Session Information
Title: Epidemiology and Health Services Research VI: Risk Factors in Rheumatic Disease Susceptibility
Session Type: Abstract Submissions (ACR)
Background/Purpose: Family history (FH) of autoimmunity, genetics, and environmental factors have been associated with RA. The area under the receiver operating characteristic curve (AUC) can measure the ability of prediction models to discriminate between cases and controls. We aimed to evaluate the performance of prediction models for RA based on family history, genetics, and environmental factors.
Methods: We developed RA prediction models in a nested case-control study within Nurses’ Health Study (NHS) and replicated in women in the Epidemiological Investigation in RA (EIRA) study. NHS cases were validated by chart review and matched to controls by age, menopausal status and post-menopausal hormone use. EIRA new-onset RA cases were matched to controls by age and region. All cases were Caucasian and satisfied the 1987 ACR criteria for RA classification. FH data were obtained from questionnaires (NHS) and registries (EIRA). Serologic status was defined by +RF/ACPA for NHS and by +ACPA for EIRA. Weighted genetic risk scores were calculated for cases and controls based on 39 genetic markers associated with RA in prior studies. Logistic regression models were used to calculate the AUC and 95% confidence intervals (CI) for seropositive/ACPA+ RA and seronegative/ACPA- RA. Model components were based on family history (FH, first-degree relative with RA or lupus in NHS and RA in EIRA), environmental (E: age, smoking pack-years, body mass index, alcohol intake, education, and parity), genetic factors (G), and HLA shared epitope-smoking interaction (GEI). Analyses stratified by FH were performed using models based on E, G, and GEI components.
Results: We analyzed 492 cases and 501 controls in NHS women and 1,244 cases and 971 controls in EIRA women with FH data. The complete model (FH+E+G+GEI) for seropositive/ACPA+ RA had an AUC of 0.71 (95% CI 0.67-0.75) in NHS and 0.78 (95% CI 0.76-0.80) in EIRA (Table 1). For women with +FH, the complete model (E+G+GEI) had an AUC of 0.85 (95% CI 0.77-0.92) for seropositive RA in NHS and an AUC of 0.85 (95% CI 0.78-0.92) for ACPA+ RA in EIRA (Table 2). For seronegative/ACPA- RA with +FH, E+G+GEI models had an AUC of 0.85 (95% CI 0.63-0.72) in NHS and an AUC of 0.80 (95% CI 0.69-0.91) in EIRA.
Conclusion: We have developed and replicated prediction models for RA in women using family history, environment, and genetics. These models had higher discrimination for seropositive/ACPA+ RA and performed best in stratified analysis for those with family history. For women with family history of RA or lupus, environmental and genetic data may be able to predict the development of RA.
|
Table 1. Areas under the receiver operating characteristic curves (AUC) for RA prediction models using family history (FH), environment (E), genetics (G), and gene-environment interaction (GEI) in the Nurses’ Health Study (NHS) and the Epidemiologic Investigation of RA (EIRA) study. |
||||
|
|
NHS |
EIRA |
||
|
|
AUC (95% CI) |
AUC (95% CI) |
||
|
Models |
Seropositive RA |
Seronegative RA |
ACPA+ RA |
ACPA- RA |
|
FH |
0.58 (0.55-0.61) |
0.59 (0.56-0.62) |
0.53 (0.52-0.55) |
0.51 (0.50-0.52) |
|
E |
0.65 (0.61-0.69) |
0.58 (0.54-0.63) |
0.69 (0.67-0.72) |
0.65 (0.62-0.68) |
|
G |
0.59 (0.55-0.63) |
0.58 (0.54-0.63) |
0.72 (0.69-0.74) |
0.53 (0.50-0.56) |
|
FH+E |
0.69 (0.65-0.73) |
0.66 (0.61-0.70) |
0.71 (0.68-0.73) |
0.65 (0.62-0.68) |
|
FH+E+G |
0.70 (0.66-0.74) |
0.68 (0.63-0.72) |
0.78 (0.76-0.80) |
0.65 (0.63-0.69) |
|
FH+E+G+GEI |
0.71 (0.67-0.75) |
0.68 (0.63-0.72) |
0.78 (0.76-0.80) |
0.66 (0.63-0.69) |
|
E models: age, cigarette smoking pack-years, body mass index, alcohol intake, education, and parity |
||||
|
FH models: family history |
||||
|
G models: weighted genetic risk scores based on 39 RA associated markers |
||||
|
GEI models: HLA shared epitope-cigarette smoking interaction |
||||
|
Table 2. Areas under the receiver operating characteristic curves (AUC) for RA prediction models stratified by family history (FH) using environment (E), genetics (G), and gene-environment interaction (GEI) in the Nurses’ Health Study (NHS) and the Epidemiologic Investigation of RA (EIRA) study. |
||||
|
|
|
NHS |
|
|
|
|
|
AUC (95% CI) |
|
|
|
|
+FH |
-FH |
||
|
Models |
Seropositive RA |
Seronegative RA |
Seropositive RA |
Seronegative RA |
|
E |
0.82 (0.74-0.90) |
0.78 (0.68-0.87) |
0.67 (0.62-0.71) |
0.61 (0.55-0.66) |
|
G |
0.63 (0.52-0.75) |
0.64 (0.52-0.75) |
0.58 (0.54-0.63) |
0.58 (0.53-0.62) |
|
E+G |
0.84 (0.77-0.92) |
0.79 (0.70-0.88) |
0.68 (0.64-0.72) |
0.62 (0.57-0.67) |
|
E+G+GEI |
0.85 (0.77-0.92) |
0.85 (0.77-0.92) |
0.68 (0.64-0.73) |
0.63 (0.58-0.68) |
|
|
|
EIRA |
|
|
|
|
|
AUC (95% CI) |
|
|
|
|
+FH |
-FH |
||
|
Models |
ACPA+ RA |
ACPA- RA |
ACPA+ RA |
ACPA- RA |
|
E |
0.77 (0.68-0.86) |
0.78 (0.67-0.90) |
0.69 (0.67-0.72) |
0.65 (0.62-0.68) |
|
G |
0.74 (0.65-0.84) |
0.57 (0.44-0.71) |
0.71 (0.69-0.73) |
0.53 (0.50-0.56) |
|
E+G |
0.84 (0.76-0.91) |
0.80 (0.69-0.92) |
0.77 (0.75-0.79) |
0.65 (0.62-0.68) |
|
E+G+GEI |
0.85 (0.78-0.92) |
0.80 (0.69-0.91) |
0.77 (0.75-0.79) |
0.66 (0.63-0.69) |
|
E models: age, cigarette smoking pack-years, body mass index, alcohol intake, education, and parity |
||||
|
G models: weighted genetic risk scores based on 39 RA associated markers |
||||
|
GEI models: HLA shared epitope-cigarette smoking interaction |
||||
Disclosure:
J. A. Sparks,
None;
C. Y. Chen,
None;
X. Jiang,
None;
L. T. Hiraki,
None;
L. Klareskog,
No own commercial interests,
2;
L. Alfredsson,
None;
K. H. Costenbader,
None;
E. W. Karlson,
None.
« Back to 2013 ACR/ARHP Annual Meeting
ACR Meeting Abstracts - https://acrabstracts.org/abstract/performance-of-prediction-models-for-rheumatoid-arthritis-serologic-phenotypes-among-women-using-family-history-genetics-and-environmental-factors/