Alcohol consumption reduces the risk of development of rheumatoid arthritis (RA) and significantly attenuates the development of erosive arthritis in animal models. It remains unknown whether alcohol consumption influences joint damage progression in RA. This study was undertaken to compare the rates of radiographic damage progression in alcohol drinkers and nondrinkers in a large prospective cohort of patients with RA.
All patients in the population-based Swiss Clinical Quality Management in RA registry database with at least 2 sequential radiographs were included. Joint erosions were assessed in 38 joints in the hands and feet using a validated scoring method. The rate of progression of erosions was analyzed using multivariate regression models for longitudinal data and was adjusted for potential confounders.
The study included 2,908 patients with RA with a mean of 4 sequential radiographs and 3.9 years of followup. A trend toward reduced radiographic progression existed in drinkers compared with nondrinkers, with a mean rate of erosive progression of 0.99% (95% confidence interval [95% CI] 0.89–1.09) and 1.13% (95% CI 1.01–1.26) at 1 year, respectively. Alcohol consumption displayed a J-shaped dose-response effect, with a more favorable evolution in occasional consumers (P = 0.01) and daily consumers (P = 0.001) as compared with nondrinkers, while heavy drinkers demonstrated worse radiographic evolution (P = 0.0001). We found significant effect modification by sex, with male drinkers displaying significantly less erosive progression compared with male nondrinkers (mean 0.86% [95% CI 0.70–1.03] versus 1.35% [95% CI 1.02–1.67]; P = 0.007).
Our findings indicate a trend toward reduced radiographic progression in alcohol drinkers compared with nondrinkers, specifically in occasional and daily alcohol consumers. In particular, male patients with RA who consume alcohol demonstrate less radiographic progression than do male nondrinkers.
Although the cause of rheumatoid arthritis (RA) remains unknown, environmental risk factors are thought to be important in its pathogenesis, with smoking representing the best-established risk factor (1). The rate of radiographic progression of joint damage can be used as a proxy measure for RA severity, representing cumulative disease activity for an individual patient. Many studies have demonstrated that radiographic joint damage is associated with long-term disability (2, 3), loss of employment status (4), and social security disability status (5). Despite the plethora of new and effective therapies for RA, remission is not achieved in many patients and there are ongoing signs of joint destruction. A better understanding of the genetic and environmental risk factors is therefore essential for effective management.
In addition to smoking, numerous environmental risk factors have been cited that may influence the susceptibility for developing RA, including alcohol, coffee, tea, oral contraceptive use, and socioeconomic status, with varying levels of evidence. A recent study demonstrated a protective effect of alcohol consumption on the risk of developing RA, with a 40–50% reduction in risk for the highest consumers (6), although some smaller studies, either exclusively in women or with alcohol consumption as a secondary outcome, found no association between alcohol and RA onset (7, 8). In murine collagen-induced arthritis, the development of erosive arthritis is almost totally abolished in mice exposed to low levels of alcohol on a daily basis (9). Nevertheless, it remains unknown whether alcohol influences disease severity or disease progression in established RA. The aim of this study was to assess the influence of alcohol consumption on disease progression in a large prospective cohort of patients with RA.
PATIENTS AND METHODS
The Swiss Clinical Quality Management (SCQM) project for RA is a longitudinal population-based cohort of RA patients. Currently, more than two-thirds of practicing rheumatologists in Switzerland are contributing patients to the database. The SCQM database collects information on sociodemographic factors, lifestyle characteristics (including alcohol consumption), antirheumatic therapies, disease activity, and radiographic damage at least annually. All patients provided written informed consent prior to inclusion in the database.
We conducted a longitudinal observational study of a population-based cohort of patients with RA. The analysis included data collected between March 1996 and February 2009. All patients in the database who met the inclusion criteria of a confirmed diagnosis of RA by a rheumatologist, known alcohol consumption status, and at least 2 consecutive sets of radiographs of the hands and feet were included. Patients were excluded if their alcohol status was unknown or if followup radiographs were missing.
The primary outcome of the study was the progression of radiographic joint damage, as measured by changes from baseline in radiographic damage scores. Radiographic damage was assessed prospectively by a single evaluator who was blinded with regard to clinical information; a validated scoring system (the Ratingen score) was used (10). This technique of radiologic scoring was chosen because it is sensitive to change and is less susceptible to the ceiling effects seen in advanced disease because of a true ordinal rating system (11). The Ratingen scoring method has excellent reliability, with an intraclass correlation coefficient (ICC) for intrarater reliability of 0.8–0.9 and an ICC for interrater reliability of 0.7–0.9 (10, 12). The minimal detectable radiographic change for this method has been determined to be 3.3% of the maximum score (10). For our study assessor, the intrarater reliability was good, with an ICC of 0.94 for a cross-sectional assessment and an ICC of 0.71 for change scores.
