Send Orders of Reprints at Reprints@benthamscience.org Waterpipe Smoking and Dependence Are Associated with Chronic Obstructive Pulmonary Disease: a Case-control Study

Introduction: Waterpipe smoking gained popularity during recent years. Although waterpipe smoking exposes people to the same noxious substances found in cigarettes, popular belief considers it harmless. Our objective was to evaluate the association between waterpipe smoking and dependence, and COPD. Methods: We conducted a case-control study in two tertiary care hospitals. Cases were included if diagnosed as COPD by a pulmonologist and confirmed by post-bronchodilator FEV1/FVC<0.7; controls were included if free of any respiratory disease or symptom. After oral consent, a standardized questionnaire was administered and spirometry results were collected. Results: 211 COPD cases and 527 controls were studied. In previous smokers, any smoking type was associated with COPD. (p<0.001) for mixed smoking. Nevertheless, we found in waterpipe current smokers, an OR=8.9[3.9-20.7] (p<0.001) for the association between dependence evaluated by LWDS-11 scale, and COPD. These results were confirmed by stratified and multivariate analysis, after adjustment for cigarette smoking and confounding variables. A cumulative smoking of one waterpipe per week for 20years (or its equivalent) was predictive of higher risk of COPD. Discussion: Whereas evidence showing harmful effects of waterpipe smoking is sparse, this study showed a high OR between the risk of developing COPD and being an ex-smoker of waterpipe, or a current waterpipe dependent individual. Additional studies are necessary to confirm our results.


INTRODUCTION
Waterpipe (WP) smoking is quickly spreading in Middle Eastern countries and the world [1,2].Details can be found in several references [3][4][5][6].Because the produced smoke passes through water in the device, many smokers think that it is less harmful than cigarette smoking.Nevertheless, WP smokers are potentially exposed to the cigarettes substances that are harmful, and although studies concerning WP smoking effects are not extensive, they suggest that it leads to cancer, genetic damage, lung diseases and other conditions in active and passive smokers [7][8][9][10][11][12].It is now considered a global public health threat by the World Health Organization [3], and given the large number of WP smokers, it is essential to study the health effects of waterpipe smoking with renewed emphasis [13].In a crosssectional study, we had found prevalence ratios of 2.5 for *Address correspondence to this author at the Jdeidet El Meten, Chalet Suisse Street, Ramza Azzam bldg, 5th Floor, ZIP Code: 00961, Beirut, Lebanon; Tel: 009613385542; Fax: 009611696600; E-mails: psalameh@ul.edu.lb,pascalesalameh1@hotmail.com physician diagnosed respiratory diseases and chronic bronchitis symptoms in waterpipe smokers versus never smokers [14].
Since waterpipe smoking is a habit with physiological, psychological and social dimensions [15], it may manifest in several patterns.Around 80% of waterpipe smokers are not dependent on it [14] and may thus smoke irregularly.So, we would consider waterpipe smoking and waterpipe dependence as interesting exposure variables, to be linked with chronic obstructive pulmonary disease (COPD).
The primary objective of this study is thus to evaluate the association between WP smoking, WP dependence and COPD.

Study Design
It is a multicenter case-control study, performed in two tertiary care hospitals in Beirut, comparing a group of patients having COPD with a control group.WP exposure, WP dependence and socio-demographic characteristics were compared in both groups.Since the study was observational and there was no traceability of patients, the Internal Review Board (IRB) waived the need for an official approval to perform the study, provided the study respected patients' autonomy and confidentiality.

Study Population
The whole study population was composed of nonhospitalized individuals.The COPD group was composed of consecutive newly diagnosed outpatient cases of mild, moderate or severe COPD.Cases were included if they were 40 years of age, free of any other respiratory disease, diagnosed as COPD by a chest physician, and had a postbronchodilator FEV1/FVC <0.7.
Consecutive outpatients consulting for various extrapulmonary problems were included as controls if they were 40 years of age, free of COPD or any other respiratory disease (no asthma, no lung cancer, no lung fibrosis…), and not complaining of any chronic respiratory symptom (no chronic cough, no chronic sputum production, no wheezing, no dyspnea at rest or on exertion); no other exclusion criterion was applied.They were taken from all outpatient clinics situated close to the respiratory clinic: patients attending clinics of cardiology, endocrinology, dermatology, nephrology, hematology, urology, gastroenterology, gynecology, ophthalmology, pre-surgery consultation, and Ear, Nose and Throat were approached.Moreover, people accompanying cases (family members or friends) could also be taken as controls.For validation purposes, a random sample of 30 control individuals was tested with postbronchodilator spirometry, to ensure the absence of COPD.

