDS risk factors of sepsis, pneumonia, and aspiration has not been well-studied. The goal of this study was to test the association between biomarker-confirmed cigarette smoking and ARDS in a diverse cohort.

Design, Setting, Patients

We obtained smoking histories and measured urine NNAL (a biomarker of cigarette smoke exposure) in 426 patients with ARDS risk factors (excluding trauma and transfusion) in a prospective cohort of critically ill patients at a single tertiary care center and tested the association between smoking and ARDS.

Interventions

None.

Measurements and Main Results

The association between cigarette smoke exposure and ARDS differed based on ARDS risk factor (p<0.02 for interaction). In patients with non-pulmonary sepsis as the primary ARDS risk factor (n=212), 39% of those with ARDS were current smokers by history, compared with 22% of those without ARDS (odds ratio 2.28 (95% CI 1.24–4.18); p=0.007). Likewise, cigarette smoke exposure as measured by urine NNAL was significantly associated with ARDS in this group. The increased risk of ARDS in non-pulmonary sepsis was restricted to patients with NNAL levels consistent with active smoking and was robust to adjustment for other ARDS predictors. Cigarette smoke exposure as measured by history or NNAL was not associated with ARDS in patients with other risk factors (e.g. pneumonia, aspiration).

Conclusions

Cigarette smoking measured both by history and by biomarker is associated with an increased risk of ARDS in patients with non-pulmonary sepsis. This finding has important implications for tobacco product regulation and for understanding the pathogenesis of ARDS.

INTRODUCTION

Despite advances in supportive care, the acute respiratory distress syndrome (ARDS) remains a major cause of morbidity and mortality in critically ill patients, with recent mortality rates of 20–40%.(1) Furthermore, the search for specific pharmacologic therapies for this syndrome remains fruitless. As a result, increased attention has been focused on the development of preventative approaches, exemplified by the recent shift in focus of the National Institutes of Health’s ARDS Clinical Trials Network to prevention and early treatment.(2)

One approach to ARDS prevention is to focus on potentially modifiable risk factors (“first hits”) that may increase the risk of developing non-cardiogenic pulmonary edema in the setting of a “second hit” ARDS risk factor such as sepsis, pneumonia, aspiration, or trauma. In animal models and some human studies, cigarette smoke exposure causes significant alterations in lung epithelial and endothelial function similar to those observed in ARDS.(3, 4) Likewise, cigarette smoking has potent effects on neutrophil trafficking and function, humoral and cell-mediated immunity, and alveolar macrophages that could contribute to ARDS development.(5, 6) We previously reported that cigarette smoke exposure is associated with substantially increased risk of ARDS after severe blunt trauma or blood transfusion.(7, 8) In contrast, the association between cigarette smoke exposure and ARDS in broader populations of at-risk patients, including those with the most common ARDS risk factors of sepsis, pneumonia, and aspiration, has not been well-studied in rigorous, prospective analyses with quantitative measures of exposure.

Defining the association between smoking and ARDS also has significant implications for tobacco product regulation. Prior analyses of the economic and public health effects of smoking have not accounted for most short-term pulmonary effects of cigarette smoke exposure, including ARDS.(9–11) With the large immediate health and financial burden of ARDS and the potentially rapid impact of regulation-related changes in cigarette smoke exposure on the incidence of ARDS, quantifying the association between smoking and ARDS may have important implications for the cost-benefit analysis of proposed regulations and their subsequent impact on health of the population.(10, 12)

This study was designed to test the hypothesis that cigarette smoke exposure is associated with an increased risk of ARDS in a diverse sample of critically ill patients with a variety of common ARDS risk factors. Because cigarette smoke exposure histories in critically ill patients may be inaccurate,(13) we supplemented smoking history with quantitative measurement of urinary total 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL), a well-validated marker of cigarette smoke exposure.(14) Some of these data have been previously presented in the form of an abstract.(15)

METHODS

Subjects

This analysis was a prospective sub-study within the Validation of biomarkers in Acute Lung Injury for Diagnosis (VALID) study,(16) a prospective cohort of critically ill patients at Vanderbilt University Medical Center, a tertiary medical center in Nashville, Tennessee. The inclusion and exclusion criteria for VALID have been described previously(16) and are summarized in the online supplement. The informed consent process has also been described previously.(16) The study was approved by the Vanderbilt Institutional Review Board.

