Coherent coding and classification schemes are pivotal to the avoidance of the duplication of effort, the indexing of registers, such as CCTR or Central and the integration of medical evidence into the clinical environment. The semantics of clinical trial structures has already been elicited, however the choice of source terminologies to populate these types remains an open question. Three issues are at stake: the choice between enumerative and combinatorial systems, the conceptual cover of the existing terminologies and the adequation of 'grammars' for the combination of terms. We addressed the second of these questions - the conceptual cover of the existing terminologies.

Methods: Entries from small set of clinical trial summaries entered in TriSum, without prior knowledge of the reviewer, were hand coded into Medra, Snomed, UMLS and Read. The adequation of the coding to the source was then judged, specifically in terms of the availability of terms to represent the entries.

Results/Discussion: Preliminary results suggest that no system is capable of coding over 70% of entries in TriSum. Moreover the lack of clear rules for the combination of concepts is brought clearly to light.