Cross-sectional observational studies using Waves 2 and 3 data from the Population Assessment of Tobacco and Health Study have found an association between e-cigarette use and respiratory symptoms. One longitudinal W3-W4 PATH Study analysis found no relation between exclusive e-cigarette use and incident respiratory symptoms but suggested that dual users of cigarettes and e-cigarettes had significantly higher risk for symptom onset compared to exclusive cigarette users.Finally, one prospective study of young adults found an association between cannabis vaping and respiratory symptoms.There are many design issues that make these studies hard to compare. The clinical importance of the respiratory outcome is not clear in most cases because the multiple wheezing questions are analyzed in isolation from each other, or an endorsement of only one item is considered symptomatic. Many of the studies included adults with COPD, which is a diagnosis strongly linked to a history of cigarette smoking, and many people with COPD have chronic severe wheezing and dyspnea. Another concern is residual confounding: Most of the studies showing an association between e-cigarette use and respiratory symptoms failed to adjust for cigarette smoking history and concurrent marijuana use, both associated with respiratory problems and concurrent e-cigarette use. Finally, few studies addressed alternative tobacco product categories besides e-cigarettes. To better understand these divergent findings on how tobacco product use relates to respiratory health, we analyzed W2 and W3 data from the PATH Study.

We developed a dependent variable that incorporated all available questions on wheezing and nighttime cough and determined cut-off values associated with functional outcomes. We focused on both cross sectional and longitudinal associations between functionally-important respiratory symptoms and ten mutually exclusive tobacco product use categories,grow rack adjusting for past cigarette smoking history and concurrent marijuana use. We also examined results for two different cut-off values for a respiratory symptom index to test for sensitivity to symptom severity. Recruitment for W1 of the PATH Study employed stratified address-based, are a probability sampling with oversampling of adult tobacco users, young adults , and African-Americans. An in-person screener selected youths and adults from households at W1, and audio computer-assisted self-interviews collected data on tobacco use and health outcomes. Respiratory symptoms were assessed in W2 and W3 , including 28,362 and 28,148 adult participants, respectively . Mean time between W2 and W3 adult interviews was 53.8 weeks. Our analyses utilized the adult W2 and W3 Restricted Use Files. We selected all W2 adults without COPD or other nonasthma respiratory diseases . We report a complete case analysis excluding participants lost to follow up at W3 and those with missing data on any variables with final analytic sample of 16,295. PATH Study design and methods,interviewing procedures, questionnaires, sampling, weighting, and response rates are in the PATH Study Restricted Use Files User Guide.All respondents provided informed consent; Westat’s IRB approved the study. The PATH Study utilized the seven wheezing/cough questions from the International Study of Allergies and Asthma in Childhood core wheezing module.Responses to the ISAAC questions were used to create a respiratory symptom index . This index was validated in the PATH Study adult sample based on its internal consistency, test-retest reliability, and its strong association with self-reported physician diagnosis of asthma. Respondents over cut-off values of ≥2 and ≥3 had significantly higher risk for physical limitations, fatigue, and poorer perception of health assessed by items from the Patient Reported Outcomes Measurement Information System .

The full validation of this measure is published elsewhere.Because the validation supported cut-off values of of ≥2 and ≥3, we examined both as a test of the sensitivity of the findings to respiratory symptom severity. Covariates were derived from W1 and W2, and included variables associated with both tobacco exposure and functionally-important respiratory symptoms.Sociodemographic variables included age, sex, race/ethnicity, education, income, and urbanicity. Medical conditions that could result from tobacco use and also cause respiratory symptoms included asthma, congestive heart failure, heart attack, diabetes, cancer, being overweight, and use of antihypertensives known to cause coughing or wheezing . Smoke-related exposures included pack-years of cigarette smoking, second-hand smoke exposure, and marijuana use. Calculating pack years of smoking We were particularly concerned with adjusting results carefully for each individual’s cigarette smoking history, an important predictor of respiratory outcomes. We derived lifetime pack years to account for cigarette smoking history in this analysis. Lifetime pack years is a clinical metric calculated by multiplying the number of packs of cigarettes per day someone smokes by the number of years they have smoked cigarettes. The following text annotates the algorithm to calculate Wave 1 lifetime pack years. Data from Wave 1 lifetime pack years was used in conjunction with variables describing subsequent cigarette use to determine lifetime pack years at W2 and beyond. Never smokers were assigned a pack years value of zero. All questions used in the algorithm and response categories are listed in Supplemental Table 3. Because of routing instructions in the PATH Study interview, only those respondents who said that they have smoked cigarettes “fairly regularly” were asked about how long they have smoked or did smoke . For any respondent at Wave 1 who currently smokes regularly or formerly smoked fairly regularly, lifetime pack years was calculated by multiplying the number of cigarette packs smoked per day by the number of years they have smoked fairly regularly. Two different formulas were used for this calculation, depending on answers to the questions for variable R01_AC9004 and R01_AC9009 .