For this sub-study, we included patients with an identified risk factor for ARDS who were enrolled in VALID within one week of hospital admission between September 2007-December 2009 and had a urine sample available for NNAL measurement (n=896). We excluded patients whose primary ARDS risk factor was severe trauma (n=362) or blood transfusion (n=108), due to the established link between cigarette smoke exposure and ARDS in these populations(7, 8) and potential clinical and biological differences between trauma-associated ARDS and other forms of ARDS, leaving 426 patients for analysis (Figure S1).(17)

ARDS and Risk Factor Definitions

Risk factors for ARDS were categorized as sepsis, pneumonia, aspiration, pancreatitis, near drowning, drug overdose, or other and were adjudicated by the VALID principal investigator (LBW). Sepsis was defined by consensus definitions.(18) Patients with sepsis due to pneumonia were classified as having pneumonia as their ARDS risk factor. ARDS was assessed daily through ICU day 5 using American European Consensus Conference definitions (i.e. PaO2/FiO2 ratio < 300) by two-physician review of radiographs and charts to exclude pure cardiogenic pulmonary edema.(19) Both mechanically ventilated and not mechanically ventilated patients with ARDS were included. If an arterial blood gas was not available, the SpO2/FiO2 ratio was used to assess the level of hypoxemia.(20)

Smoking and Alcohol History

Smoking and alcohol use histories, including administration of the Alcohol Use Disorders Identification Test), were obtained from patients or surrogates as detailed in the Supplement. If patients or surrogates were unable or unavailable to provide this history, it was obtained from the medical record.

Urine NNAL

NNAL is a metabolite of NNK (nitrosamine 4-(methylnitrosamino)-1-(3-pyridly)-1-butanone, or nicotine-derived nitrosamine ketone), a potent carcinogen found only in tobacco products. Urine NNAL is a well-established, highly specific marker of NNK uptake with a half-life of 10–18 days. (21–23) Because of its long half-life, NNAL detects tobacco exposure for many weeks and serves as an integrated marker of exposure over time. NNAL levels were measured in urine collected at study enrollment using liquid chromatography-tandem mass spectrometry.(24) Urine NNAL levels ≥ 47.3 pg/ml were considered consistent with active smoking, and NNAL levels < 47.3 pg/ml but above the limit of quantitation (1 pg/ml) were considered consistent with passive smoke exposure, based on the finding in 601 outpatients that a urine NNAL cutoff of 47.3 pg/ml accurately distinguishes active from passive smokers (sensitivity 87.4%, specificity 96.5%).(14) Because the urine may be variably concentrated in critical illness, as a sensitivity analysis, NNAL levels were normalized to urine creatinine, for which a cutoff of 64 pg NNAL/mg Cr was used to distinguish active from passive exposure.(25)

Statistical Analysis

Full details of the statistical methods are in the Supplement. Because NNAL is not normally distributed, NNAL levels were log-transformed or classified into a categorical variable for regression analysis. Multivariable logistic regression was performed to adjust the association between smoking (as measured by either history or by NNAL) and ARDS for established predictors of ARDS, specifically diabetes, race-ethnicity, alcohol abuse, and severity of illness.(26) In addition, we controlled for time elapsed between hospital admission and enrollment because NNAL levels decline over time.

RESULTS

Study Subjects

Overall, there were no differences in baseline demographics between patients with and without ARDS (Table 1). As in prior studies, ARDS subjects had higher severity of illness and a lower prevalence of diabetes than those without ARDS. While the proportion of patients with a history of current smoking was numerically higher in those with ARDS than those without ARDS (36% vs 29%), this difference was not significant (p=0.12).

Table 1.