Data Collection
COPD patients were evaluated for dyspnea, chronic cough and expectorations.They were classified by spirometry according to GOLD guidelines [16].After an informed consent, a standardized questionnaire evaluated age, gender, height, weight, education, work status, marital status, detailed cigarette and waterpipe smoking history (current, previous or never smokers), passive smoking, declared exposure to smokes and fumes at work and home, polluted areas residency history and consumption of fruits and vegetables.
Education was divided into five classes of illiterate, primary, complementary, secondary and academic education levels.Work status was categorized into currently working, retired, not finding a job, and never working, while marital status was categorized into married, single and widow/divorced.Cigarette smoking was defined as smoking more than 1 pack in a lifetime, while current waterpipe smoking was defined as a positive answer to the question "do you regularly smoke waterpipe" and previous waterpipe smoking as a positive answer to the question "were you a regular smoker of waterpipe".Frequency and duration of smoking were recorded for both cigarette and waterpipe.Cumulative smoking doses were subsequently calculated, as frequency multiplied by mean quantity: for cigarettes, it was the number of packs per day multiplied by the duration of smoking in years (packs*years); likewise, for waterpipe, it was the mean number of waterpipes smoked per week multiplied by the duration of smoking in years (waterpipes*years).
Passive smoking was addressed by questions about the number of smokers at home and at work, and the mean number of hours of exposure to their smoke.For exposure to smokes at home, questions about heating and cooking were asked as independent variables: the uses of diesel, wood, hot air, butane gas, electricity, central heating were addressed.Polluted areas were assessed using the following items: living close (<100m) to a busy and heavy traffic road, and living close (<100m) to a local power plant (which is a diesel exhaust source).Frequency of consumption of fruits and vegetables was also evaluated.
Cigarette and waterpipe dependence scores according to well known Fagerström Test for Nicotine Dependence (FTND) [17] and LWDS-11 [15], respectively.LWDS has already been developed and validated in the Lebanese population: it is composed of four subscales, the first representing nicotine dependence, the second negative reinforcement, the third psychological craving, and the fourth reflecting positive reinforcement.It has adequate internal consistency and test-retest reproducibility.Results were biologically and psychologically sound.It is a useful clinical and epidemiological tool.LWDS-11 score 10 defines WP dependence [15].

Sample Calculation
Sample size calculation was performed with an alpha risk of 5% and a beta risk of 20%.The mean percentage of Lebanese who smoke and could theoretically be exposed to WP is around 50% [18] and 20% of these are dependent to waterpipe; [14] thus, 10% of the general population is considered WP dependent.To show a twofold increase in risk of COPD due to WP dependence, we needed a sample of 721 subjects: 515 controls for 206 cases.

Statistical Analysis
Independent laypersons entered the data.SPSS software, version 13.0 was used.Quality control of data entry and data cleaning were carried out by researchers.A p-value 0.05 was considered significant.Missing data were not replaced since they were fewer than 10% of values.For continuous variables used in comparison of groups, we used ANOVA or students' tests for variables with adequate normal distribution.For non-normally distributed continuous variables and for non-continuous quantitative and ordinal variables, the Kruskal-Wallis test was used.For categorical variables, the Chi2 and Fisher exact tests were used when applicable.
The major dependent variable was: being diagnosed with COPD or not, based on post bronchodilator FEV1/FVC values.Independent variables were waterpipe smoking and dependence score.Cumulative WP exposure was expressed in WP-years (Number of WP smoked per week x the duration of WP smoking in years = number of WP-years).Dose-effect relationships were evaluated with trend tests and served to determine a cut-off point of cumulative waterpipe exposure.A stratified analysis was carried out over cigarette smoking, and multivariate results were presented for cigarette and non cigarette smokers.
Multivariate analysis compared measures between groups of comparison, taking into account potential confounding variables of a p-value <0.20 in bivariate analysis: ever cigarette smoking, gender, residency, age, height, weight, body mass index, education, work status, marital status and number of smokers to which a person was potentially exposed (passive smoking at home and work), in addition to factors that affect COPD (even living close to a busy road, living close to a diesel exhaust source, heating home or cooking on diesel or solid fuels, and eating low quantities of fruits and vegetables).Logistic regressions were chosen because of the dichotomous dependent variable (COPD), while three major independent variables were evaluated: current waterpipe smoking, waterpipe dependence and previous waterpipe smoking (in separate models).Moreover, a first order interaction term was introduced, respectively (waterpipe smoking*ever smoking cigarette), (waterpipe dependence*ever smoking cigarette) OR (previous waterpipe smoking*ever smoking cigarette).Stepwise descendant likelihood ratio logistic regressions were applied on the global population, on cigarette smokers and non cigarette smokers.The final models were retained.Point estimates, 95% confidence intervals and p-values were presented.

