A number of studies have suggested that smoking is a risk factor in susceptibility to rheumatoid arthritis (RA) (1–10). While there have been fewer investigations of the impact of smoking on disease outcome, 3 recent studies have suggested that heavy smoking may influence overall RA severity (11–13). All of these studies reported a relationship between smoking and rheumatoid factor (RF) positivity, nodule formation, and radiographically determined joint damage. Smoking was also associated with increased functional impairment (as determined by scores on the Health Assessment Questionnaire [HAQ]), lower grip strength, and more pulmonary disease (12, 13), although these observations may be related to comorbidity effects of smoking rather than the severity of RA per se. In one study, the RF concentration was found to be positively correlated with the number of years smoked (12), while another study showed that the levels of IgA-RF and IgM-RF, but not IgG-RF, were associated with the number of pack-years smoked (13).
The mechanism by which smoking influences RA susceptibility/severity is unclear at present, although it may have direct effects on the disease process by inducing and/or increasing the production of RF or by producing alterations in the immune system (3, 14–17). The influence of genetic factors on the association between smoking and RA is unknown, although a recent study reported that heavy cigarette smoking was particularly associated with RA in patients who had no family history of the disease (9). We postulate that genes associated with detoxification or activation of chemicals in tobacco smoke will be important. In this regard, our previous finding that increased severity in RA is associated with a null polymorphism at the glutathione S-transferase (GST) M1 locus is of particular interest (18).
The GSTs are a widely expressed group of enzymes that catalyze the detoxification of xenobiotics via glutathione conjugation. They are also believed to play an important role in detoxifying products generated by the activity of reactive oxygen species (ROS). There is evidence that some allelic variants are associated with differences in detoxification efficiency, and various cancer studies have suggested that polymorphism of GSTM1 may influence the ability to detoxify chemicals in cigarette smoke (19–21).
The GSTM1 enzyme detoxifies known or suspected carcinogens found in tobacco smoke. These include benzo[α]pyrene and other polycyclic hydrocarbons (19). The GSTM1 gene has 2 functional alleles (GSTM1*A and GSTM1*B) and a nonfunctional null allele caused by deletion of the GSTM1 sequence. Homozygosity for GSTM1-null occurs in ∽45–55% of Caucasians. The GSTM1-null polymorphism has been associated with increased risk of smoking-related cancers (19–25), and a synergistic interaction has been observed between smoking and GSTM1-null on the risk of coronary disease (26).
In this study, we sought to confirm that cigarette smoking was associated with more severe disease in RA and to test the hypothesis that the impact of smoking on disease outcome is associated with polymorphism of the GSTM1 gene which is involved in the detoxification of chemicals in cigarette smoke. The hypothesis was developed a priori with GSTM1-null being considered as the putative high-risk genotype. The study was performed only on female RA patients. Most of the male RA patients (∽90%) in the population had been smokers, and it was therefore not possible to test for statistically significant differences between male patients who had smoked and those who had not.
PATIENTS AND METHODS
- Top of page
- PATIENTS AND METHODS
Patients. The association between GSTM1 genotypes, smoking, and disease severity was studied in 164 unrelated female RA patients residing in North Staffordshire, UK. All patients were Northern European Caucasians. The characteristics of these patients are shown in Table 1.
Table 1. Demographic and clinical characteristics of the 164 female patients with rheumatoid arthritis
|Median (range)||No. (%) of patients|
|Age, years||57.0 (32–79)||–|
|Age at disease onset, years||46.0 (19–73)||–|
|Disease duration, years||11.0 (5–30)||–|
|Rheumatoid factor positive||–||102 (62.2)|
|Nodular disease||–||31 (18.9)|
|Erosive disease||–||148 (90.2)|
|Previous joint surgery||–||54 (32.9)|
The patients fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 criteria for RA (27) and were recruited in a clinic that had been established to examine the effects of disease-modifying antirheumatic drugs (methotrexate, sulfasalazine, hydroxychloroquine, gold, and D-penicillamine). About 6.5% of the patients were being treated with corticosteroids. Therapy was administered as clinically indicated.