All main analyses were weighted using the W3 longitudinal full-sample and replicate weights to adjust for the complex sample design and loss to follow up. Variances were estimated using the BRR method with Fay’s adjustment set to 0.3 to increase estimate stability. Pack-years of cigarette smoking, tobacco product P30D frequency variables , and second-hand smoke exposure were Winsorized at the 95th, 95th and 99th percentiles, respectively,microgreens shelving to address outliers.We examined unadjusted associations between tobacco product use at W2 and the presence of functionally-important respiratory symptoms then used multivariable weighted Poisson regression to obtain adjusted risk ratios and 95% CIs for each dichotomous outcome.Next, we evaluated longitudinal associations between W2 tobacco product use and changes in respiratory symptoms from W2 to W3. Symptoms “worsened” if the symptom score was <3 at W2 but > 3 at W3. Symptoms “improved” if the symptom score was ≥ 3 at W2 but < 3 at W3. Finally, we tested the sensitivity of the findings to symptom severity level by rerunning all analyses at a cutoff level of ≥ 2. For each multivariable analysis, post-hoc two-group comparisons were completed to determine if the adjusted risk for each tobacco product use category was significantly different from exclusive cigarette users. All analyses used Stata survey data procedures, version 15.1; standard errors forTables 3 and 4 and estimates for all covariates for Tables 2-4 are included in Supplemental Tables 4-6.At W2, the prevalence of functionally-important respiratory symptoms was 7.2% . Table 1 shows that respiratory symptoms were more common in the four categories of tobacco use that included cigarettes , compared to never tobacco use, and among those who used marijuana. Functionally-important respiratory symptoms were much more common among those with asthma, and also more common among those with comorbid conditions, obesity, and those using medications known to cause coughing or wheezing . Figure 1 illustrates the unadjusted linear relationship between frequency of cigarette use and proportion of persons with functionally-important respiratory symptoms for the four use categories featuring cigarettes. The shape of the dose-response lowess lines were almost identical and the 95th percentile for cigarette use intensity was essentially the same for all four groups, regardless of what other tobacco products were added to cigarettes, emphasizing the importance of cigarettes in these four most prevalent categories of tobacco use. In the full, adjusted, multi-variable cross-sectional model , all four tobacco use categories that featured cigarette smoking were associated with a doubling of the risk of functionally-important respiratory symptoms vs. never tobacco users , and risk for the multiple use categories were not significantly different from exclusive cigarette use .

As illustrated in Figure 2, we observed a significant positive dose-response relationship for current use of cigarettes . Exclusive use of non-cigarette products not associated with added risk Compared to never users, the risk of functionally-important respiratory symptoms were not significantly different for exclusive users of e-cigarette, cigar, hookah and smokeless tobacco; moreover post hoc testing indicated that risk ratios for each of these categories were significantly lower compared to exclusive cigarette use . None of these cross-sectional results changed when the analysis was repeated at a respiratory index cut-off level of ≥2. Testing sensitivity to key confounders of the e-cigarette—respiratory symptom association Cigarette smoking pack-years, second-hand smoke exposure, and marijuana use were also associated with functionally-important respiratory symptoms . Table 2 highlights the importance of cigarette smoking pack-years and past-month marijuana use as confounders of the association between tobacco product use and respiratory symptoms. Cigarette pack-years was a particularly strong confounder; adding this variable alone to the cross-sectional multi-variable model attenuated association estimates for cigarettes and cigarettes+e-cigarettes by 30% and for exclusive e-cigarettes by 25%. That was partly because all three groups had a similarly long cigarette smoking history—weighted mean 13.4 cigarette pack-years for exclusive cigarette smokers, 12.9 for the dual users, and 10.8 for exclusive e-cigarette users. Similarly, 19.2% of exclusive e-cigarette users also currently used marijuana; adding P30D marijuana use to the multi-variable model attenuated association for e-cigarettes by 9%. Adding all three confounders together attenuated the e-cigarette-respiratory symptom association RR from 1.53 to 1.05. The categorical analysis did not address whether functionally-important respiratory symptoms increased with increasing frequency of use. Figure 2 explored this for cigarettes and e-cigarettes, adjusting for cigarette smoking history. For cigarettes, there was a significant linear increase in the percent with functionally-important respiratory symptoms with higher intensity of use; prevalence of respiratory symptoms was less than 5% for never users and over 30% for those smoking a pack a day or more. There was also an increase in respiratory symptoms with higher intensity of e-cigarette use, but the trend did not reach statistical significance .Table 3 gives results for the two longitudinal models for worsening respiratory symptoms. Symptoms worsened for 5% and 8%, respectively, for cutoff levels of ≥ 3 and ≥ 2. Symptom worsening was most common in the four categories featuring cigarette use, with risk ratios for worsening symptoms for the four categories ranging from 1.64 to 2.80, and always significantly higher than for never users, regardless of threshold. Also regardless of threshold, post hoc testing indicated that risk ratios for dual use of cigarettes+e-cigarettes were never different compared to exclusive cigarette use, whereas combustible plus noncombustible use was always associated with lower risk. Cigarette pack-years, second-hand smoke exposure, and marijuana use at W2 were also associated with symptom worsening at W3, at both cutoff levels. There were no statistically significant associations between exclusive use of cigars, smokeless tobacco or hookah and worsening of respiratory symptoms compared to never users. Post hoc testing indicated that risk ratios were significantly smaller than for exclusive use of cigarettes, regardless of cutoff level for the respiratory symptom outcome . In contrast, findings for exclusive e-cigarette use were sensitive to symptom severity, showing a significant association with worsening symptoms at a threshold of ≥2 , but not at a symptom threshold of ≥3 . Table 4 gives results for the two longitudinal models for improving respiratory symptoms . Symptoms improved for 21% and 29%, respectively, for cutoff levels of ≥2 and ≥3. In contrast to symptom worsening models, tobacco use was less apt to be associated with improvement and more sensitive to cutoff threshold. Categories of use featuring cigarettes were not reliably less likely to be associated with symptom improvement compared to never users; only exclusive use of cigarettes at a threshold of ≥2 was associated with lower risk ratio for symptom improvement . At this threshold, former smokers, e-cigarette, cigar and smokeless tobacco users were all significantly more likely to show symptom improvement compared to exclusive cigarette users . This was also true for e-cigarette users at a threshold of ≥3, where e-cigarette users were also more likely show symptom improvement compared to never users . This study underscores the adverse consequences of continued cigarette smoking among people without COPD or other non-asthma respiratory disease on functionally-important respiratory symptoms.