Arthritis & Rheumatism, Volume 62,
November 2010 Abstract Supplement

Abstracts of the American College of
Rheumatology/Association of Rheumatology Health Professionals
Annual Scientific Meeting
Atlanta, Georgia November 6-11, 2010.


Modeling Smoking in Systemic Sclerosis: A Comparison of Different Statistical Approaches.

Hudson3,  Marie, Lo2,  Ernest, Steele3,  Russell, Baron1,  Murray

Jewish General Hospital, Montreal, QC, Canada
Lady Davis Institute, Jewish General Hospital
McGill University

Background:

Vascular disease is ubiquitous in SSc and cigarette smoking is well known to contribute to vascular disease. There is no standard way of modelling smoking in epidemiologic studies. The purpose of this study was to demonstrate the impact of different methods of modelling smoking when determining vascular outcomes in SSc.

Methods:

We undertook a study using data from the Canadian Scleroderma Research Group Registry. Patients self-reported their smoking history, including intensity, duration and time since cessation. Vascular outcomes were severity of Raynaud's phenomenon, presence of finger ulcers and severity of finger ulcers. Five regression models using various definitions of smoking were developed: Model 1 compared ever to never smoking; Model 2 compared current and past smoking to never smoking; Model 3 modeled smoking using polynomial contrasts, treating never, past and current smoking as ordered values of a single variable; Model 4 modeled smoking status as current versus non-current; and, Model 5 represented smoking using the Comprehensive Smoking Index, which integrates smoking intensity, duration and time since cessation into a single covariate (Leffondre K et al. 2006 Stat Med 25:4132–46). All regression models were adjusted for age, sex, ethnicity, disease duration and limited or diffuse skin involvement.

Results:

This study included 606 SSc patients, of which 87% were women, 90% were white, mean age was 55 (± 12) years, mean disease duration was 11 (± 9) years, and 36% had diffuse disease. Of these, 16% were current, 42% past and 42% never smokers. Current and past smokers smoked a mean of 25 (±17) and 17 (±18) pack-years, respectively. Smoking duration varied from 1 to 60 years, with past smokers having a shorter duration than current smokers (18.3 vs. 31.7 years, respectively). Past smokers stopped smoking approximately 16 (±12) years prior to their baseline registry visit, although this varied from 1 to 50 years. The results of the regression analyses for the five models of smoking are summarized in Table 1.

Model 1 did not detect any significant effect of smoking on the vascular outcomes. Models 2 and 3 showed a 'U-shaped' trend in the effect of smoking, with past smokers appearing healthier than both never and current smokers, with this effect reaching statistical significance in Model 3 only. Model 4 detected a negative effect of smoking on the severity of vascular outcomes, reaching statistical significance for the severity of Raynaud's and showing a strong trend for the presence of finger ulcers. The results of Model 5 were similar to those of Model 4, but with much stronger statistical significance.

Conclusions:

These findings highlight the importance of proper modeling of smoking. Simple models may mask important effects of smoking on vascular outcomes in SSc. The CSI was the most sensitive model because it accounts for the wide range of smoking exposure.

Table 1. Summary of the regression results of the 5 different models of smoking

Model 1: Ever vs. Never Smoking      
Outcome Variableb-EVER(a)p-value    
Severity of Raynaud's (0–10)0.0070.92    
Presence of finger ulcers-0.030.88    
Severity of finger ulcers (0–10)-0.010.69    
Model 2: Current vs. Never and Past vs. Never      
Outcome Variableb-CURRENT(b)p-value b-PAST(c)p-value 
Severity of Raynaud's (0–10)0.200.05.-0.060.41 
Presence of finger ulcers0.400.15 -0.180.36 
Severity of finger ulcers (0–10)0.040.35 -0.030.33 
Model 3: Polynomial Contrasts      
Outcome Variableb-LINEAR(d)p-value b-QUAD(e)p-value 
Severity of Raynaud's (0–10)1.140.18 1.850.03*
Presence of finger ulcers2.190.35 4.410.05*
Severity of finger ulcers (0–10)0.190.65 0.660.10*
Model 4: Current vs. Non-current Smoking      
Outcome Variableb-CURRENT(f)p-value      
Severity of Raynaud's (0–10)0.230.01*   
Presence of finger ulcers0.500.06.    
Severity of finger ulcers (0–10)0.060.16    
Model 5: Comprehensive Smoking Index (CSI)      
Outcome Variableb-CSI(g)p-value    
Severity of Raynaud's (0–10)0.520.002**   
Presence of finger ulcers1.210.002**   
Severity of finger ulcers (0–10)0.120.07*   
. p < 0.1,
*p < 0.05,
**p < 0.01
(a)Estimated regression coefficient for Ever vs. Never smokers;
(b)Estimated regression coefficient for Current vs. Never smokers;
(c)Estimated regression coefficient for Past vs. Never smokers;
(d)Estimated regression coefficient for the linear contrast covariate;
(e)Estimated regression coefficient for the quadratic contrast covariate;
(f)Estimated regression coefficient for Current vs. Non-current smokers;
(g)Estimated regression coefficient for the CSI

To cite this abstract, please use the following information:
Hudson, Marie, Lo, Ernest, Steele, Russell, Baron, Murray; Modeling Smoking in Systemic Sclerosis: A Comparison of Different Statistical Approaches. [abstract]. Arthritis Rheum 2010;62 Suppl 10 :577
DOI: 10.1002/art.28346

Abstract Supplement

Meeting Menu

2010 ACR/ARHP