A secondary outcome of the study was the progression of functional disability, measured as change from baseline in the Health Assessment Questionnaire (HAQ) disability index (13). The HAQ score ranges from 0 to 3, where 3 represents the maximum possible disability. This score has been demonstrated to predict work disability (14), medical expenditure (15), and mortality (16).
Exposure variable and predictors.
All patients were categorized as drinkers or nondrinkers of alcoholic beverages, based on the self-reported drinking status in the patient questionnaire. Patients who discontinued drinking alcohol and those who started drinking during the observation period were categorized as drinkers. In order to explore a potential dose-response effect of drinking, we further stratified drinkers as occasional drinkers, daily drinkers (consumption of alcoholic beverages on 1 occasion per day), and heavy drinkers (consumption of alcoholic beverages on several occasions per day). Alcohol drinkers who changed their category of consumption during the observation period were classified into the higher category. Other important predictors of disease progression in RA, such as disease activity measures, self-assessed symptom questionnaires, various disease characteristics, and demographic characteristics, were extracted from the database and included in the analyses. We determined the dominant antirheumatic treatment regimen used during the time span between consecutive radiographic assessments and used this variable to control for the analysis of disease-modifying antirheumatic drug (DMARD) and biologic agent use. The dominant antirheumatic treatment was defined as the regimen used for >50% of the observation period.
Based on previous studies with this cohort (17, 18), we estimated that a sample size of 100 patients per group would provide 90% power with an alpha error of less than 0.05 to detect a 2-sided difference of at least 1 Ratingen score unit (0.5% of the maximum score) in radiographic damage progression per year. Differences in baseline disease characteristics between the groups were compared using Student's t-test for normally distributed variables, Wilcoxon's rank sum test for non-normally distributed continuous variables or categorical variables, and Pearson's chi-square test for dichotomous variables. Less than 5% of covariates were sporadically missing. To minimize potential bias, we used the population average of the respective covariates as a substitute.
Confounding was a concern in this study, since it is possible that RA patients who regularly consume alcohol differ from patients who do not drink alcohol with regard to important disease characteristics. Given that these differences could substantially influence disease progression, we used multivariate adjustments to correct for such confounding effects. Radiographic damage progression and functional disability evolution were analyzed using generalized mixed models for longitudinal data (19). We verified that the multivariate normal assumptions for longitudinal models were satisfied and examined whether time as a linear trend or as a polynomial function best fit the data. We adjusted the analysis for differences in baseline disease characteristics. Rheumatoid factor (RF), age, sex, education level, Disease Activity Score in 28 joints (DAS28) (20), HAQ status, DMARD therapy, and corticosteroid use at baseline were considered to be a priori confounders and were entered into the model. We tested other covariates using a backwards stepwise selection approach. The estimated rate of radiographic progression at baseline was calculated, since this is one of the strongest predictors of future radiographic progression (21).
The final model was adjusted for differences in baseline damage scores, DAS28, functional disability (as measured by the HAQ), presence of RF, sex, age, disease duration, tobacco smoking, education level, and medications (including DMARDs, biologic agents, and corticosteroids). Point estimates of the regression model at every year for the median duration of followup were used to produce the graphs showing the results. We considered the possibility of a dose response by alcohol intake by classifying the alcohol exposure variable into 4 categories (none, occasional, daily, and heavy [several drinks per day]), according to the categories on the patient questionnaire. We further investigated whether effect modification of alcohol existed in various subgroups, such as those classified by sex, RF status, smoking status, anti–cyclic citrullinated peptide (anti-CCP) status (present, absent, or missing), or therapy with biologic agents. All statistical tests were 2-sided and were evaluated at the 0.05 significance level. Statistical analysis was performed with Stata, version 9.2 for Windows (Stata Statistical Software).
Of the 4,834 patients with RA in the SCQM registry, a total of 2,908 with a median of 4 sequential radiographs and 3.9 years of followup met the study inclusion criteria. Excluded patients were missing either followup radiographs (n = 1,774), alcohol consumption status (n = 149), or both (n = 3). Because missing radiographic followup was related to recent enrollment in the database (since excluded patients enrolled, on average, 3.3 years later than included patients [P < 0.0001]), we assumed absent followup to be missing at random.
