Pancreatic cancer: Survival in clinical trials versus the real world.
216 Background: Physicians often use clinical trial data to estimate expected survival for their patients, however clinical trial results may not generalize to a broader population. Given the need for accurate prognostication, we examined whether the survival of clinical trial pancreatic adenocarcinoma patients was similar to “real world” patients. Methods: We conducted a literature search using Pubmed to identify pancreatic adenocarcinoma Phase III trials published between 2005 and 2012. We excluded secondary or pooled analyses, trials with second-line treatments, and non-randomized studies. We compared clinical trial patients to a cohort of patients from the Surveillance, Epidemiology, and End Results (SEER) program, matching for diagnosis year, age and stage of disease. We evaluated for differences in median survival, 1-year survival, and 2-year survival. Results: We identified 27 trials (3 metastatic, 16 mixed metastatic and locally advanced, 3 locally advanced, and 5 resectable) that fit our search criteria, consisting of 8,438 patients. Compared to SEER the median survival was higher in 98% of the individual clinical trial arms by an average of 3.7 months (p = 0.001). The 1-year overall survival rate was higher in 100% of the clinical trial arms by an average of 12.3% (p <.0001). The 2-year overall survival was higher in 76% of the clinical trial arms by an average of 5.1% (p <.0001). The survival differences between SEER and clinical trials were greatest for patients with metastatic pancreatic cancer. Conclusions: We found that clinical trial patients have profoundly improved survival compared to patients in the SEER database, suggesting that clinical trial survival estimates may not generalize to “real world” patients with pancreatic cancer. Patient performance status, underlying comorbidity, and differences in treatment could not be assessed in SEER, though these factors likely explain a portion of the survival differences. These findings suggest that physicians should use caution when extrapolating survival estimates from clinical trials to the real world.