Objectives : The purpose of this study was to develop, implement, and assess an estimation procedure for preventing adverse drug reaction by subjective symptoms (complaints) of patients. This time, we carried out this study focusing on drug-induced leucopenia. Methods : We have built a database for CARPIS(Case Reports of Adverse Drug Reaction and Poisoning Information System) since 1987, and the total number of case reports of adverse drug reaction accumulated in the CARPIS database amounts to about 16000. Using the date obtained from 139 cases of drug-induced leucopenia accumulated in the CARPIS database, we investigated a relationship between leucopenia and patients. The evaluation scores were created based on the subjective symptoms and backgrounds of the patients. Results : We could estimate 91 case (65.5%) in 139 cases to be drug-induced leucopenia by the use of these evaluation scores. The validity of this evaluation scores was estimated to be as follows; sensitivity=65.5%, specificity=80.0% and predictive value of positive test(PVP)=82.0%. The positive likelihood(LR) ratio was 3.3 and negative likelihood ratio 0.43. Conclusions : In this study, PVP and LR values were not good, because among the symptoms of leucopenia a very few specific symptoms could be detected. But we reported previous by the evaluation scores about drug-induced liver disorders. Therefore, in order to apply these evaluation scores to the clinical practice, we prepared an evaluation form for subjective symptoms and backgrounds of the patients with adverse drug reactions. In the future, we plan to examine other adverse reactions which adds the data to this form.