In this RA population, 1,084 patients (37%) reported no alcohol consumption and 1,824 patients (63%) reported at least some consumption of alcohol. The alcohol drinkers included more male patients than the nondrinkers, were younger, included a greater percentage of smokers, and had shorter disease durations, lower DAS28 scores, and lower HAQ scores, with consequently less joint erosion at baseline (Table 1). The median estimated erosive progression at baseline was, however, similar between drinkers (1.01 [interquartile range 0.31–2.24]) and nondrinkers (1.04 [interquartile range 0.32–2.40]). Among the drinkers, 1,486 (81.5%) were occasional drinkers, 272 (14.9%) were daily drinkers, and 57 (3.1%) were heavy drinkers. Nine of the 2,908 patients (0.3%) indicated that they drank alcohol, but did not specify their intake category and were therefore excluded from the dose-effect analyses. Four patients changed their drinking classification during followup, with 2 patients reclassifying themselves as drinkers and 2 patients becoming nondrinkers. These patients were considered to be drinkers in the analyses. Other important risk factors of disease progression, such as RF seropositivity and type of antirheumatic therapy, did not differ significantly between drinkers and nondrinkers. The use of oral corticosteroids was, however, less common in the alcohol drinkers than in the nondrinkers (46.1% versus 53.1%; P < 0.001).
Table 1. Baseline characteristics of the patients with RA*
Except where indicated otherwise, values are the median (95% confidence interval). RA = rheumatoid arthritis; RF = rheumatoid factor; IQR = interquartile range; DAS28 = Disease Activity Score in 28 joints; HAQ = Health Assessment Questionnaire; erosion score = percentage of the maximum possible Ratingen damage score.
Alcohol drinkers included occasional, daily, and heavy drinkers.
By Wilcoxon's rank sum test for non-normally distributed variables, Student's t-test for normally distributed continuous variables, and the chi-square test for dichotomous variables.
Disease-modifying agent use either as monotherapy or in combination with other agents.
Use of any tumor necrosis factor inhibitor or rituximab.
In the crude analysis, the unadjusted evolution of radiographic damage was significantly less in the alcohol drinkers compared with the nondrinkers (P = 0.029). In the fully adjusted model, there was a tendency toward reduced radiographic progression in the drinkers, with radiographic damage at 1 year having progressed by a mean of 0.99% (95% confidence interval [95% CI] 0.89–1.09) in the drinkers and by 1.13% (95% CI 1.01–1.26) in the nondrinkers. The strongest predictors of radiographic damage progression were disease duration, estimated baseline radiographic damage, and RF.
Alcohol consumption displayed a J-shaped dose-response effect, with a more favorable evolution in occasional consumers and daily consumers compared with nondrinkers, while disease in heavy drinkers evolved less favorably. Radiographic progression at 1 year was 0.99% (95% CI 0.89–1.11) in the occasional consumers and 0.92% (95% CI 0.70–1.13) in the daily consumers, compared with 1.13% (95% CI 1.00–1.23) in the nondrinkers and 1.29% (95% CI 0.82–1.76) in the heavy consumers (Figure 1). Four years after inclusion, radiographic progression was 3.85% in the heavy drinkers, 3.19% in the nondrinkers, 3.06% in the daily drinkers, and 2.96% in the occasional drinkers. Although the differences between the 4 groups were not consistently different at specific time points, using direct pairwise comparisons, the longitudinal evolution of the group curves were significantly different. The adjusted evolution of radiographic damage was significantly less in occasional drinkers compared with nondrinkers (P = 0.01) and in daily drinkers compared with nondrinkers (P = 0.001), while the radiographic evolution was significantly worse in heavy drinkers compared with occasional and daily drinkers (P = 0.0001).
Moreover, we found evidence of effect modification by sex (P = 0.05). Men who drank alcohol demonstrated a more pronounced reduction in radiographic progression compared with men who did not drink alcohol (0.86% [95% CI 0.70–1.03] versus 1.35% [95% CI 1.02–1.67], P = 0.007) at 1 year, an effect which was less evident in women (1.04% [95% CI 0.92–1.15] versus 1.10% [95% CI 0.96–1.23], P = 0.49) (Figure 2). No effect modification existed with RF positivity (P = 0.62), smoking status (P = 0.25), or use of biologic agents (P = 0.13).
