Results
A total of 90 ADR involving 64 patients were included in the analyses. Patients presented a median age of 63 years (IQR: 51.0-70.8) and were predominantly male (51.6%). Descriptive data and agreement analyses of ADR were presented in Table 1 . Twelve different types of clinical manifestations were found, resulting from 30 suspected drugs to cause ADR. Most patients presented only one ADR (n=45; 70.3%). Thirteen patients presented two ADR (20.3%) and five patients presented three ADR (7.8%) during the follow up. Maximal number of ADR per patient was four, identified in only one patient (1.6%). Mean ADR duration was three days, and 13 ADR (14.4%) lasted for more than five days.
A total of 270 assessments were performed using the WHO-UMC system, as a result of the evaluation of 90 ADR by the three judges involved in the adjudication process. According to global (majority) classification, 47 (52.2%) ADR were categorized as possible and 37 (41.1%) as probable. Conditional and unlikely were classified in three (3.3%) and two (2.2%) cases, respectively. Only one ADR was classified as definite. Similar categorization of the ADR, according to Naranjo algorithm was also presented in Table 1.
Slight agreement was found in the comparison of all judges in the pairwise analysis, and also for multiple judges, indicating poor reproducibility of WHO-UMC system (Table 1) . Thirteen cases were classified as unassessable by at least one judge and they were excluded for kappa analysis.
More than a half of the studied cases of ADR were categorized by the judges with some level of preventability. ADR was considered definitely preventable in 29 (32.2%) cases and probably preventable in 23 (25.6%) cases. Thirteen eight (42.2%) cases of ADR were considered not preventable by the participating judges. Kappa Fleiss also revealed a slight concordance among the judges (k=0.134; CI 95%: 0.018-0.225).