Description of the Population
Five hundred and seventy persons with chronic respiratory symptoms were approached in total; among them, 549 agreed to participate (96.3%).Non participants gave the following reasons for their non participation: no time (70%) and no particular reason (30%).Post bronchodilation spirometry was performed for all 549 patients (100%).After bronchodilation, 211 patients (38.4%) fulfilled the definition of COPD (post-bronchodilator FEV1/FVC <0.70), while the rest, i.e. 338 (61.6%) persons with respiratory symptoms, did not fulfill this definition, and thus could not be classified; they were subsequently removed from the analysis.Thus, for the 211 COPD cases, the mean postbronchodilator FEV1/FVC for COPD patients was 0.60 (SD=0.08):they were classified as follows: there were 38(17.9%)stage I, 121(57.6%)stage II, 45(21.2%)stage III, and 7(3.3%) stage IV COPD severity class.

Smoking History Details
Compared to never smoking, all types of smoking were associated with a higher risk of COPD (Table 2).Among exsmokers, mixed smokers (cigarettes & waterpipe) had the highest association with COPD (OR=44.1),followed by cigarette smokers (OR=28.99),and waterpipe smokers (OR=11.67).
Among current smokers, cigarette smokers have the highest association (OR=20.5),followed by mixed smokers (OR=9.37)and waterpipe smokers (OR=1.78).However, when looking at dependence, mixed waterpipe and cigarette dependence showed the highest association with COPD (OR=42.88),followed by cigarette dependence (OR=42.32)and waterpipe dependence (OR=3.14).Moreover, being exposed to passive smoking at home increases the risk of COPD (OR=2.51;p<0.001).
When comparing waterpipe dependent current smokers with never smokers and waterpipe non dependent smokers, an OR of 8.93 [3.85-20.68]was found (p<0.001); in this case, when stratifying for cigarette smoking, no qualitative interaction was found (tests of homogeneity of the Odds Ratio between strata (p-value=0.083); in non cigarette smokers stratum, OR=1.91 and in cigarette smokers stratum, OR=16.25).A quantitative interaction was suspected since ORs between strata are highly heterogeneous.The other way round, cases had a mean LWDS-11=18.44, while controls had 11.59 (p<0.001).
Moreover, when comparing previous waterpipe smokers versus never smokers, an OR=22.13 [9.82-49.89]was found (p<0.001).Similar results were found for stratified analysis, and a quantitative interaction was also suspected.Fig. (1) presents the relationship of COPD with waterpipe smoking and dependence.

Dose-Effect Relationships
Significant dose-effect relationships were found for all kinds of smoking: higher cumulative doses of previous cigarette, current cigarette, previous waterpipe frequency and duration, and current waterpipe frequency and duration are associated with increased risk of COPD (p<0.001 for all).Current low doses of waterpipe have no association with COPD; high cumulative doses have a strong to unlimited one (Table 3).A cut-off point of 20 waterpipe-years for current and previous smoking was chosen for further analysis.Fig. (2) shows waterpipe dependence dose-effect relationship, with never smokers as the reference category.The higher the dependence score on LWDS-11, the higher the risk of COPD: for non dependent waterpipe smokers (LWDS-11=1-9), OR=0; for moderate dependent (LWDS-11=10-16), OR=5.56, and for highly dependent (LWDS-11>16), OR=12.24(p<0.001 for trend).