Progression of functional disability.
To examine the consistency of the radiographic data, we repeated the analysis with the HAQ score as the outcome. Patients had a mean of 5.7 sequential HAQ assessments during the observation period. As with the radiographic data, there was a more favorable evolution in the drinkers, with an increase in HAQ scores at 5 years of 0.36 (95% CI 0.32–0.41) in the drinkers and 0.41 (95% CI 0.35–0.46) in the nondrinkers, although this did not reach statistical significance (P = 0.18) (Figure 3). There was no significant effect modification in HAQ scores by the quantity of alcohol consumption, sex, RF positivity, or the use of biologic agents. The strongest predictors of the progression of functional disability were high baseline HAQ score, female sex, RF positivity, and lower education levels.
In this large observational study, we found a trend toward a reduction in the progression of radiographic damage in alcohol drinkers, most specifically in those with low to moderate consumption. The evolution of functional capacity revealed a similar trend, although it did not reach significance. We observed an unexpected effect modification by sex, with a highly significant reduction in radiographic progression in male drinkers. Overall, this suggests that in addition to its role in the development of RA, alcohol may also be involved in the evolution of the erosive process. To our knowledge, this is the first study to explore the effect of alcohol consumption on the progression of joint damage.
A recent study examined the effect of ethanol in type II collagen–induced arthritis (9). The development of an erosive arthritis was almost completely abolished in the exposed mice, without any sign of toxic effects on the liver. The antiinflammatory and antidestructive properties of ethanol were mediated by significant reductions in interleukin-6 (IL-6), macrophage inflammatory protein 1α, tumor necrosis factor α (TNFα), NF-κB nuclear translocation, down-regulation of leukocyte migration, and up-regulation of IL-10 production.
The chronic abuse of alcohol has been clearly shown to cause immunosuppression, as well as an increase in overall morbidity and mortality (22). Subjects who abuse alcohol show elevated levels of the proinflammatory cytokines TNFα, IL-1, and IL-6 (23). However, long-term intake of low to moderate amounts of alcohol has been shown to reduce levels of TNFα (24, 25), IL-1 (24), IL-6 (26, 27), C-reactive protein (CRP) (26, 28, 29), and NF-κB activation (30, 31) and to increase levels of the antiinflammatory cytokine IL-10 (23, 32, 33).
The protective effect of alcohol consumption on radiographic progression observed in men but not in women may be partially explained by sex differences in RA severity. A multinational cohort including >6,000 RA patients demonstrated more active disease (in terms of the DAS28), increased functional impairment (as measured by the HAQ score), less remission, and more erosion in female patients (34). Nevertheless, several other authors have demonstrated no sex differences in terms of radiographic outcomes (14, 35). Moderate alcohol consumption also appears to have a greater influence on systemic markers of inflammation, such as CRP and leukocyte count, in men than in women (29). Another hypothesis for the observed effect modification of alcohol by sex could be related to the alcohol dose. In general, men consume alcohol more regularly and in greater quantities than women (36). In this population, male drinkers were twice as likely to consume alcohol daily (27% of males compared with 14% of females), which is the category in which the beneficial effect of alcohol on radiographic progression was most pronounced.
Although the effect size is relatively small over a short timeframe, the clinical importance may well be of greater significance over the course of a lifetime in a chronic disease such as RA. Differences of this magnitude are probably clinically meaningful for men over the long term, but not necessarily for women. Given our findings, it would appear prudent to include alcohol consumption in future studies investigating radiographic progression in RA patients, particularly when analyzing the effect of sex.
Our study does have several limitations. Approximately one-third of the patients in the registry did not have sequential radiographic followup and were excluded. Our results could potentially be biased if missing radiographic followup was associated with both alcohol consumption and more severe radiographic progression. There are a number of reasons why we think that the results are unlikely to be due to selection bias. First, the predominant reason for missing followup radiographs was recent enrollment in the database with insufficient time for subsequent radiographs. Second, there are no data to suggest that missing radiographic followup is related to more severe disease progression. In fact, the excluded patients in this analysis had baseline characteristics that were very similar to those of the included patients, with no differences in terms of estimated baseline radiographic damage (P = 0.99), sex (P = 0.86), or RF status (P = 0.41), and only minor, clinically irrelevant differences in age (56.1 versus 54.4 years), DAS28 scores (4.31 versus 4.47), and HAQ scores (1.25 versus 1.11).
