Customize your JAMA Network experience by selecting one or more topics from the list below.
Boeckler P, Cosnes A, Hedelin G, Lipsker D. Association of Cigarette Smoking but Not Alcohol Consumption With Cutaneous Lupus Erythematosus. Arch Dermatol. 2009;145(9):1012–1016. doi:10.1001/archdermatol.2009.199
To ascertain whether smoking or alcohol consumption is associated with lupus erythematosus (LE), because this topic is still subject to debate and part of the debate could be related to the fact that smoking and alcohol consumption are specific risk factors for cutaneous LE.
Prospective multicenter case-control study.
Three French university hospitals.
One hundred eight patients with LE and 216 control subjects.
Standardized questionnaire evaluating cigarette smoking and alcohol consumption.
Main Outcome Measures
The statistical significance of smoking history and alcohol consumption as associated risk factors for LE by estimating matched case-control odds ratios and their 95% confidence intervals, using multiple conditional logistic regression and the Breslow-Day test to investigate differences in quantities of cigarette and alcohol consumption.
Of the LE patients, 73.1% smoked compared with 49.5% of controls, (odds ratio, 2.77; 95% confidence interval, 1.63-4.76). There was no significant difference in alcohol consumption between LE patients and controls. Among the 79 LE patients who smoked, 72 (91.1%) had started smoking before the first manifestation of LE (mean delay between initiation of smoking and first signs of LE, 14.1 years). The LE patients smoked significantly more than controls did (11.7 vs 7.0 pack-years; P = .002). The prevalence of smoking among patients who met more than 4 American College of Rheumatology (ACR) criteria and/or with antinuclear DNA antibodies was lower than the prevalence in patients who met fewer than 4 ACR criteria or than the prevalence in controls (P < .001).
Cigarette smoking is associated with LE, but alcohol consumption is not. The risk conferred by cigarette smoking seems highest in patients who meet fewer than 4 ACR criteria and/or who do not have antinuclear DNA antibodies.
The etiology of lupus erythematosus (LE) has not yet been elucidated in detail, although genetic and environmental factors play a role in its development. Commonly accepted environmental etiological factors include exposure to UV light and drugs.1,2 Some epidemiological studies have suggested that smoking tobacco is a risk factor for LE, probably in genetically predisposed individuals.3-7 However, whether cigarette smoking increases the risk of developing systemic LE (SLE) remains controversial. To date, 3 case-control studies3-5 have reported significantly increased odds ratios (ORs) for the development of SLE in smokers, whereas 6 others have not found a clear association.8-13 Thus, the precise effect of cigarette smoking and/or alcohol consumption on LE is not clearly established. Most studies focused on SLE, and the patients included were almost exclusively women. We and other authors6,7,14 have previously shown that cigarette smoking could be specifically involved in the pathogenesis of cutaneous LE. Thus, the effect of cigarette smoking could specifically involve a subgroup of patients with cutaneous LE, and analysis of this effect should not be restricted to patients who meet 4 or more American College of Rheumatology (ACR) criteria. On the other hand, alcohol consumption seems to be associated with a decreased risk of LE.4,5 We conducted a multicenter prospective study to evaluate cigarette smoking and alcohol consumption as risk factors for developing cutaneous LE in a large series of patients compared with control subjects. In this report of our results, we also discuss the physiopathological role of cigarette smoking in LE.
From 2004 to 2006, cases and controls were recruited from outpatients of 3 French university hospitals that are strongly involved in the care of patients with connective tissue diseases: the Dermatology Clinic of the University of Strasbourg (center 1), the Dermatology Service of the University of Créteil (center 2), and the Internal Medicine Service of the University of Pitié/Tenon (center 3). All consecutive patients presenting with LE who underwent evaluation at the centers were included in the study.
