|Home | About | Journals | Submit | Contact Us | Français|
Marijuana smoke is very similar to tobacco smoke, but whether it has similarly adverse effects on pulmonary function is unclear.
To analyze associations between marijuana (both current and lifetime exposure) and pulmonary function
We used repeated measurements of pulmonary function and smoking collected over 20 years in the Coronary Artery Risk Development in Young Adults (CARDIA) Study. Mixed linear modeling was used to account for individual age-based trajectories of pulmonary function and other covariates including tobacco use, which was analyzed in parallel as a positive control.
4 US cities, 1985–2006
Black and white men and women recruited at age 18–30 years and followed for 20 years
Forced expiratory volume in 1 second (FEV1) and forced vital capacity (FVC)
Marijuana exposure was nearly as common as tobacco exposure, but was mostly light (median 2–3 episodes per month). Tobacco exposure, both current and lifetime, was linearly associated with lower FEV1 and FVC. In contrast, the association between marijuana exposure and pulmonary function was non-linear (p<.001): at low levels of exposure, FEV1 increased by 13 ml/joint-year (95% confidence interval (CI): 6.4 – 20) and FVC by 20 ml/joint-year (95%CI:12 – 27); but at higher levels of exposure, these associations leveled off or even reversed. The slope for FEV1 was −2.2 ml/joint-year (95%CI:−4.6 – 0.3) at >10 joint-years, and −3.2 ml per marijuana smoking episode/month (95%CI:−5.8 – −0.6) at >20 episodes/month. The net association with FEV1 declined to or below baseline with very heavy use, but FVC remained significantly elevated in even heavy users (e.g., 76 ml [95%CI:34 – 117) at 20 joint-years).
Occasional and low cumulative marijuana use was not associated with adverse effects on pulmonary function.
Exposure to tobacco smoke causes lung damage with clinical consequences that include respiratory symptoms, chronic obstructive pulmonary disease (COPD), and lung cancer1,2. COPD and lung cancer are leading causes of death2,3, and smoking tobacco cigarettes is the most important preventable cause of death in the United States4,5.
Marijuana smoke contains many of the same constituents as tobacco smoke6, but it is unclear whether smoking marijuana causes pulmonary damage similar to that caused by tobacco. Prior studies of marijuana smokers have demonstrated consistent evidence of airway mucosal injury and inflammation7–9, and increased respiratory symptoms such as cough, phlegm production and wheeze similar to tobacco smokers10–12. However, analyses of pulmonary function and lung disease have failed to detect clear adverse effects of marijuana use on pulmonary function10–13. It is possible that cumulative damage to the lungs from years of marijuana use could be masked by short-term effects; prior analyses have not attempted to disentangle these factors. Smoking marijuana is increasingly common in the United States14, and understanding whether it causes lasting damage to lung function has important implications for public health messaging and medical use of marijuana15,16.
The Coronary Artery Risk Development in Young Adults (CARDIA) Study collected repeated measures of tobacco and marijuana smoking as well as pulmonary function over 20 years in more than 5000 study participants. We estimated both current intensity and lifetime cumulative exposure to tobacco and marijuana smoking, and analyzed their associations with spirometric measures of pulmonary function over the 20 years of follow-up.
CARDIA is a longitudinal study designed to measure risk factors for coronary artery disease in a cohort of African-American and European-American women and men (n=5115) aged 18–30 years and healthy at enrollment in 198517,18. Participants were sampled from 4 US communities without selection for smoking behaviors, and comprise a broad cross-section of typical tobacco and marijuana use patterns. With the informed consent of participants and the approval of Institutional Review Boards at each study center (Oakland, Chicago, Minneapolis, and Birmingham), participants underwent a baseline examination and 6 follow-up examinations, with 69% retention at Year 20. Pulmonary function testing was performed at years 0, 2, 5, 10 and 20. For this investigation, we included all visits for which pulmonary function, smoking behavior, second hand smoke exposure, height and waist circumference were available.