Description of study population (n=426) stratified by ARDS

Characteristic	Patients with ARDS (n=216)	Patients without ARDS (n=210)	p-value
Age	56 ± 16	56 ± 15	0.86
Male gender	114 (53%)	102 (47%)	0.54
White race	191 (88%)	170 (81%)	0.23
APACHE II score	28 ± 8	26 ± 8	0.003
Comorbidities*
Diabetes	61 (28%)	82 (39%)	0.02
Chronic liver disease	14 (7%)	17 (8%)	0.52
Chronic kidney disease	30 (14%)	34 (16%)	0.51
History of current smoking	77 (36%)	60 (29%)	0.12
Charted alcohol abuse	42 (19%)	43 (21%)	0.79
Current alcohol abuse**	24 (11%)	37 (12%)	0.66
Smokeless tobacco use	3 (1.4%)	3 (1.4%)	0.44
Source of ICU admission
Emergency Room	70 (32%)	94 (45%)
Floor	65 (30%)	37 (18%)	0.01
Outside hospital	59 (27%)	51 (24%)
Operating Room	19 (9%)	25 (12%)
Primary ARDS Risk
Factor	86 (40%)	126 (60%)
Non-Pulmonary Sepsis	74 (34%)	42 (20%)
Pneumonia	1 (0.5%)	9 (4%)	<0.001
Pancreatitis	2 (1%)	13 (6%)
Drug Overdose	44 (20%)	20 (10%)
Aspiration	9 (4%)	0 (0%)
Other

^*No difference in: Metastatic malignancy, leukemia, COPD, asthma, organ or bone marrow transplantation, HIV, coronary artery disease

^**Charted alcohol abuse includes current or former. Current alcohol abuse includes charted history of current abuse or high score on AUDIT questionnaire

In keeping with the diverse nature of this cohort, there were significant differences in source of ICU admission between those with and without ARDS (p=0.01). Specifically, those with ARDS were more likely to have been admitted to the ICU from the hospital floor and less likely to have been admitted from the emergency room. Similarly, there were substantial differences in ARDS risk factor between those with and without ARDS (p<0.001), with pneumonia and aspiration making up a larger proportion of those with ARDS than those at risk for ARDS.

Prevalence of cigarette smoke exposure by NNAL

NNAL levels were consistent with active smoking (≥47.3 pg/ml) in 38% of the cohort, passive smoking (above LOQ to 47.3 pg/ml) in 28%, and no cigarette smoke exposure in 34% (below LOQ). For reference, the prevalence of cigarette smoking in Tennessee in 2008 was 23%.(27) The distribution of NNAL levels by smoking history is shown in Figure 1. As in our prior studies,(7, 13) subjects identified by history as current smokers were highly likely to be confirmed as active smokers by NNAL; in addition, NNAL levels consistent with active smoking were not uncommon in subjects identified by history as non-smokers or former smokers, or whose smoking status was unknown.

Figure 1. NNAL Levels By Smoking History in Entire Cohort (n=426).

Urine NNAL levels ≥ 47.3 pg/ml were considered consistent with active smoking.

Association of cigarette smoke exposure with ARDS

In the overall cohort of 426 patients, there was no statistical association between NNAL levels and ARDS. This finding was consistent whether NNAL was analyzed as a continuous variable (p=0.67) or as a categorical variable (p=0.89).

Because there were significant differences in ARDS risk factors between subjects with and without ARDS, we tested for the presence of multiplicative interaction between cigarette smoke exposure and ARDS risk factors. There was a strong interaction between non-pulmonary sepsis as an ARDS risk factor and cigarette smoke exposure that was present whether cigarette smoke exposure was classified by smoking history (p=0.01) or by NNAL (p=0.004). Specifically, in patients with non-pulmonary sepsis (n = 212, excluding pneumonia and aspiration), both current smoking by history (odds ratio [OR] for ARDS 2.28; 95% CI 1.24–4.19; p=0.008) and increasing NNAL levels (OR per one-log increase 1.13; 95% CI 1.00–1.27; p=0.047) were associated with development of ARDS. In this group, 39% of subjects with ARDS were current smokers by history, compared with 22% of subjects without ARDS (odds ratio 2.28 (95% CI 1.24–4.18); p=0.007). These patients are further described in Table S1. In patients with other ARDS risk factors, there was no association between ARDS and either a history of smoking (OR 0.79, 95% CI 0.45–1.40; p=0.42) or urine NNAL (OR 0.91; 95% CI 0.82–1.02; p=0.11).