The diagnosis of discoid LE (DLE), disseminated DLE, tumid LE, and acute LE was made according to established clinical and histopathological criteria.15 The diagnosis of subacute LE was established according to the definition of Sontheimer et al.1 The patients who fulfilled 4 or more ACR criteria were defined as having SLE.16 Each patient was examined by a skilled practitioner involved in the care of LE patients (including A.C., C.F., and D.L.). We recorded the type of cutaneous lesion of LE, the date of the diagnosis of LE and of its first manifestations, the duration of the disease, the type and the number of ACR criteria, the presence and level of antinuclear DNA (anti-nDNA) antibodies and their specificities, and the levels of C3 and C4. Systemic involvement was quiescent in all patients during the preceding 6 months, and treatment was unchanged during that period.
For each case, 2 controls matched for sex and year of birth (±5 years) were randomly selected from the relatives (mainly husband and wife) of the cases included in the study and other patients undergoing surgery (mainly excision of nevi).
A self-administered questionnaire was obtained from the LE patients and from the controls. We asked the patients about their smoking habits (the number of cigarettes smoked a day and the patients’ ages when they started and stopped smoking) and their alcohol consumption (frequency, type, and quantity of alcoholic drinks consumed). Regular cigarette smoking was defined as smoking at least 1 cigarette a day for at least 3 months. Cigarette smoking was quantified in pack-years, and alcohol consumption was quantified in grams per year. We also considered the time between the diagnosis of LE and the beginning of the use of tobacco and alcohol.
The statistical significance of smoking history and alcohol as associated factors for LE was analyzed by estimating matched case-control ORs and their 95% confidence intervals (CIs) with the use of multiple conditional logistic regression and multiple stratified logistic regression. To investigate differences in the quantities of cigarettes smoked and alcohol consumed, the Breslow-Day test was used.17
All patients and controls gave oral consent to participate in the study. Under French law, a study that relies on a questionnaire only does not need approval of an ethics committee or institutional review board.
We included 108 LE patients (28 male and 80 female; female to male ratio, 3.3) and 216 controls in the study. The mean age of patients was 40.1 (range, 17-73) years. Subtypes of LE included DLE or disseminated DLE in 48.1%, subacute LE in 24.1%, and acute LE in 15.7% (Table 1). The remaining patients had tumid LE or SLE with no specific cutaneous lesions (Table 1). The latter group included 6 patients with SLE who had nonspecific cutaneous lesions of LE such as pyoderma gangrenosum, aseptic pustules, or vasculitis. Mean duration of the disease was 7.4 years (range, 1 month to 45 years), and the mean time from the first manifestation of LE to the diagnosis was 1.5 years (range, 1 month to 20 years).
There was a great disparity between the number of male patients included in the different centers. Center 1 included 53 patients with a female to male ratio of 1.2, whereas center 2 included 33 female patients and only 1 male patient (female to male ratio, 33.0), and center 3 included 21 patients with a female to male ratio of 6.0.
Table 2 shows the distribution of smoking and alcohol status among the cases and controls. Overall, significantly more of the LE patients (73.1%) were smokers compared with the controls (49.5%) (P < .001) (OR, 2.77; 95% CI, 1.63-4.76). The Breslow-Day test showed an effect of center (P < .001) (Table 2). Thus, the ORs for each center are 9.26 (95% CI, 3.65-26.5) for center 1, 1.20 (95% CI, 0.47-3.14) for center 2, and 1.00 (95% CI, 0.31-3.26) for center 3.
There was no significant difference in alcohol consumption between LE patients and controls in the entire population or in the different centers.
Among the 79 LE patients who smoked, 72 (91.1%) had started smoking before the first manifestation of LE (mean delay between initiation of smoking and the first signs of LE, 14.1 years [range, 6 months to 49 years]). Only 2 patients were former smokers, and 7 started smoking after the diagnosis of the disease. At the time of the study, they had smoked for a mean of 20 years (22 years for male and 18.5 years for female patients). Mean cigarette consumption was 15 cigarettes a day (15.4 cigarettes a day for male and 14.9 cigarettes a day for female patients).
