Novel uses of coded data in clinical trials.

Abstract

6143 Background: The EU Clinical Trials Directive has had an adverse impact on academic clinical research by increasing regulatory burdens that fall to staff whose primary roles are not trial related. In particular, documenting of clinical events is costly and often incomplete. One way to improve compliance whilst decreasing costs would be to use automated National Health Service (NHS) coded data. We set out to assess completeness of reporting of skeletal related events (SREs) post-progression within a clinical trial in bony metastatic prostate cancer (Trapeze, ISRCTN12808747). As these occur outside of the trials clinic environment, we examined the viability of using coded data in place of conventional case report form (CRF) collection. Methods: Hospital notes were reviewed to determine the occurrence of SREs in 86 men (median age 69; range 49-84) enrolled in Trapeze. These figures were used as our gold standard. SREs were classified into: (a) radiotherapy (RT) to treat bone metastases and (b) other (hypercalcaemia, cord compression, fracture and change of cancer therapy). SRE incidence was also determined by review of CRFs, and interrogation of the RT & NHS coding databases. These were both compared to the gold standard and with each other. Results: Sixty SREs were detected by notes review: 39/60 (65%) comprised RT to treat bone metastases, whilst 21/60 (35%) comprised others. Trial CRFs missed 36/39 (92%) RT SREs and 13/21 (62%) non-RT SREs. The departmental RT database detected 39/39 (100%) of RT SREs, whilst the NHS coding database missed 18/21 (86%) non-RT SREs. No SREs were detected by coding that did not appear in the notes review. Conclusions: In this study, post progression CRFs were highly inaccurate and trial procedures are being modified. For bone RT as an SRE, coded data matched the gold standard and could safely be substituted for standard CRF collection. For non-RT SREs both trial CRFs and NHS coded data were inaccurate in comparison with notes review. Data captured in academic trials needs to be carefully targeted to ensure maximum accuracy and minimal impact on personnel not directly associated with the study. For selected events, traditional CRFs can be reliably replaced with automated data capture from NHS databases, and this practice should be encouraged and expanded. No significant financial relationships to disclose.