A randomised trial of adjuvant chemotherapy vs placebo in operable stomach cancer recruited 249 patients from the West Midlands Region between 1976-1980. A Cancer Registry survey identified a further 1261 suitable concurrent cases. Trial patients were compared with the 960 non-trial cases from participating Districts. Only 493 (51%) non-trial cases passed all of the prospective trial selection criteria for entry. Stage and fitness caused the majority of exclusions and were also highly prognostic. A univariate analysis comparing eligible patients with the trial showed the two groups to be balanced for the significant independent prognostic factors of the trial. However, differences in patient age and the surgery performed indicate that recruitment may have been influenced by unknown selection factors. This survey highlights the difficulty of retrospective selection and confirms the need for randomised controls. Data available from specialist Registries may be used to help develop new protocols and to verify and extend trial results.