Current intensity of tobacco use (cigarettes smoked per day) was assessed at each examination. These data, along with baseline examination data on past years of smoking, were used to estimate cumulative lifetime exposure to cigarettes in terms of “pack-years”, with 1 pack-year of exposure equivalent to 7300 cigarettes (1 year * 365 days/year * 1 pack/day * 20 cigarettes/pack). Misclassification of smoking exposure by self report, measured by comparisons with serum cotinine levels, is uncommon19.
Current intensity of marijuana use (episodes in the last 30 days) was also assessed at each examination. Using baseline examination data on past lifetime exposure to marijuana, current marijuana use intensity, and another question designed to assess number of joints or pipe fulls smoked per episode (see detailed assumptions in Online Appendix), we calculated total lifetime exposure to marijuana joints in “joint-years”, where 1 joint-year of exposure is equivalent to 365 joints or pipe fulls smoked (1 year * 365 days/year * 1 joint/day), as has been described previously20.
Study outcomes were forced vital capacity (FVC) and forced expiratory volume in 1 second (FEV1) measured by forced spirometry. These were collected using a Collins Survey 8-liter water-sealed spirometer and an Eagle II Microprocessor (Years 0, 2, 5 and 10) and then an OMI rolling seal spirometer (Year 20); a comparability study performed on 25 participants demonstrated an average difference of <1% for both measurements. Standard quality control and testing procedures were maintained according to established guidelines21,22.
CARDIA was designed to recruit approximately equal numbers of self-identified “Black, not Hispanic” and “White, not Hispanic” men and women in order to assure an adequate sample of the largest minority group in the US at that time. Height and waist circumference were measured at each examination. As a proxy for socio-economic status, we used the maximum educational grade attained for each participant. Second hand smoke exposure in hours/week (sum of exposure in the home, small enclosed spaces and large spaces) was assessed at each exam, with linear interpolation for missing data. Asthma was self-reported at each exam; we used the baseline assessment. We obtained average annual city-specific levels of airborne particulate matter <10 microns (PM10) and <2.5 microns (PM2.5) in size23 around the four CARDIA study centers from the Environmental Protection Agency24 (see Online Appendix for details).
Participants were categorized by whether they ever reported current use of tobacco, marijuana or both at a CARDIA examination, and compared across these categories using descriptive statistics. We then categorized participants according to degree of current and lifetime tobacco and marijuana exposure at each examination, and described pulmonary function (FEV1 and FVC) across categories before and after adjustment. Tests of trend and interaction were performed in fully adjusted models (see below).
The categorized exposure models described above represent a standard approach to multivariable-adjusted association testing. Categorization models, however, use necessarily arbitrary category thresholds and do not take full advantage of the continuous exposure measurements for estimation or adjustment purposes. To fully explore and test potential nonlinear associations, we modeled tobacco and marijuana exposure variables as flexible cubic splines (see Online Appendix for details) in adjusted models to allow associations with pulmonary function to take different shapes at lower versus higher levels of exposure25.
For each adjusted analysis described above, we used mixed models accounting for repeated measures of pulmonary function within subjects, with a random intercept and a random 3-knot age spline within each individual, and an unstructured variance-covariance structure. Fully adjusted models included fixed effects for year, center and center-year (their interaction), race-sex category, education, asthma; cubic splines for age, height, waist circumference, second hand smoke exposure, PM10 and PM2.5; and interactions between the age spline variables and race-sex, asthma, waist spline variables, and height spline variables to allow for differing flexible age-based trajectories of pulmonary function for participants with differing characteristics. Models were queried to produce adjusted estimates of slope (reflecting the incremental difference in pulmonary function observed with additional tobacco or marijuana smoking) and net association (reflecting the net observed difference between persons with a particular level of consumption and persons with none) at various points along the association curve. All analyses were executed with Stata 11 (College Station, Texas) and used 2-sided tests for significance at the 0.05 level (and 95% confidence intervals [CI]).
The 5115 CARDIA participants recruited in 1985–6 contributed 20,777 total visits that included pulmonary function testing. Of these, 959 visits were excluded for lack of complete information on smoking behavior, 114 for lack of height or waist measurements, and 1 for an unknown visit date, leaving 19703 visits (95%) with complete data from 5016 participants (98%). Participants contributed 3.9 visits/participant on average; attrition was more common in tobacco smokers but not associated with marijuana use. FEV1 and FVC varied across participants, rose slightly with age through the late 20’s, and declined slowly thereafter (Figure 1).
More than half of participants (54%, mean age 25 at baseline) reported current marijuana or tobacco smoking or both at one or more examinations (Table 1). Smoking patterns differed by race and sex, with black women most likely to smoke tobacco only, white men most likely to smoke marijuana only, and black men most likely to smoke both. Tobacco smokers tended to have lower education and income and to be slightly shorter and less active, whereas marijuana smokers tended to be taller and more active. The median intensity of tobacco use in tobacco smokers was substantially higher (8–9 cigarettes per day) than the median intensity of marijuana use in marijuana smokers (2–3 episodes in the last 30 days). While marijuana and tobacco exposures were strongly correlated, our sample included 91 participants with no tobacco exposure and more than 10 joint-years of marijuana exposure (contributing 153 observations of pulmonary function), 40 (56 observations) of whom had more than 20 joint-years of exposure.
In fully adjusted models that considered 4-level categorizations of current and lifetime exposure to tobacco and marijuana, tobacco smoking (both current and lifetime) was associated with a lower FEV1, and current smoking with a lower FVC (Table 2). For example, compared with zero exposure, FEV1 was 63 ml lower (95%CI: −89 – −36) and FVC was 69 ml lower (95%CI: −97 – −41) with current tobacco exposure of >20 cigarettes per day, and 101 ml lower (95%CI: −136 – −65) with lifetime tobacco exposure >20 pack-years. In contrast, exposure to marijuana (both current and lifetime) was associated with a higher FVC, and lifetime exposure with higher FEV1. For example, compared with zero exposure, FVC was 59 ml higher (95%CI: 12 – 107) and FEV1 36 ml higher (95%CI: −6.5 – 79) with lifetime marijuana exposure >20 joint-years, and FVC was 20 ml higher (95%CI: −5.2 – 49) with >20 episodes of marijuana use/month. We found no statistically significant interactions between tobacco and marijuana exposure for either FEV1 or FVC.
When we modeled current and lifetime tobacco and marijuana exposure as continuous exposures and permitted flexible non-linear associations (via splines), we again found strong, dose-related associations (p<.001) between increasing exposure to tobacco and lower FEV1 and FVC (Figure 2), with no evidence of non-linearity (Table 3). Declining slopes ranged as steep as −2.8 ml (95%CI: −4.8 – −0.7) per additional cigarette smoked per day and −7.0 ml (95%CI: −10 – −3.7) per additional pack-year for FEV1, and were of similar magnitude for FVC (Table 3). At 50 pack-years of exposure, FEV1 was on average 332 ml lower (95%CI: −401 – −263) and FVC was 229 ml lower (95%CI: −310 – −147) compared with no exposure.
For marijuana, we found strong statistical evidence that associations between marijuana use and pulmonary function were non-linear (Figure 2, Table 3). At low lifetime exposure levels, increasing marijuana use was associated with a steep increase in both FEV1 (13 ml/joint-year higher [95%CI: 6.4–20]) and FVC (20 ml/joint-year FVC [95%CI: 12 – 27]); but at higher levels of exposure (>7 joint-years), the slope leveled out or even turned downwards. At >10 joint-years of lifetime exposure, we found a declining slope for FEV1 of borderline significance (−2.2 ml/joint-year [95%CI: −4.9 – 0.36], p=0.079), and a significant decline in FEV1 at more than 20 episodes of marijuana use per month (−3.2 ml/episode [95%CI: −5.8 – −0.6], p=.017). While net associations with FEV1 became negative at very high exposure levels (>40 joint-years/>25 episodes/month), these negative deflections were not statistically significant (Table 3). FVC remained significantly elevated in even heavy users (e.g., 76 ml [95%CI:34 – 117) at 20 joint-years).
In this 20-year study of marijuana and pulmonary function, we confirmed the expected reductions in FEV1 and FVC from tobacco use. In contrast, marijuana use was associated with higher FEV1 and FVC at the low levels of exposure typical for most marijuana users. With up to 7 joint-years of lifetime exposure (e.g., 1 joint/day for 7 years or 1 joint/week for 49 years), we found no evidence that increasing exposure to marijuana adversely affects pulmonary function. This association, however, was non-linear: at higher exposure levels, we found a leveling off or even a reversal in this association, especially for FEV1. While our sample contained insufficient numbers of heavy users to confirm a detrimental effect of very heavy marijuana use on pulmonary function, our findings suggest this possibility.
The associations we found between tobacco and pulmonary function are consistent with a large body of prior research on the adverse pulmonary consequences of tobacco smoking. The high prevalence of tobacco smoking, the wide range of exposure intensity among smokers, and tobacco’s legality have made it an easy target for observational epidemiology. Exposure predicts reduced expiratory flow and air trapping, gas exchange abnormalities and emphysema1, and smoking cessation interventions reduce the rate of FEV1 decline in smokers26 (i.e., these associations are likely causal). Our findings of a linear dose-response relationship showing lower FEV1 and FVC with increasing tobacco exposure, consistent with prior findings, represent a positive control for our study of the association between marijuana smoking and pulmonary function.
Prior studies of marijuana and pulmonary function have yielded apparently conflicting results10–13. Many studies have focused on the ratio of FEV1/FVC, lower values of which suggest the presence of airway obstruction, and have found either no association10,20,27 or lower FEV1/FVC with marijuana use28–32. Lower FEV1/FVC in marijuana smokers, however, can be explained at least partly by a tendency towards higher FVC or total lung capacity28,29,32. A recent longitudinal study, which demonstrated significantly higher FVC and total lung capacity with marijuana exposure, strongly supports this notion13,20, as does our study.
Marijuana’s potential association with FEV1 has been even less clear. Tobacco smoking reduces FEV1, but despite the similarities in the constituents of marijuana smoke and tobacco smoke and a priori expectations that marijuana smoking might have similar effects, prior research has not demonstrated this. In studies that report FEV1 in association with marijuana use, findings have mostly been null20,28,32–35, though one study reported the apparently paradoxical finding of a lower FEV1 with past marijuana use, but a non-significantly higher FEV1 with current use29.
Our study suggests a way to reconcile these findings. Because of the many thousands of measurements obtained over 20 years among over 5000 participants with a wide range of smoking habits, we could simultaneously account for levels of current and past lifetime use of both marijuana and tobacco, and test for non-linearity in their associations with pulmonary function in order to disentangle short-term and long-term effects. We found highly significant non-linearity, with a positive association for both FEV1 and FVC at low levels of exposure that reversed in direction towards a possibly negative association for FEV1 at higher levels of exposure (see Figure 2 and slopes in Table 3). These findings could explain the past/current paradox previously noted29, and are also consistent with the average null association reported in studies20,28,32–35 that either dichotomized marijuana exposure (user/non-user)28–31,33,36 or constrained the association to be linear across all levels of exposure10,20,32,35. When we looked at “marijuana only” smokers (Table 2, first panel), we also found a null association with FEV1 and FVC. Only after parsing out the association at different levels of exposure, with careful control for confounding, did the suggestion of a negative association for FEV1 at high levels of exposure emerge.
These findings suggest that marijuana smoking could influence pulmonary function via multiple mechanisms. To explain the higher FVC previously observed in marijuana smokers20,32, some have proposed that the deep inspiratory maneuvers practiced by marijuana smokers could stretch the lungs13,20, resulting in larger lung volumes20,32. Another speculative possibility is strengthening of chest wall musculature or another “training” effect that allows marijuana users to inspire more fully (closer to total lung capacity) on spirometry testing. A non-destructive stretch or training effect is consistent with previously reported findings in marijuana smokers of lower lung density32 and a lack of emphysematous change32 or diminished diffusion capacity20,27,32,36. This mechanism would explain our FVC results, and could explain the positive deflection of FEV1. The functional impact of this association on lung health or respiratory function in daily life is unclear13. An alternate explanation is the acute bronchodilatory effect of marijuana use that has been directly observed in some studies11. This effect, however, is transient (lasting approximately 60 minutes11), and seems unlikely to explain higher lung volumes measured during the CARDIA examination unless many marijuana users smoked immediately before the examination.
The suggestion of a negative association with FEV1 at higher exposure levels could reflect mixing of this putative stretch/training effect with a second mechanism operating on a different time-exposure scale. A negative association with heavy marijuana smoke exposure aligns with our a priori hypothesis that marijuana smoking should produce damage to the airways and accelerated loss of lung function similar to that caused by tobacco smoking. Hypothetically speaking, a positive effect from marijuana in the short term (the stretch/training effect) and a negative effect in the long term (damage from smoke exposure), should result in a non-linear association such as the one we observed. According to this explanation, the predominant effect for FEV1 at very high exposure (over 40 joint-years) reflects cumulative damage; the predominant effect for FVC at all levels of exposure is from the stretch/training mechanism.
Our study has limitations. While CARDIA offers longitudinal spirometry measurements, it lacked body plethysmographic measurements of static lung volumes (total lung capacity and residual volume) and measures of diffusing capacity and radiographic emphysema. A minority of our participants reported very high levels of marijuana exposure (and a smaller minority of these were non-smokers of tobacco), so our estimates at high marijuana exposure levels are imprecise. The self-reported measures of marijuana and tobacco smoking are certain to include recall error, both random and systematic, and do not include any indication of smoking method (joint, pipe, “bong”, etc). It is unlikely, however, that such error would differentially occur in association with pulmonary function, and non-differential error would most likely bias results towards of the null. Our mixed modeling approach is ideal for filtering out random error and taking advantage of individual-level correlations in the data. As with any observational analysis, unmeasured or inadequately modeled confounding effects could be mixed with our estimates, but our study’s extensive covariate measurements and large sample permitted more extensive efforts to control confounding than were possible in previous studies.
Marijuana may have beneficial effects on pain control, appetite, mood, and management of other chronic symptoms15,16. Our findings suggest that occasional use of marijuana for these or other purposes would not have significant adverse consequences on pulmonary function. It is more difficult to estimate the potential effects of regular heavy use, as this pattern of use is relatively rare in our study sample; however, our findings do suggest an accelerated decline in pulmonary function with heavy use, and a resulting need for caution and moderation when marijuana use is considered.
This work was supported by the National Institute on Drug Abuse (R01-DA-025067) and the National Heart, Lung and Blood Institute (N01-HC-95095 and N01-HC-48047). NHLBI helped design CARDIA and funds data collection, supports a Publications and Presentations Committee that reviews and approves all publications, and provides a representative that sits on that committee, but does not otherwise control publication. NIDA funded this analysis but did not participate in CARDIA or review this manuscript.
Dr. Sidney helped lead data collection for CARDIA, Dr. Kertesz obtained funding for this analysis, Dr. Vittinghoff developed and helped implement the statistical methodology, and Dr. Pletcher conceived the analysis with Dr. Kertesz, analyzed the data and drafted the manuscript. All authors participated in interpretation of the data and critical review of the manuscript and approved the final draft. Dr. Pletcher had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Dr. Kalhan has served as a consultant for Boehringer-Ingelheim, Forest Laboratories, and AstraZeneca; has received honoraria for lectures from GlaxoSmithKline and AstraZeneca; has received honoraria for development of educational materials from Quantia Communications and Medscape Education; and has received industry-sponsored grants from GlaxoSmithKline and Boehringer-Ingelheim. Dr. Kertesz chaired a committee that advised the Drug Treatment Task Force for the Chief Justice of the State of Alabama. Dr. Kertesz is currently an employee of the Department of Veterans Affairs; views expressed in this manuscript do not reflect positions of either the Department of Veterans Affairs or any other entity of the federal government.
Besides the federal funding listed above, the authors have no other conflicts of interest to report.