Abstract
Objective. While cigarette smoking is the best-studied environmental factor contributing to rheumatoid arthritis (RA), no study to date has examined the influence of smoking cessation on disease activity. We examined this relationship in an observational cohort of patients with RA in the United States.
Methods. Patients enrolled in the Consortium of Rheumatology Researchers of North America registry (CORRONA) were stratified into never, former, and current smokers at enrollment. Current smokers were further stratified into continued and ceased smoking groups during their followup in the registry. The primary outcome was change in Clinical Disease Activity Index (CDAI) at last visit in a multivariate, random-effects regression model accounting for multiple timepoints.
Results. At last visit, there was no significant change in CDAI between ceased smokers and continued smokers (coefficient −0.00091, SE 0.0033, p = 0.7834). The study did confirm prior cross-sectional studies that current smokers have worse disease activity than former or never smokers.
Conclusion. In the short term, smoking cessation did not appear to influence change in disease activity over time.
Rheumatoid arthritis (RA) is a chronic inflammatory disease of multifactorial etiology. There is evidence of a role for both genetic factors and environmental factors in causation1,2,3. Cigarette smoking is the best-studied environmental factor contributing to RA. Recent studies demonstrated positive relationships between smoking, seropositive RA, and disease severity4. Additionally, smoking is related to rheumatoid factor (RF) and anticitrullinated protein antibodies (ACPA) isotype and titer4.
To our knowledge, to date, no studies have examined the effect of smoking cessation on RA disease activity. We examined this relationship in an observational cohort of patients with RA in the United States.
MATERIALS AND METHODS
Population and cohort definitions
This analysis used data from patients enrolled in the Consortium of Rheumatology Researchers of North America (CORRONA) registry. Full details of the CORRONA registry have been reported5. In brief, patients with a diagnosis of RA, either new-onset or longstanding, are enrolled at an initial study visit. Clinical history, including duration of symptoms and date of diagnosis, is recorded, in addition to demographic information and RA medication history. A tender and swollen joint count is performed by a clinician. Patients are then seen every 3 months, although some patients may go longer between visits if deemed appropriate by their treating clinician. All data are collected by patient or physician report. Pertinent clinical history, tender and swollen joint counts, patient and physician global assessment, presence of erosions as reported by the treating rheumatologist, erythrocyte sedimentation rate (ESR), and C-reactive protein (CRP) are recorded at each visit, as are changes in RA medications. At each visit, patients are also asked if they are current, former, or never-smokers.
The study population included individual patients with smoking status information. Patients were stratified into never, former, and current smokers at their initial visit. Patients who reported current smoking at enrollment in the registry were further stratified into continued and ceased smoking groups during their followup in the registry. The study was approved by the institutional review boards (IRB) of participating academic sites, and a central IRB for community-based private sites.
Definition of smoking cessation
“Ceased smokers” were defined as patients who had reported current smoking who subsequently reported cessation of smoking on at least their last 2 consecutive CORRONA evaluations. A patient who noted smoking cessation only at their last followup visit was defined as a continued smoker, as many patients who attempt to quit smoking will frequently resume smoking shortly afterward. Therefore, persistence of smoking cessation was required to define a patient as a “ceased smoker.” In addition, patients who initially reported active smoking, then had multiple consecutive ceased smoking visits, then again reported active smoking, were classified as active smokers. Only patients who had continued cessation for multiple consecutive visits and including the last 2 followup visits were defined as ceased smokers. As the “ceased-smoker” definition required at least 3 CORRONA visits, the comparison group, “continued smokers,” also required at least 3 visits. Patients from both groups were required to not be in remission as determined by Clinical Disease Activity Index (CDAI) at their index visit.
Matching
In order to control for duration of followup within CORRONA, ceased smokers were matched with continued smokers based on the total number of followup visits. All patients who noted active smoking at enrollment in CORRONA were analyzed; all continued smokers were matched with all ceased smokers. For most patients, study visits occurred at 3-month intervals.
Index visit
The index visit for ceased smokers was defined as the last visit while smoking status was reported as positive prior to the first reported cessation visit. Therefore, all patients were active smokers during the index visit. For example, if a patient had 5 CORRONA visits, and reported active smoking at visits 1 and 2, and smoking cessation at visits 3 through 5, the index visit was defined as visit 2. The index visit for continued smokers was determined by matching continued smokers with ceased smokers for total number of study visits. For example, if a continued smoker was matched with a ceased smoker where both had 5 visits and the index visit for the ceased smoker was the third visit, then the third visit for the continued smoker was used as the index visit.
Study outcomes
The primary outcome for the longitudinal analysis was change in CDAI score over time from the index visit. CDAI is a composite measure of RA disease activity consisting of the 28 tender joint count, 28 swollen joint count, patient global health score visual acuity scale, and physician global health visual acuity scale. The CDAI has been shown to correlate highly with the Disease Activity Score 28-joint count (DAS28)6. As the CDAI values had a skewed distribution, for the regression analysis they were transformed to square-root CDAI, which had a more normal distribution. The outcome variable is therefore the interaction between smoking cessation status and time.
