[The minimization method in a medium size clinical trial. Study of immunochemotherapy with PSK in gastric cancer]

One of the objectives of randomization is to balance the number of patients in each level and combination of levels of prognostic factors. However, as the number of possible prognostic factors increases, some imbalance may occur and make the results of the analysis less interpretable. In order to balance the marginal distribution of each factor and to facilitate more interpretable trial results, the "minimization method" reported by Pocock was applied in our multi-center immunochemotherapy clinical trial. Balancing in every institution was achieved by using Zelen's prespecified value d = 3 for replacing provisional assignment in the individual institution. Nonmetric principal component analysis in S, N factors which were used in minimization process showed good equilibration between groups. Also, in uncontrolled background factors, such as cancer pathology or Borrmann types, acceptable balance was obtained spontaneously at the end of the trial.