Abstract Introduction PTPN22, PADI-4, and CTLA-4 have been associated with risk for rheumatoid arthritis (RA). We investigated whether polymorphisms in these genes were associated with RA in Caucasian women included in two large prospective cohorts, adjusting for confounding factors and testing for interactions with smoking. Methods We studied RA risk associated with PTPN22 (rs2476601), PADI-4 (rs2240340), and CTLA-4 (rs3087243) in the Nurses' Health Study (NHS) and NHSII. Participants in NHS were aged 30 to 55 years at entry in 1976; those in NHSII were aged 25 to 42 years at entry in 1989. We confirmed incident RA cases through to 2002 in NHS and to 2003 in NHSII by questionnaire and medical record review. We excluded reports not confirmed as RA. In a nested case-control design involving participants for whom there were samples for genetic analyses (45% of NHS and 25% of NHSII), each incident RA case was matched to a participant without RA by year of birth, menopausal status, and postmenopausal hormone use. Genotyping was performed using Taqman single nucleotide polymorphism allelic discrimination on the ABI 7900 HT (Applied Biosystems, 850 Lincoln Centre Drive, Foster City, CA 94404 USA) with published primers. Human leukocyte antigen shared epitope (HLA-SE) genotyping was performed at high resolution. We employed conditional logistic regression analyses, adjusting for smoking and reproductive factors. We tested for additive and multiplicative interactions between each genotype and smoking. Results A total of 437 incident RA cases were matched to healthy female control individuals. Mean (± standard deviation) age at RA diagnosis was 55 (± 10), 57% of RA cases were rheumatoid factor (RF) positive, and 31% had radiographic erosions at diagnosis. PTPN22 was associated with increased RA risk (pooled odds ratio in multivariable dominant model = 1.46, 95% confidence interval [CI] = 1.02 to 2.08). The risk was stronger for RF-positive than for RF-negative RA. A significant multiplicative interaction between PTPN22 and smoking for more than 10 pack-years was observed (P = 0.04). CTLA-4 and PADI-4 genotypes were not associated with RA risk in the pooled results (pooled odds ratios in multivariable dominant models: 1.27 [95% CI = 0.88 to 1.84] for CTLA-4 and 1.04 [95% CI = 0.77 to 1.40] for PADI-4). No gene-gene interaction was observed between PTPN22 and HLA-SE. Conclusion After adjusting for smoking and reproductive factors, PTPN22 was associated with RA risk among Caucasian women in these cohorts. We found both additive and multiplicative interactions between PTPN22 and heavy cigarette smoking.