Analysis
Chi-square statistics were performed for categorical variables, and analysis of variance for continuous variables. Index and last visit variables were assessed and controlled for in multivariate models, including age, sex, race/ethnicity, RF status, duration of disease, duration of followup, medication use [methotrexate (MTX) and prednisone usage and dose, other disease-modifying antirheumatic drug (DMARD) and biologic DMARD use]. Individual components of the CDAI were independently assessed, as were the modified Health Assessment Questionnaire, patient pain assessment, ESR, and CRP (when available). Assessment of change in CDAI over time was performed using the continued and ceased-smoker populations with a random-effects regression model, controlling for age, sex, RF status, race/ethnicity, disease duration, and medication use. The random-effects regression model assesses multiple covariates at each timepoint in followup, and calculates a coefficient for each. CDAI scores were transformed for this analysis to square-root CDAI to permit linear modeling. A sensitivity analysis random-effects regression model limiting followup to 2 years was also performed, as the number of patients at each timepoint beyond 24 months became significantly smaller, and could have had a disproportionate effect on the results. All statistics were analyzed using SAS version 9.1.
RESULTS
Figure 1 shows smoking status at enrollment of all patients in CORRONA at the time of this analysis, as well as how patients were stratified into continued and ceased smokers.
Table 1 shows baseline characteristics of patients with RA included in the study stratified by smoking status. There were significant differences between never, former, and current smokers in sex, race/ethnicity, and mean age. Former and current smokers were more likely to be RF-positive than never-smokers; current smokers were more likely to be RF-positive than former smokers. Never and former smokers were more likely to be in remission or have low disease activity as measured by CDAI than current smokers (never-smokers, 17.2% and 31.0%, respectively; former smokers, 18.6% and 32.3%; current smokers, 13.1% and 25.8%; p < 0.0001). Mean CDAI and DAS28 scores at baseline were lowest among former smokers [13.5 (SD 12.9) and 3.5 (SD 1.6), respectively] and never-smokers had lower scores [14.2 (SD 12.8) and 3.7 (SD 1.6)] compared with current smokers [16.8 (SD 13.9) and 3.9 (SD 1.6)].
Table 2 shows the subgroup of current smokers, stratified into continued versus ceased smokers, with characteristics at index visit (while all were still smoking) and last visit. Of the 2328 current smokers at enrollment in CORRONA, 1258 had both 3 or more visits and followup data on smoking status and were included in this analysis. There were significant differences at index visit in age, disease duration, and duration of followup within CORRONA between the 925 continued smokers and the 333 ceased smokers. There were no significant differences in sex, race/ethnicity, RF status, CRP, or ESR between the continued smokers and ceased smokers.
Disease activity at the index visit was significantly different between the 2 groups, measured by the CDAI [ceased, 12.9 (SD 12.1) vs continued, 16.5 (SD 13.1); p < 0.0001]. There were no differences at index visit in use of MTX, other traditional DMARD, prednisone or biologic DMARD, MTX or prednisone dose, history of erosions, or subcutaneous nodules (data not shown).
At last visit, mean CDAI scores were lower for ceased than for continued smokers [11.5 (SD 10.9) vs 14.3 (SD 12.6), respectively; p = 0.0002]. There was no significant difference in CRP levels or ESR.
Table 3 shows a random-effects regression model assessing change in CDAI score over time. In order to be included in the random effects regression model, all data at all data points had to be present; patients with missing data were excluded. Of the 1258 patients in the unadjusted analysis, 53 patients were eliminated due to missing data. The primary outcome variable, “effect of smoking on change in square-root CDAI over time in months,” showed that over time, there was no significant difference in the square-root CDAI between patients who stopped smoking compared to patients who continued smoking (coefficient −0.00091, SE 0.0033, p = 0.7834). Baseline disease activity adjusted for year of baseline visit was significantly lower in current smokers who ceased smoking (coefficient −0.345, SE 0.104, p = 0.0009). CDAI did not change significantly over time (coefficient −2.761, SE 2.591, p = 0.2866). The influence of smoking cessation was also not significant (coefficient −0.00091, SE 0.0033, p = 0.7834). Variables that significantly affected change in CDAI included biologic DMARD use (coefficient +0.569, SE 0.0849, p < 0.0001), use of MTX only (coefficient +0.255, SE 0.0928, p = 0.0064), year of index visit (coefficient −0.066, SE 0.0318, p = 0.0382), using only a traditional DMARD (coefficient +0.471, SE 0.134, p = 0.0008), prednisone use (coefficient −0.237, SE 0.066, p = 0.0004), and disease duration (coefficient −0.0207, SE 0.00446, p < 0.0001).