We performed a sensitivity analysis comparing the classification of occasional drinkers as drinkers or nondrinkers and consequently classified this subgroup as drinkers. Inclusion and followup in the SCQM database are unrelated to alcohol consumption. Enrollment in the database is determined by both physician and treatment choices. Enrolled patients tend to have more severe disease and receive more biologic therapies than RA patients in the general population (37). It is therefore highly unlikely that alcohol drinkers or nondrinkers were differently enrolled. However, one must keep in mind that these results relate to a single population of predominantly white patients in Switzerland and may not necessarily be directly extrapolated to other populations.
We used self-reported alcohol consumption status, which may be prone to underreporting and misclassification. While patients may have underreported their alcohol consumption to “please” their physicians, it is very unlikely that alcohol consumption was differentially misclassified by levels of disease severity. Recall bias is unlikely to have been a concern, since we targeted the overall pattern of alcohol consumption rather than the exact amount over a predefined period. Another potential problem is that RA patients may change their level or pattern alcohol consumption over time; however, only a few participants in our cohort changed their alcohol drinking status during followup. In a Scandinavian cohort, alcohol consumption by RA patients with a disease duration of <6 months did not differ from that reported by patients with a longer disease duration (6).
Another potential bias may arise in RA patients treated with potentially hepatotoxic DMARDs or nonsteroidal antiinflammatory drugs (NSAIDs), who may be advised to stop or reduce their alcohol consumption. While this is a theoretical concern, it has been demonstrated that no significant differences exist in alcohol consumption between RA patients taking either methotrexate (MTX) or NSAIDs and those who were not (6). In addition, in a UK study of 200 RA patients who were given clear advice to cease alcohol consumption prior to the introduction of a DMARD, the vast majority of patients treated with MTX continued to drink alcohol (38). In our study, we investigated the pattern of DMARD use between drinkers and nondrinkers and did not find any significant differences, which is consistent with the results of previous studies. However, the differences in transaminase levels with regard to varying levels of alcohol consumption have yet to be analyzed. Patients taking MTX who drank alcohol did not discontinue MTX more frequently than nondrinkers (P = 0.64 by Cox proportional hazards model), suggesting that moderate alcohol consumption did not result in significantly higher elevations of transaminase levels compared with nondrinkers who were taking MTX.
The effect of certain environmental risk factors, most specifically smoking, appears particularly strong in the setting of anti-CCP–positive RA. Unfortunately, anti-CCP status was available in only 24% of the patients in our study and could therefore be an unmeasured confounder. In this subset, the proportion of alcohol drinkers who were anti-CCP positive was very similar to the proportion of the nondrinkers who were anti-CCP positive (58.5% and 57.9% respectively, P = 0.881). Moreover, we found no evidence of effect modification by anti-CCP status (positive, negative, or missing).
One of the strengths of our analysis is the use of a large, prospective, population-based cohort of RA patients. Quantification of radiographic status was performed using well-validated and reproducible methods, with high intrarater and interrater correlations. The statistical power of our study was sufficient to detect small differences in rates of radiographic progression. With the use of a longitudinal regression analysis, we adjusted for a large number of potential confounders, such as the use of various therapeutic agents, RF, smoking, and socioeconomic status.
In conclusion, a trend toward reduced radiographic progression was observed in drinkers compared with nondrinkers, particularly in the occasional and daily consumers. Men who drank alcohol demonstrated a more pronounced reduction in radiographic progression compared with men who did not drink, an effect which was not seen in women. Although the effect size is relatively small over a short timeframe, the clinical importance may well be of greater significance over the course of a lifetime, particularly in male RA patients. While the careful surveillance of alcohol consumption, particularly in combination with drugs such as NSAIDs and DMARDs, remains an essential component of good management, perhaps the universal recommendation to cease low to moderate alcohol consumption may be questioned for selected RA patients. Further research is required to better understand the impact of alcohol consumption on RA in terms of the type of alcohol, drinking pattern, volume, drug interactions, and the ethnicity of the population.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Nissen 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.
Study conception and design. Nissen, Gabay, Finckh.
Acquisition of data. Nissen, Scherer.
Analysis and interpretation of data. Nissen, Finckh.
The authors are grateful to all of the physicians who contributed valuable time to the enrollment and followup of this group of patients in the SCQM, as well as to all of the RA patients in the database, without whom this type of study would not have been possible.