S (ACE) AEP Vol. 20, No. 9 September 2010: 691–724 698 data and CHIS self-report data. PAMF rates were sex and age standardized to the CHIS population. RESULTS: PAMF and CHIS rates were similar for men (77% vs. 79%) and women (76 vs. 78%) and those over (84% vs. 86%) and under (73% vs. 75%) 65 years of age. In nearly every racial/ethnic subgroup, PAMF and CHIS rates were comparable and not statistically significantly different. CONCLUSION: CRC screening rates based on clinical records in a geographically limited healthcare organization were comparable to those obtained from a statewide selfreport survey inCalifornia. This suggests that self-report bias in CHIS may be minimal with regard to CRC everscreening in sex-, age-, and racial/ethnic subgroups. It also suggests that PAMF screening rates reflect screening rates in the general population in California. P21 SURVIVAL AND PROGNOSTIC FACTORS OF PANCREATIC CANCER PATIENTS DIAGNOSED FROM 1995–1999 IN THE SAN FRANCISCO BAY AREA: A POPULATIONBASED STUDY WITH ACTIVE FOLLOW-UP Z Gong, EA Holly, PM Bracci, Department of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, CA PURPOSE: Pancreatic cancer is one of the most fatal cancers in the U.S with a 5-year relative survival rate !5% based on cancer registry data. For a more complete assessment of pancreatic cancer survival in the population, we conducted active patient follow-up in our large populationbased study of pancreatic cancer to compute overall survival rates and assess survival associated with demographic and prognostic clinical factors. METHODS: Active follow-up was used to determine vital status and survival for 1,961 pancreatic cancer patients identified as part of our population-based case-control study. Survival rates were compared across subgroups of patients using Kaplan-Meier methods. Multivariable Cox proportional hazards models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for survival associated with demographics, tumor characteristics and treatment. RESULTS: The overall 5-year relative survival rate of the 1,961 patients was 1.7% with a median survival of 3.9 months. The 5-year survival rate was greater in patients who were younger (!50 years: 9.6%), diagnosed with localized disease (9.4%), had well-differentiated tumors (6.4%), and had had surgical resection (10.9%). Shorter survival was associated with older age at diagnosis (HR for age 80+ yrs: 2.1; 95% CI, 1.6–2.6), male sex (HR, 1.1; 95% CI, 1.0–1.2), distant/metastatic disease at diagnosis (HR, 2.0; 95% CI, 1.6–2.5), and poorly differentiated tumors (HR, 1.5; 95% CI, 1.2–1.9). Longer survival was observed for Asian/Pacific Islanders (relative to non-Hispanic whites) (HR, 0.75; 95% CI, 0.65–0.87) and patients who received any active treatment regardless of tumor stage at diagnosis (HR ranged: 0.40–0.80). CONCLUSIONS: Using active follow-up methods resulting in !1% of patients lost to follow-up, results from our large population-based study show that an overall survival rate for pancreatic cancer patients is lower than that reported in cancer registry data. RESULTS: suggest that pancreatic cancer survival rates have changed little over time, and that early detection and improved treatment strategies are needed to improve prognosis for this deadly disease. P22 INNATE IMMUNITY GENES AND RISK OF ACUTE LYMPHOBLASTIC LEUKEMIA K Bartley, L-I Hsu, A Chokkalingam, J Wiemels, L Barcellos, C Metayer, P Buffler, School of Public Health, University of California, Berkeley, Berkeley, CA PURPOSE: Childhood acute lymphoblastic leukemia (ALL) may be impacted by abnormal responses to microbial challenges due to poor immune priming in childhood. The purpose of this analysis is to determine the association of ALL with single nucleotide polymorphisms (SNPs) in innate immunity genes, which may play a role in immune development. METHODS: Among 377 childhood ALL cases and 448 controls, single SNP statistical analyses were conducted using unconditional logistic regression, assuming a logadditive inheritance model, for 27 genes. Multi SNP analyses were conducted using haplotype sliding windows, with blocks of 2–5 SNPs. Gene-gene interactions were assessed using logic regression and cross validation methods. All models were adjusted for sex, age, race and Hispanic status. RESULTS: Among the 143 SNPs, 6 individual SNPs in 5 genes (CCL8, CX3CR1, CXCL12, FASLG, ITGB1) were significantly associated with childhood ALL (p ! 0.05), but did not remain significant after adjustment for multiple testing. Haplotype analyses revealed regions of interest near the 3’ end of both CXCL12 and CX3CR1; however, no statistically significant gene–gene interactions were identified using logic regression methods. CONCLUSIONS:While the results of this analysis do not generally support a role for innate immunity genes in the development of ALL, regions of interest identified in the CXCL12 and CX3CR1 genes should be further explored to identify specific functional SNPs.