All patients had been examined annually for at least 5 years, and their disease had been extensively characterized. Outcome measures were recorded at the final review and consisted of assessments of functional status, using the HAQ (28), and radiographic outcome, scoring radiographs of the hands and feet according to the standard radiographs of Larsen et al (29). RF levels were measured by nephelometry; a level >60 IU/ml was considered positive (30).
The Carstairs Deprivation Index (31, 32) was used to categorize the socioeconomic status of the patients. Carstairs scores were obtained from the 1991 census small-area statistics for the UK and assigned to each patient based on their enumeration district of residence, which was identified from the postal code address. The index is based on a composite of 4 variables: male unemployment, social class of head of household, overcrowding, and access to a car. The Carstairs scores in the RA population for this study ranged from −4.6 (least deprived) to +5.81 (most deprived).
Smoking history. A current and past smoking history was obtained from each patient. Patients were initially classified by whether they had ever smoked or never smoked. Ever smokers were those who had smoked at least 1 cigarette/day for ≥1 year, but not necessarily during the time of the study. All patients who had ever smoked had started smoking before the onset of RA. Those who had ever smoked were further categorized into past and current smokers. Past smokers were those who had stopped smoking at least 3 months before entry into the study. Of the 35 past smokers, 21 had stopped smoking before the onset of RA. The extent of smoking was quantified in pack-years. One pack-year is equivalent to 20 cigarettes/day for 1 year.
GST typing. Leukocyte DNA was extracted from blood samples that had been collected into tubes containing EDTA. GSTM1 genotypes were defined using a polymerase chain reaction assay that identifies the GSTM1*0, GSTM1*A, and GSTM1*B alleles (33). Patients were classified into those who expressed (GSTM1-1) and those who did not express (GSTM1-0) GSTM1.
Statistical analysis. The association of disease severity measures with GSTM1 genotypes and smoking was assessed using multiple regression analyses, with adjustment for the independent variables, age, and disease duration where appropriate. The influence of RF and the Carstairs Deprivation Index were also examined by inclusion as independent variables in some of the regression models. Additional analyses were performed after stratification by GSTM1 status. In some analyses, we also examined the effect of interaction between variables by use of multiple regression models that contained the interaction term as well as the corresponding main effects. Where appropriate, correction for potential multiple testing errors was performed using Holm's procedure (34). All analyses were carried out using the Number Cruncher Statistical Package for Windows (version 6.0.4; NCSS, Kaysville, UT), or the PEPI software package (version 2.0) for epidemiologic analysis (35).
- Top of page
- PATIENTS AND METHODS
Frequency and extent of smoking in female RA patients.About one-half (51.2%) of the patients had ever smoked (Table 2). The number of smokers had dropped to 29.9% by the time of the study. In this group, the mean duration of smoking and the mean number of pack-years were 38.9 years and 29.3 years, respectively (Table 3), which were significantly greater than in past smokers (23.5 years and 18.5 years, respectively). There was a significantly higher frequency of seropositive (RF+) disease among current smokers than among those who had never smoked.
Table 2. Frequency of cigarette smoking and seropositivity in 164 female patients with rheumatoid arthritis
|Smoking status||No. (%) of patients||% positive for rheumatoid factor|
Table 3. Extent of cigarette smoking in female patients with rheumatoid arthritis
|Smoking status||No. of patients||Years smoked, mean ± SD||Pack-years, mean ± SD|
|Ever||84||32.5 ± 13.3||24.8 ± 16.1|
|Past||35||23.5 ± 13.4||18.5 ± 17.5|
|Current||49||38.9 ± 9.7*||29.3 ± 14.4†|
Relationship between smoking and disease severity in female RA patients. The mean Larsen score in patients who had ever smoked was significantly higher than that in patients who had never smoked (Pcorr = 0.01, corrected for age and disease duration) (Table 4). The trend was similar in current smokers (Pcorr = 0.05). The highest mean Larsen score (112.2) was found in past smokers, and this was significantly higher than the Larsen score in patients who had never smoked (Pcorr = 0.006). Using multiple regression analyses (corrected for age and disease duration), we found that in patients who had ever smoked, there was no association between the Larsen score and the number of pack-years smoked (P= 0.43).
