To the Editor: When the many published reports on the efficacy of long-term anticoagulant therapy in preventing recurrences of myocardial infarction are studied, one cannot fail but to be impressed by a curious phenomenon: that as experimental design has become more and more rigorous, the documentation of the benefits derived from this particular form of therapy have become more and more difficult as the differences between control and treated patients tend to vanish Thus, in Bjerklund's study and in the British Medical Research Council Study, the probability value for successful therapy in males of the most favorable category (selected by age) calculates to be about 0.1 (at times, slightly less, but never at the 0.05 level). When control of observer bias is added as well to the experimental design, as was done by Seaman and by our group using the double-blind placebo technique, no benefit from anticoagulant therapy was demonstrated