A sensitivity analysis limiting followup to 24 months, to eliminate the effect that few patients with greater followup could have, showed no difference from the primary model (coefficient −0.0045, SE 0.006, p = 0.4641).
DISCUSSION
This is the first study to examine the relationship between smoking cessation and change in disease activity in patients with RA enrolled in a longitudinal observational registry. In our study, smoking cessation did not appear to significantly influence disease activity over time. This analysis did confirm prior cross-sectional studies that current smokers have greater disease activity and are less likely to be in states of remission or low disease activity than former or never-smokers.
Several studies have examined relationships among smoking and disease activity and severity; most had cross-sectional design and were performed in Europe7,8,9,10,11. These studies found smoking was associated with nodulosis and lower grip strength, and was variably related to disease activity, swollen joint count, radiographic erosions, ESR, and functional disability. North American studies found similar results12,13,14.
We are unaware of any other study that reported on a cohort of patients with RA who are current smokers and explored differences between those who would quit smoking versus those who continued. There were some demographic differences between these groups; why these differences were associated with subsequent smoking cessation is unclear.
The most clinically important finding was the absence of the effect of smoking cessation on change in CDAI over time in longitudinal analysis. A strength of this model was the adjustment for multiple variables at multiple timepoints, including baseline disease activity, that varied based on year of enrollment and smoking status. A surprising finding was the negative effect of use of biologics on CDAI values. However, this likely reflects that our study design did not measure change from predrug baseline assessments, but rather indicated prevalent use. Of note, recent studies reported that smoking was associated with decreased response to anti-tumor necrosis factor-α therapy and MTX15,16,17.
There are several possible reasons why our study failed to reveal a relationship between smoking cessation and improvement in disease activity. One possibility is that this is a true finding, and that smoking cessation does not influence disease activity; once smoking has helped initiate the disease process, it may not play a role in the perpetuation of the disease. Of the 2328 patients who were current smokers at baseline, 1070 were not included in the analysis as they did not have sufficient visits within CORRONA. Had there been sufficient data on these patients, it is possible the results would have shown a benefit for smoking cessation. Since previous reports noted that the effect of smoking is greatest among patients who are ACPA-positive, another possibility is that an analysis stratified by ACPA status might show different results. The CORRONA dataset began before ACPA antibodies were routinely assessed in clinical practice, so for many patients, ACPA status is unavailable. We look forward to performing that analysis in the future.
Additionally, smoking status in CORRONA is recorded as yes/no. If the patients who continued smoking had a smaller pack-year history than those who ceased, this might mitigate the difference in disease activity between the 2 groups. The overall duration of smoking cessation may not have been long enough to reveal an effect on RA disease activity. Finally, there was a fundamental difference in baseline disease activity between those who would cease smoking and those who continued smoking, comparing the last visit prior to smoking cessation. There may be fundamental differences between these 2 groups that go beyond sex, race, or RF status, but do relate to age, disease duration, duration of followup within CORRONA, or some other unassessed variable that may explain the differences. There also may be other reasons why patients decide to quit smoking, including other health issues. These might include development of comorbid medical problems such as coronary artery disease, diabetes, hyperlipidemia, or hypertension, or development of greater concern for overall health. This concern could lead to other changes such as exercise, which might be associated with improvement in disease activity.
Strengths of our study include the prospective longitudinal design and the cohort size of the CORRONA registry. There are few other RA cohorts as large that include both smoking status and physician-derived outcome measures, including tender joint count and swollen joint count. Our study is appreciably larger than other investigations of the effects of smoking on RA outcomes.
Among the limitations, the observational design of the registry, like other registries, makes residual confounding a study limitation. Second, these analyses are limited by absence of smoking duration and pack-years. That information was not recorded in earlier versions of the registry forms, and therefore was not available for analysis. Finally, some studies suggest smoking status is most germane among patients with the shared epitope and ACPA18. As ACPA status was not routinely measured in clinical practice until recently, it is not available for many patients enrolled in CORRONA prior to 2006.
This is the first study assessing the relationship between smoking cessation and RA disease activity. Smoking cessation does not appear to influence change in disease activity over time.
Footnotes
-
Dr. Greenberg is supported by research grants from the US National Institutes of Health (NIH; K23AR054412), the Arthritis Foundation, and Bristol Myers Squibb and has served on advisory boards for BMS, Centocor, Genentech, Roche, and UCB. Research support was provided to Dr. Kremer by Amgen, Abbott, Centocor, BMS, Roche, and Genentech, and honoraria from Abbott, Centocor, BMS, Roche, and Genentech. Dr. Hochberg served as a consultant for Amgen, Bristol Meyers Squibb, Roche, and UCB. CORRONA has received support from Abbott, Amgen, BMS, Centocor, Genentech, Lilly, and Roche.
- Accepted for publication December 14, 2011.