Multivariable Models

To determine whether the association between smoking and ARDS in non-pulmonary sepsis could be confounded by demographic or clinical factors associated with ARDS, we created multivariable logistic models to adjust for other ARDS predictors. The association between NNAL and ARDS in patients with non-pulmonary sepsis was not significantly attenuated by adjustment for severity of illness as measured by APACHE II score, race-ethnicity, diabetes, or alcohol abuse (Table 2). Analyses using categorized NNAL as the primary predictor showed similar results (Table S2). Likewise, the association between smoking history and ARDS was robust to adjustment for these predictors (Table 3).

Table 2.

Association between NNAL (Continuous) and ARDS, Unadjusted and Adjusted, In Patients with Non-pulmonary Sepsis (n=212)

Predictors of ARDS	Odds Ratio for ARDS (95% CI)	p-value
UNADJUSTED MODEL
Log-NNAL, per one-log increment	1.13 (1.00–1.27)	0.047
MULTIVARIABLE MODEL*
Log-NNAL, per one log increment	1.16 (1.01–1.33)	0.037
APACHE II score	1.09 (1.04–1.13)	<0.001
African-American race	0.34 (0.13–0.91)	0.03
Hispanic ethnicity	0.31 (0.03–3.20)	0.33
Current alcohol abuse	1.77 (0.62–5.08)	0.29
Diabetes	0.65 (0.35–1.23)	0.19
Time elapsed between admission and enrollment	1.19 (0.99–1.44)	0.06

^*Referent ethnicity is white.

Table 3.

Association between Smoking History and ARDS, Unadjusted and Adjusted, In Patients with Non-pulmonary Sepsis (n=212)

Predictors of ARDS	Odds Ratio for ARDS (95% CI)	p-value
UNADJUSTED MODEL
Current smoker by history*	2.28 (1.24–4.19)	0.008
MULTIVARIABLE MODEL**
Current smoker by history*	2.15 (1.07–4.32)	0.03
APACHE II score	1.08 (1.04–1.13)	<0.001
African American race	0.38 (0.14–1.01)	0.053
Hispanic ethnicity	0.38 (0.04–3.97)	0.42
Current alcohol abuse	1.77 (0.62–5.08)	0.29
Diabetes	0.69 (0.37–1.30)	0.25
Time elapsed between admission and enrollment	1.18 (0.98–1.42)	0.09

^*Referent group is all current non-smokers by history, including former smokers, and those with unknown smoking history.

^**Referent race-ethnicity is white.

Of note, using previously validated NNAL cutpoints, the risk of ARDS was restricted to active smoking in both unadjusted and adjusted models (Table S2). A smoothed Lowess scatterplot of the association between NNAL levels and ARDS is shown in Figure 2. This analysis suggests that increased ARDS risk may begin at levels of NNAL consistent with moderate to heavy passive smoke exposure.

Figure 2. Association between NNAL and ARDS in patients with non-pulmonary sepsis (n=212).

This figure demonstrates a locally weighted scatterplot smoothing (Lowess) analysis in which the continuous relationship between NNAL levels on the x-axis (log-scale) and development of ARDS on the y-axis is depicted (line). Dots represent individual patients; those at the top of the figure have ARDS, those at the bottom do not.