Table 4. Relationship between smoking and radiographic or functional outcome in female patients with rheumatoid arthritis*
|Outcome measure||Smoking status|
|Never (n = 80)||Ever (n = 84)||Past (n = 35)||Current (n = 49)|
|Larsen score||83.1 ± 47.2||104.7 ± 49.9†||112.2 ± 47.7‡||99.3 ± 51.3§|
|HAQ score||1.39 ± 0.8||1.77 ± 0.8‡||1.86 ± 0.7¶||1.71 ± 0.8#|
The HAQ scores showed a trend similar to that of the Larsen scores. Ever, past, or current smokers had significantly increased HAQ scores compared with those who had never smoked (Table 4). Again, no association was found between the HAQ score and the number of pack-years smoked (P= 0.68).
Influence of GSTM1 polymorphism on Larsen and HAQ scores. Comparison of Larsen scores in GSTM1-null and GSTM1-1 patients by multiple regression analysis revealed a trend toward more severe disease in GSTM1-null patients, which is similar to that reported previously (18), although this did not achieve significance after correction for age and disease duration (99.4 versus 86.8; P= 0.1). There was no significant difference in the HAQ scores between the 2 subgroups (1.57 versus 1.59).
To investigate whether polymorphism of the GSTM1 gene influenced the response of RA to smoking, we stratified patients by GSTM1 status (those expressing versus those not expressing GSTM1) as well as by never smoking or ever smoking. GSTM1-null patients who had ever smoked had significantly higher Larsen (Pcorr = 0.012) and HAQ (Pcorr = 0.003) scores than GSTM1-null patients who had never smoked (Table 5). Radiographic outcome in GSTM1-null patients who had ever smoked was also worse than that in GSTM1-expressing patients who had ever smoked (Pcorr = 0.05), although there was no significant difference in the HAQ scores. In GSTM1-expressing patients, there were no significant differences in the Larsen or HAQ scores between ever smokers and never smokers.
Table 5. Relationship between ever smoking and radiographic or functional outcome in female patients with rheumatoid arthritis, stratified by GSTM1 status*
| ||Smoking status|
|No. of patients||Mean ± SD||No. of patients||Mean ± SD|
| GSTM1-0||43||82.8 ± 49.7||53||112.8 ± 53.3†‡|
| GSTM1-1||37||83.5 ± 44.9||31||90.8 ± 40.9|
| GSTM1-0||41||1.28 ± 0.8||50||1.81 ± 0.8§|
| GSTM1-1||37||1.50 ± 0.9||31||1.71 ± 0.7|
Association between GSTM1 polymorphism, smoking, and RF production. There was a significant association between ever or current smoking and RF positivity (P= 0.006 and 0.004, respectively) after correction for age and disease duration (Table 2). In all smokers, the RF level was also associated with the number of years of smoking (P= 0.02). A linear relationship between RF level and the number of pack-years was not found, but smokers with >20 pack-years had significantly higher RF levels than those with <20 pack-years (264.3 versus 99.9 IU; P< 0.001).
We postulated that the development of more severe disease in GSTM1-null patients who smoked might be associated with a difference in their production of RF. We therefore examined the association between RF status and smoking in patients stratified by GSTM1 genotype. Using logistic regression analysis (adjusted for age and disease duration), we found no significant difference in RF status between ever smoking and never smoking (67.7% versus 51.4%; odds ratio [OR] 1.9, P= 0.2) in GSTM1-1 patients. However, in GSTM1-null patients, there was a difference in RF status between ever and never smoking (75.5% versus 51.2%; OR 3.1, P= 0.01). Division into past and current smokers revealed that this difference was specifically due to a difference between current and never smoking in these patients (83.9% versus 51.2%; OR = 5.1, P= 0.006) (Table 6). No significant difference in RF status was found between current smoking and never smoking in GSTM1-1 patients (64.7% versus 51.4%; OR = 1.7, P= 0.4).
