OBJECTIVE: To verify the reliability of the methodology used in the manufacturers’ Budget Impact Analyses (BIAs) submitted in reimbursement dossier to Polish HTA Agency, we compared public payer (National Health Fund – NHF) actual expenditures on selected drugs with their estimated costs. METHODS: BIAs of medicines reimbursed for at least one year and assessed previously in Agency were selected for the analysis. The BI estimated by the manufacturer in the first year after product’s introduction to reimbursement was used. Estimated size of target population was compared with the actual one. The actual data was obtained from the NHF. BIAs for selected medicines were critically appraised to determine any variables that may have affected its reliability. RESULTS: 20 BIAs met inclusion criteria and were included into the study. Forty-five percent (9/20) of the BI expenditures were underestimated. Median and mean difference between actual and estimated expenditure were 17,22% and 30,25% respectively. The overestimation was found in 55% BIAs (11/20); median = 262,26%, mean difference = 2267,25%. Population size was underestimated in 65% (13/20) of the BIAs. Median and mean difference between actual and estimated population were 25,61% and 33,93% respectively. Overestimation of the population was found in 35% (7/20) of the BIAs; median = 566,67%, mean difference = 573,28%. The main factors that could influence differences between predictions and actual spending were: underestimated (65%) or overestimated (35%) number of patients eligible for treatment, overestimated (30%) or underestimated (25%) market share, wrong assumption on 100% compliance (15%).   CONCLUSION: The study has demonstrated large variances between predicted budget impact and actual expenditures on drugs. It also revealed significant weaknesses in the quality of submitted BIAs, e.g.: errors in calculations, very limited data provided by manufacturer that unable to complete revision and reproduce of figures in the calculation.