Multivariate Analysis
A step-wise descendent logistic regression was run, taking smoking of 20 waterpipe-years or more as a major independent variable (n=734; 210 COPD cases and 524 controls), the retained model adequately fitted the data: R 2 of Nagelkerke=0.613and Hosmer Lemeshow test p-value was 0.438; moreover, 85.8% of individuals were adequately classified by the model.Again, we found no interaction between waterpipe and cigarette smoking (the interaction term was removed from the model).There was however a significant positive association between ever waterpipe smoking and COPD (ORa=2.43[1.18-4.97];p=0.015).Older  4).

Subgroups Multivariate Analyses
In Table 4, we also present multivariate analyses for the association between COPD and waterpipe smoking in subgroups.Cigarette smokers subgroup: when taking smoking 20 waterpipe-years or more as a major independent variable (n=406; 196 COPD cases and 210 controls), the retained model adequately fitted the data: R 2 of Nagelkerke=0.476and Hosmer Lemeshow test p-value=0.124;moreover, 79.3% of individuals were adequately classified by the model.There was no significant relationship between waterpipe smoking and COPD (ORa=1.78[0.79-4.03];p=0.164).Never cigarette smokers subgroup: when taking smoking of 20 waterpipe-years or more as a major independent variable (n=328; 14 COPD cases and 314 controls), the retained model adequately fitted the data: R 2   When taking waterpipe dependence as the major independent variable (n=355; 26 COPD cases and 329 controls), the retained model generated by the descendant step-wise logistic regression adequately fitted the data: R 2 of Nagelkerke=0.625 and Hosmer Lemeshow test p-value=0.863.Moreover, 95.5% of individuals were adequately classified by the final model.There was a significant positive and strong association between current waterpipe dependence and COPD (ORa=5.88[1.11-31.05];p=0.037).