Cigarette Smoke Exposure and ARDS Outcomes

As a secondary analysis, we tested whether cigarette smoke exposure was associated with severity of lung injury and/or clinical outcomes in subjects with ARDS. In both the non-pulmonary sepsis subset and the overall cohort, cigarette smoke exposure as measured by NNAL was not associated with ARDS severity (Table S3). Furthermore, as in our prior study of the association between NNAL levels and ARDS outcomes,(28) cigarette smoke exposure as measured by NNAL was associated with significantly lower mortality in ARDS patients (Table S4). Also as in our prior work, the association between smoking and lower mortality was no longer statistically significant after adjusting for other predictors of mortality, including age, APACHE score, comorbidities, code status on ICU admission, and shock (Table S5).

Sensitivity Analysis Correcting NNAL for Creatinine

As described in the methods, because the urine may be variably concentrated in critical illness, urine NNAL levels were normalized to urine creatinine for sensitivity analysis. These analyses revealed no substantive differences in results (data not shown).

DISCUSSION

To our knowledge, this analysis is the first to examine prospectively the risk of ARDS associated with biomarker-confirmed cigarette smoke exposure in critically ill subjects with a variety of ARDS risk factors. Cigarette smoke exposure as assessed by either smoking history or urinary NNAL levels was associated with approximately double the odds of developing ARDS in subjects with non-pulmonary sepsis, even after controlling for other ARDS predictors including alcohol abuse, diabetes, and severity of illness. No association was detected in subjects at risk for ARDS from pneumonia or aspiration. As sepsis is both common and increasing in incidence, these findings have important implications for ARDS prevention, public health, regulation of tobacco products, and the global burden of disease attributable to tobacco.

The association between smoking and ARDS has strong biologic plausibility in light of considerable mechanistic research on the effects of smoking on inflammation and lung epithelial and endothelial function. Smoking enhances both lung epithelial and endothelial permeability,(3, 4, 29) modifies the quantity and function of pulmonary neutrophils and alveolar macrophages,(5, 6) and promotes platelet dysfunction,(30) pathways central to the pathogenesis of ARDS. In 298 explanted human lungs rejected for transplantation, we found that smokers had more pulmonary edema as measured by lung weight than non-smokers; moreover, smokers with the highest level of exposure had impaired alveolar fluid clearance.(31) These experimental studies provide supportive evidence that the association identified in this study between smoking and ARDS in non-pulmonary sepsis is important and likely to be causal.

Why might the association between smoking and ARDS differ based on ARDS risk factor? For one, ARDS due to non-pulmonary sepsis (an “indirect” lung injury) may be biologically different from ARDS due to “direct” lung injury (e.g. pneumonia and aspiration). We recently reported that ARDS due to indirect lung injury from non-pulmonary sepsis is characterized by more severe endothelial injury and inflammation when compared with ARDS due to direct lung injury, which is characterized by more severe lung epithelial injury.(32) If smoking primes the lung to develop ARDS via enhanced endothelial permeability, as suggested by some experimental studies,(3) the effects of smoking on ARDS susceptibility may be more pronounced in indirect lung injury, as in non-pulmonary sepsis. Alternatively, it may be that the effects of smoking on lung epithelial permeability(4, 29) combined with increased endothelial permeability due to severe sepsis culminate in poor barrier function across the alveolar-capillary interface and a higher propensity to develop the increased permeability pulmonary edema that characterizes ARDS.

A second potential explanation for the interaction between smoking and ARDS risk factors derives from reported associations between smoking and increased susceptibility to bacterial pneumonia and influenza.(6) Since smoking is already associated with an increased risk of pneumonia, any smoking-associated increased risk of ARDS in pneumonia patients may either be negligible or difficult to detect. Further, because development of ARDS in patients with pneumonia leads to ICU admission, many patients with less severe pneumonia (and no ARDS) were likely never admitted to the ICU and therefore would not have been enrolled in this study, potentially obscuring associations between smoking and ARDS susceptibility. Larger studies in a broader cohort of hospitalized pneumonia patients will be needed to further evaluate the association between smoking and ARDS in this subgroup.

