Geographic and network analysis of oncology trials: Portfolio assessment of ClinicalTrials.gov.
6047 Background: Increasing complexity and specificity of cancer trials (CT) necessitates collaboration across the CT network to optimize research efforts. The relationship among CT sites and networks provide unique insights into improving coordination and accrual. Methods: 5971 interventional CTs registered between 2007 and 2010 were aggregated by trial and location. CTs in the Aggregate Analysis dataset from ClinicalTrials.gov (AACT) were identified using MeSH terms. The distribution of mid-phase (MP)(phase I/II,II) and late-phase (LP)(PII/III,PIII) CTs for the ten most represented cancer types by number of sites was assessed using network graph theory and geographic analysis comparing distribution of trials across metropolitan statistical areas. Results: 66,566 CT sites were identified across the sample, 59.6% were in the United States (MP: 50.2%; LP: 42.9%). Geographical availability of CTs and local cancer incidence rates were highly correlated (0.797, p≤ 0.001) but varied depending upon disease type. Network density (the degree of dyadic interconnections between sites studying similar cancer types) showed overall that MP trials were less dense with sparser interconnections among sites than LP trials. Network density of LP trials was higher for cancer types that had poorer correlation between geographic distribution of incidence and enrolling sites (-0.777, p=0.008). MP trials did not show a similar trend. Conclusions: The relationship between the distribution of CT and site location can be envisaged through geographic and network analysis of CT registry data. LP high-density networks should strive to diversify trial locations to better meet regional incidence rates. [Table: see text]