Table 6. Relationship between GSTM1 status, current smoking, and RF production in female patients with rheumatoid arthritis*
| ||RF frequency, no. (%) of patients||RF levels, mean ± SD IU|
|Current smokers||Never smoked||Current smokers||Never smoked|
|GSTM1-0||26/31 (83.9)†||19/37 (51.2)||284.7 ± 302.9‡||69.8 ± 137.8|
|GSTM1-1||11/17 (64.7)||21/41 (51.4)||175.4 ± 289.5||74.7 ± 138.1|
With regard to the amount of RF produced, there was a significant association between current smoking and RF levels in GSTM1-null patients (P= 0.002), but no association in GSTM1-1 patients (P= 0.4). Confirmation of the relationship between current smoking and RF status in GSTM1-null patients (P= 0.005) but not GSTM1-1 patients (P= 0.3) was obtained from a separate cohort of RA patients (n = 134) with early disease (median disease duration 1 year) (data not shown).
Multiple regression models which included RF (positive or negative) as well as GSTM1-null + ever smoking as independent variables showed that RF status had the strongest association with the Larsen score after correction for age and disease duration (Table 7). Nonetheless, after correction for RF status, GSTM1-null + ever smoking remained associated with radiographic outcome, although the significance levels were greatly reduced. A similar observation was found for the HAQ score, although the association with RF status was weaker and age had a more significant effect than disease duration (Table 8). These data suggest that the association of GSTM1-null + ever smoking with disease severity is not due solely to RF status. However, the possible contribution of other, as yet undiscovered, confounder or modifier variables to the association of this combination variable with the outcome measures cannot be excluded.
Table 7. Multivariate determinants of the Larsen score in female patients with rheumatoid arthritis*
|Variable||Regression coefficient||Standard error||P|
|GSTM1-0 + ever smoked||17.097||7.016||0.033|
Table 8. Multivariate determinants of the HAQ score in female patients with rheumatoid arthritis*
|Variable||Regression coefficient||Standard error||P|
|GSTM1-0 + ever smoked||0.434||0.141||0.002|
|Carstairs Deprivation Index score||0.040||0.023||0.08|
Influence of socioeconomic status. We investigated whether socioeconomic deprivation had any effect on disease severity. There was no association between the Carstairs Deprivation Index scores and radiographic outcome in univariate or multivariate models (data not shown). In contrast, the HAQ score was associated (P= 0.015) with the Carstairs scores in models corrected for age and disease duration only, but lost significance in multivariate models containing GSTM1-null + ever smoking and RF status (Table 8). Removal of GSTM1-null + ever smoking from the latter model resulted in a weakly significant association between the Carstairs and HAQ scores (P= 0.04).
- Top of page
- PATIENTS AND METHODS
We have shown in a well-characterized group of female patients with RA that a history of cigarette smoking is associated with more severe disease, although the mechanism for this effect remains unclear. We also provide evidence that the relationship between smoking and disease severity in these patients is associated with polymorphism at the GSTM1 locus.
Our data suggest that the risk of developing severe disease in female RA patients is increased in those who have the GSTM1-null polymorphism and who have also smoked. The difference between these patients and GSTM1-null patients who had never smoked was highly significant, although significance levels were reduced after correction for RF status. In contrast, there was no significant difference between smokers and nonsmokers who had the functional GSTM1-1 phenotype. Also, in nonsmokers, there was no difference in outcome between GSTM1-null and GSTM1-1 patients. Thus, deletion of the GSTM1 gene per se does not appear to influence RA severity, but does so only in individuals who also smoke. The difference in disease outcome between patients who have never smoked and those who have ever smoked appears to be accentuated in individuals with the GSTM1-null polymorphism. Given the known role of the GSTM1 enzyme in detoxifying chemicals in cigarette smoke, this result suggests that such substrates may have an important influence on the progression of RA.