In addition to identifying a link between smoking and ARDS in non-pulmonary sepsis, these analyses confirmed two of our prior findings regarding cigarette smoke exposure in critical illness. First, the prevalence of smoking in ICU patients and ARDS patients in this sample is higher than the population average.(7, 13, 28) Since NNAL levels decline over time, and subjects were enrolled on average 1–2 days after hospital admission (and concomitant smoking cessation), the prevalence of smoking measured at admission might be even higher. Second, these data confirmed our prior finding that biomarker-confirmed smoking status is associated with lower unadjusted mortality in ARDS.(28) In this study as in our prior work, we found that this lower mortality in smokers seemed to be largely explained by younger age, fewer comorbidities, and lower severity of acute illness (Table S5). A similar association, termed the “smoker’s paradox,” is observed in acute myocardial infarction (MI) and has been attributed in part to the younger age and relative paucity of comorbidities of smokers with acute MI compared with non-smokers, similar to the pattern observed in ARDS.(28, 33)

The finding that smoking is associated with development of ARDS in non-pulmonary sepsis has important implications for tobacco product regulation. The immediate effects of cigarette smoke exposure on acute cardiopulmonary disease costs and outcomes have not been well-represented in economic models, including those used in the United States Food and Drug Administration (FDA)’s cost-benefit analyses of cigarette warning labels(10) and its 2014 proposed “deeming” rule in which the FDA asserts jurisdiction over cigars, e-cigarettes, and other products.(11) Because these models time-discount costs, inclusion of immediate changes in disease burden, such as posed by ARDS, may have a substantial effect on the discounted present value of reductions in smoking. Further research will be needed to define the dose-response curve and time course of the association between smoking and sepsis-associated ARDS to fully inform regulatory decisions.

This study has several strengths, including the use of a validated biomarker to quantify cigarette smoke exposure, measurement of alcohol use with a validated survey instrument, a diverse patient sample, and a prospective design focused on testing the hypothesis under study. This study also has some limitations. First, inclusion in this analysis was predicated on the urine availability, and anuric patients were excluded. The effect of acute kidney injury (AKI) on urine NNAL excretion is unknown. To mitigate this concern, we conducted sensitivity analyses using NNAL adjusted for urine creatinine, which produced similar results (not shown) and is consistent with approaches used to analyze other urine biomarkers in AKI.(34) Furthermore, adjustment of the primary analyses for AKI or abnormal renal function did not substantively affect the results (not shown). Second, urine NNAL can be elevated by use of smokeless tobacco or other inhaled tobacco products. In this sample, 2% of subjects had a history of using smokeless tobacco; exclusion of these subjects did not change the findings. However, it is possible that some elevations in urine NNAL were due to smokeless tobacco use not captured by history; the Centers for Disease Control recently estimated the prevalence of smokeless tobacco use in Tennessee at 6.4%.(35) Third, the number of subjects with nonpulmonary sepsis and NNAL levels consistent with passive smoking is modest (n=58); thus, whether passive smoking is associated with some increase in the risk for ARDS (as suggested by Figure 2) or not (as suggested by data in Table S2) remains unclear, and further study would require a larger group of patients. Fourth, although our focus on non-pulmonary sepsis was strongly supported by highly significant tests of interaction for both a history of smoking (p=0.01) and NNAL levels (p=0.004), as a sub-group analysis, it should be interpreted with some caution and would benefit from external validation in an independent cohort. Finally, while urine NNAL is clearly a useful biomarker for clinical research purposes, it does not at present have utility in clinical practice in the ICU.

In conclusion, in this prospective cohort of heterogeneous critically ill patients at risk for ARDS, cigarette smoking is associated with a substantial increase in the risk of developing ARDS in non-pulmonary sepsis, independent of other ARDS predictors. With the widespread prevalence of both smoking (particularly in the developing world) and sepsis, these findings have important implications for public health, tobacco product regulation, and for the global burden of disease related to tobacco. Furthermore, these findings may have important implications for prevention of ARDS, particularly if the mechanisms by which smoking promotes the development of acute lung injury can be identified. Future research should focus on replicating and quantifying the association between smoking and ARDS in other diverse ARDS cohorts, including hospitalized non-ICU patients, and on identifying the underlying mechanisms of smoking-associated ARDS in order to facilitate targeted treatment strategies.