Objective.- This study was undertaken to confirm and extend our earlier observation that gender is a biological effect modifier of smoking-rheumatoid arthritis (RA) relationship in a diverse national survey sample in the United States.

Methods.- Smoking history of 644 cases of RA and 1509 geographically matched general population controls were compared using weighted logistic regression.

Results.- There were 644 respondents with RA (cases) and 1509 geographically matched controls. Cases were significantly younger, less educated, more likely to be single and female than controls. Among cases 57% were smokers while among controls 49% smoked. Among women, after adjusting for age, hysterectomy had an age adjusted odds ratio 1.45, (95% CI 0.99-2.10) and menopause an adjusted odds ratio 1.18 (95% CI 0.99-2.10) were associated with smoking. In univariable analysis ever-smoking was associated with increased risk of RA (odds ratio 1.34, 95% CI 1.0-1.81). Among the strata of smokers, there was an increasing gradient of risk with increasing exposure to smoking (P = 0.041). In separate multivariable models, smoking increased the risk in men (odds ratio 2.29, 95% CI 1.35-3.90) while in women the risk was not elevated (odds ratio 0.98, 95% CI 0.67-1.42). After adjusting for the statistically significant interaction both female gender (odds ratio 2.30, 95% CI 1.39-3.83) and having ever smoked (odds ratio 2.31, 95% CI 1.36-3.94) emerged as significant risk factors for RA.

Conclusions.- Gender interacts with smoking in by an unknown mechanism to lead to differential risk of RA.