The prevalence of smoking was higher in LE patients than in controls of either sex (Table 3). The difference was greater in male subjects. Male subjects smoked more often than female subjects among LE patients and controls (Breslow-Day test, P = .02). This could explain the center effect observed in Table 1. Indeed, recruitment was obviously very different among the 3 centers. Center 1 included significantly more male patients than centers 2 and 3, which included mostly female patients. However, this effect of center disappeared when the quantity of tobacco use (pack-years) was analyzed in addition to the prevalence (Table 4). Thus, when LE patients were smokers, they smoked significantly more than controls did (P = .002).
The prevalence of smoking among patients who met 4 or more ACR criteria and/or who had anti-nDNA antibodies was lower than in patients who met fewer than 4 ACR criteria (P = .002). Among the 43 patients (39.8%) who met 4 or more ACR criteria, 23 (53.5%) were smokers and had smoked on average 5.5 pack-years. Among the 65 patients (60.2%) who met fewer than 4 ACR criteria, 55 (84.6%) were smokers and had smoked on average 16 pack-years. Among the 39 patients with anti-nDNA antibodies, 24 (61.5%) were smokers. Among the 69 patients without anti-nDNA antibodies, 55 (79.7%) were smokers. The LE patients with cutaneous involvement, fewer than 4 ACR criteria, and no anti-nDNA antibodies were those who smoked most. Patients who met 4 or more ACR criteria and who had anti-nDNA antibodies were those who smoked the least, even when compared with controls (Table 5 and Table 6).
An analysis of the prevalence of smoking within the different subgroups of patients with cutaneous lesions of LE showed no statistical difference (data not shown) when we considered the prevalence of smoking and the quantity of cigarettes consumed in pack-years, despite a very high prevalence of smoking among LE patients with the discoid disseminated (20 patients [86.9%]) and subacute (18 [69.2%]) subtypes. This is probably the consequence of the small size of the population in the different subgroups. Multiple logistic regression models and simple analysis gave the same results. Thus, for ease of presentation, only the latter were provided.
The current view of the etiology of LE is that several environmental factors act in a genetically predisposed individual to induce the disease. Our data suggest that smoking is associated with an increased risk of LE, whereas alcohol consumption is not. In this study, we showed that cigarette smoking is particularly prevalent in patients who met fewer than 4 ACR criteria and that a dose effect exists.
Results concerning alcohol consumption and LE are conflicting. Some studies demonstrated a protective effect,3-5 whereas others showed no significant association,12,18 as in our study. In our medical practice, we have observed that men and women with severe disseminated DLE and those who met fewer than 4 ACR criteria of SLE were often alcoholic, but this finding was not confirmed in this study. Cigarette smoking, often associated with alcohol consumption, may be a confounding factor.
Some previous studies have demonstrated that smoking is a risk factor for LE.3-7 The prevalence of current smoking in LE patients varies from 17% to 46%, and a recent meta-analysis of 9 studies (7 case-control studies and 2 cohort studies)19 revealed an association between cigarette smoking and LE (OR, 1.50; 95% CI, 1.09-2.08). All of these studies focused on patients with SLE, including mainly women who met 4 or more ACR criteria. Miot et al7 explored cigarette smoking in DLE patients who met fewer than 4 ACR criteria and they found a very high prevalence of smoking (84.2%) in LE patients compared with the control population (33.5%) (OR, 14.4; 95% CI, 6.2-33.8). A strong association between LE and tobacco was also observed in the study of Gallego et al6 (OR, 12.2), who included patients comparable to those of Miot et al.7
One source of the originality and strength of this study is it was conducted in 3 centers with different patient recruitment. Center 1 included a huge proportion of male patients, and overall most of the patients had cutaneous LE without SLE (37 patients [69.8%]). Thus, center 1 is representative of patients seen in general dermatology practices. Center 2 included almost exclusively women with SLE (21 patients [61.7%]), as seen in internal medicine or rheumatology departments, and center 3 included 11 patients (52.3%) who met 4 or more ACR criteria. Statistical analysis revealed that the difference in the prevalence of cigarette smoking between the different centers seemed related not to LE subtype but to the patient's sex, with men smoking more than women. Thus, sex obviously is a confounding factor that needs to be addressed.