DISCUSSION
This is the first study that thoroughly looks at the relationship between waterpipe smoking and COPD.In previous smokers, we found a high OR between the risk of developing COPD and being an ex-smoker of waterpipe.In current smokers waterpipe smoking seems harmless at first sight (OR=1.78[0.54-5.89]): the reverse causality concept may be applied here for new waterpipe smoker, this habit being a relatively new trend in Lebanon.However, the risk of COPD was found high with waterpipe dependence and with cumulative smoking 20 waterpipe-years versus non smokers.These results were confirmed by dose-effect relationships (in terms of frequency, duration and cumulative exposure) and by multivariate analyses.
In previous cigarette, WP and mixed smokers, we found higher ORs with the risk of developing COPD than in current  smokers.This could reflect a long lasting effect of smoking even after cessation.Yet, it could also be due to the fact that COPD cases stopped smoking more than controls [19].It is noteworthy that in previous waterpipe smokers, the risk of having COPD is high (OR=11.67)although still inferior to that of previous cigarette smokers (OR=28.99).It is also remarkable that this risk increases in previous mixed waterpipe and cigarette smokers (OR=44.10),which suggest a synergistic interaction between cigarette and previous waterpipe smoking.
However, we could not find a statistically significant interaction in stratified or multivariate analysis.Although a synergistic quantitative interaction between cigarette and waterpipe smoking is possible, the sample size may have precluded confirmation of such issue in the cigarette subgroup in both analyses.Mixed smokers could be heavier smokers of cigarettes with higher risk of COPD than exclusive cigarette smokers.They could be using waterpipe only to titrate their nicotine serum levels, thus not enough to demonstrate a toxic effect; the mean cumulative dose of  cigarettes is higher in previous mixed smokers versus exclusive cigarette smokers.This issue remains to be studied in larger scale studies, specifically designed to study the interaction between waterpipe and cigarette smoking.
Nevertheless, the multivariate analysis results that show a positive and strong association between waterpipe smoking and COPD in never cigarette smokers, and between cumulative smoking > 20 waterpipe-years and COPD, are in favor of waterpipe smoking toxicity leading to COPD.Eissenberg and Shihadeh [20] found that expired-air carbon monoxide, blood carboxyhemoglobin and smoke exposure were greater with waterpipe than with cigarette smoking.A recent meta-analysis concluded that waterpipe smoking is likely to be a cause of COPD [21].
Moreover and interestingly, our study is the first that shows a positive relation between waterpipe dependence and COPD, more potent than the relation between waterpipe smoking and COPD.Although comparable results were never demonstrated in the literature, our results are analogous to those found by other researchers for cigarette smoking, where tobacco dependence increases the risk of COPD [22,23].
Situations such as ever living close to a busy road and lower education appear in our study, in certain circumstances, as risk factors for COPD.The findings of Swiss cohort studies provide strong evidence that living near busy streets leads to adverse respiratory health effects without specifying COPD itself [24].Schikowski et al. showed [25] that living near a major road might increase the risk of developing COPD and can have a detrimental effect on lung function especially in women.As for education, similar to our results, Krzyzanowski et al. [26] found a faster rate of FEV1 decline in less educated than in more educated men; the odds ratio of obstructive lung disease in primary educated versus university educated subjects was 2.9 and the corresponding odds ratio for spirometric airflow limitation was 5.2 in a Norwegian study after adjustment for occupational and smoking exposure; [27] Tabak et al. [28] investigated the longitudinal association between baseline educational level and the rate of FEV1 decline and found that less educated women have a faster decline of FEV1 than the more educated one.Moreover, as found in other countries [29][30][31][32][33][34], we noted in our study that indoor pollution (heating home not electrically) and passive smoking are associated with a higher risk of COPD.It is therefore probable, that these factors (living close to a busy road, lower education, heating home by fuel, passive smoking) may be considered as interacting with smoking to induce a synergistic effect on development of this disease.
We noticed several drawbacks in our study.The low overall rate of waterpipe smoking in this relatively older population led to a decrease in the power of the study in several types of analyses, particularly multivariate ones.Moreover, the analysis of water-pipe smoking related to the degree of dependence and possible relationship to COPD is based on a relatively small number of waterpipe smokers in the case-and control groups; this may widen confidence intervals and decrease the robustness of our results.
A selection bias might exist since controls and cases were chosen within the same hospitals; differences in hospitals and physicians' reputation may affect this selection.Unhealthy controls (39% of the control group) had mixed non-respiratory diagnosis and are probably non-comparable to the general population; their presence in the control group may lead to the underestimation of the associations we found.In addition, cases and controls may have geographical comparability or belonged sometimes to the same family: all this could induce an underestimation of the expected associations.On the other hand, it was not possible to get detailed information about non responders; however, answer refusals due to "lack of time" or "no reason" are expected to be both because of non-motivation or illiteracy, but not because of exposure status: all other factors being equivalent, exposed and unexposed people would have the same probability of responding.Persons with lower education level have generally lower probability of responding, and they also have higher odds of waterpipe and cigarette smoking.This may introduce a selection bias of more educated persons, but we do not expect this bias to be differential.Pulmonologists recruiting subjects were also unaware of their exposure status, especially when considering different types of  No COPD COPD exposure (waterpipe, cigarette and mixed smoking).This may decrease selection bias, which however cannot be totally excluded.Nevertheless, we would expect this bias be of low magnitude since we obtained a high response rate (>90%).As for the generalizability of the results, we expect it to be for all individuals who would theoretically be treated in the tertiary centers: their geographical origin is broad (from all Lebanese regions), but their socioeconomic status may be higher than that of the general population.More pronounced results would be expected lower socioeconomic status individuals.
A classification bias is also possible since exposure was only measured by the questionnaire, without further evaluation; as in all case-control studies, this may be subject to recall and subjectivity bias.On the other hand, to minimize confounding, multiple exposures that would be suspected to cause COPD were taken into account; however, residual confounding is still possible due to unmeasured variables, particularly since baseline variables differed between comparison groups.Thus, further studies are necessary to confirm the causality relationship of the association we found between waterpipe smoking and COPD, and to extend it to lower socioeconomic status individuals.
Current waterpipe smokers subgroup: When taking current smoking of more than 20 waterpipe-years as the major independent variable (n=358; 27 COPD cases and 331 controls), the retained model generated by the descendant stepwise logistic regression adequately fitted the data: R 2 of Nagelkerke=0.656and Hosmer Lemeshow test p-value=0.893.Moreover, 96.4% of individuals were adequately classified by the final model.There was however a significant positive and unlimited association between current waterpipe smoking and COPD (ORa=13.24[2.21-79.20];p=0.005).

Table 1 . Baseline Characteristics of Cases and Controls
*Numbers may sum inferior to the total (100%) due to missing values.