We confirmed the observations of other studies (11, 12), which found that patients who smoked were more likely to be RF positive than were nonsmokers and that the number of years smoked was associated with levels of RF. Of particular interest was the finding that the association between current smoking and RF status was significant only in GSTM1-null patients, where those who smoked had the highest levels of RF. The correlation between smoking and RF has made it difficult to determine the independent predictive value of each of these factors in relation to disease severity. From our data, it is possible to speculate that lack of the GSTM1 enzyme in smokers may promote increased RF production through a failure to detoxify smoke-derived chemicals (or their byproducts), which have the potential to damage IgG. Previous studies have suggested that free radical–mediated alteration of IgG may stimulate the formation of immune complexes with RF antibody, thereby promoting tissue damage during rheumatoid inflammation (36, 37).
Saag et al (11) suggested that smoking may be more important in the initiation of erosive disease than in the perpetuation of the disease process. The lack of an obvious dose-related effect of smoking on the amount of radiographic damage in patients in our study may add some weight to this idea. However, our data also suggest that if smoking is involved in the initiation process, it leads to more severe disease than it does in patients in whom smoking is not involved (particularly if the patients are GSTM1-null).
It might have been expected that increased exposure to cigarette smoke would have led to increased damage and more severe disease. However, the outcome in past smokers was as severe as in those who continued to smoke, even though the number of years smoked and the number of pack-years were significantly lower in the former group. Nonetheless, the mean duration and amount of smoking in the past smokers was relatively high and, compared with other studies, was close to the levels considered to represent a history of heavy smoking (7, 11, 12). There is evidence to suggest that individuals who smoked in the past continue to produce RF, even after cessation of smoking (12, 38). Thus, tissue damage mediated by RF complexes may continue long after the initial stimulus for RF production has been removed.
An alternative explanation for the lack of a dose-related effect could be that some other factor(s) in female smokers predisposes them to the development of more severe RA, especially if they are GSTM1-null. Potential confounders such as body mass index, estrogen levels, oral contraceptive use, or alcohol intake may be important, but without available data on these variables, we were unable to test their effect in our study. However, a recent report from Karlson et al (7) suggested that body mass index and oral contraceptive use were not significant confounders in the association between smoking and RA susceptibility. Other possible confounders such as diet or exposure to environmental toxins may also have some influence but would be difficult to quantify.
The observed associations do not appear to be directly related to social deprivation, since adjustment for socioeconomic status using the Carstairs Deprivation Index did not affect the association between smoking and disease severity in these patients. This is consistent with the report by Saag et al (11), who found that the association of smoking with radiographic changes or RF positivity did not vary with any sociodemographic factors tested. There was no association between the Carstairs score and radiographic outcome with or without inclusion of GSTM1-null + smoking in our regression models, although we did find a relationship between functional outcome and the Carstairs score after exclusion of GSTM1-null + smoking. This is consistent with previous studies that showed a relationship between functional outcome and social deprivation (39, 40). However, those studies did not investigate the influence of smoking, and our data suggest that the relationship between social deprivation and functional outcome is at least partly due to the influence of smoking. It needs to be borne in mind that the HAQ score may not be a specific indicator of the functional impact of RA severity per se. This may be particularly true among patients who smoke, given that smoking is likely to have adverse effects on general health and function.
One limitation of this study is its cross-sectional nature and the difficulty in establishing cause and effect. It has been argued that such studies are subject to misclassification of smoking status and that patients with very mild disease (who no longer seek medical care) or patients with very severe disease (who may have died) are underrepresented (41). While prospective studies may overcome some of these difficulties, they tend to be of relatively short duration and may not properly address the long-term impact of smoking on disease severity. They may also suffer from the difficulty of classifying RA in patients with early inflammatory polyarthritis. We do not believe that misclassification of smoking status is a problem in our current study, especially since we found no difference in disease outcome between past and current smokers. However, our study is clearly limited to women with well-established RA in one particular geographic area, and further studies of different populations are needed to establish the generalizability of these findings.