This effect of center also disappeared when we considered cigarette smoking according to the number of pack-years. When LE patients smoke, they smoke more than controls do. Our data also show that patients with cutaneous LE who meet fewer than 4 ACR criteria are those who smoke most compared with controls or compared with patients with anti-nDNA antibodies and/or who meet 4 or more ACR criteria. The latter group of patients are those who smoke the least, even when compared with controls. Thus, our data strongly suggest that cigarette smoking could have a greater effect on cutaneous manifestations of LE than on the extracutaneous signs of the disease such as renal involvement. This observation suggests that the genetic background is much stronger and much more important—or at least different—in patients who meet 4 or more ACR criteria than in patients with cutaneous LE who meet fewer than 4 ACR criteria, in which environmental factors like cigarette smoking could be crucial. This difference in patient recruitment could explain the discordant results of the association between LE and cigarette smoking in previous studies. The fact that LE patients with anti-nDNA antibodies are those who smoke the least contradicts a study showing that current cigarette smoking is associated with DNA antibodies.20
Ninety-one percent of LE patients who were smokers began to smoke a long time (mean, 14 years) before the first manifestations of the disease, thus minimizing bias associated with reactive habits induced by the disease. The long time lapse between the initiation of smoking and the first signs of disease suggests that there is a cumulative effect involving an important accumulated smoking exposure. The dose-response relationship between the number of pack-years of smoking and the risk of LE is further supported by the strong statistical relationship with a pack-year effect. The mean age for starting smoking was quite precocious (18 years), with a high number of cigarettes smoked a day (mean, 15 cigarettes a day).
Thus, this study provides strong evidence of a cumulative and dose effect, and cutaneous involvement seems the main target affected by cigarette smoking. Hardy et al4 suggested a dose effect, but the difference of cigarette smoking in pack-years between LE patients and controls did not reach significant difference in their study.
How cigarette smoking could trigger LE remains unclear, but some hypotheses exist. Cigarette smoke contains more than 100 toxic and carcinogenic substances. Some of them are aromatic amines, substances that are also contained in drugs known to be associated with LE,21 especially subacute LE. Exposure to UV light is also a well-documented trigger of LE, especially of the cutaneous manifestations.1 Another study has shown that cigarette smoke is phototoxic.22 From an immunological point of view, cigarette smoking has very complex immunomodulatory effects, recently reviewed by Costenbader and Karlson.23 Furthermore, it has been known for many years that antimalarials are less efficient in smokers.24,25 Although the metabolism of antimalarials is not well understood, it is possible they are inactivated by the cytochrome P450 enzyme complex. Polycyclic aromatic hydrocarbons found in cigarette smoke are known potent inducers of the P450 enzyme complex.26 Finally, cigarette smoking may not be a risk factor per se, but predisposition to LE and dependency on cigarette smoking could have a common genetic background, as discussed by Boeckler et al14 and Füst et al,27 because a cluster of genes for olfactory receptors implicated in smoking behavior and addiction is located close to the HLA antigen class I region, of which some alleles are known genetic risk factors for SLE.
Correspondence: Dan Lipsker, MD, PhD, Clinique Dermatologique, 1, Place de l’hôpital, 67091 Strasbourg CEDEX, France (email@example.com).
Accepted for Publication: February 19, 2009.
Author Contributions: All authors had full access to data. Study concept and design: Boeckler and Lipsker. Acquisition of data: Boeckler, Cosnes, Francès, and Lipsker. Analysis and interpretation of data: Boeckler, Hedelin, and Lipsker. Drafting of the manuscript: Boeckler and Lipsker. Critical revision of the manuscript: Boeckler, Cosnes, Francès, Hedelin, and Lipsker. Statistical analysis: Hedelin. Study supervision: Lipsker.
Financial Disclosure: None reported.
Create a personal account or sign in to: