OBJECTIVES: Currently in the Netherlands a randomised clinical trial is executed to compare two outpatient psychotherapies for patients with Borderline Personality Disorder (BPD). The goal of the present study was to calculate the cost-of-illness (COI) of BPD for Dutch Society. METHODS: We used a prevalence-based approach, which takes into account total yearly costs of all patients who are diagnosed with BPD at a certain point in time. COI was calculated both top-down and bottom-up. For top-down calculation, prevalence figures from existing registrations and costs of the Dutch health care system from government publications were used. Baseline cost interviews of 88 BPD-patients in the trial were used to estimate bottom-up COI. BPD was defined according to ICD-9 and ICD-10 (top-down) and DSM-IV (bottom-up) classifications. RESULTS: Based on literature, prevalence of BPD in the Dutch general population was estimated at 1.1%. For all cost items, large differences arise between the bottom-up and the top-down approach. Total yearly societal are €200,184,828 top-down, and 16 times as high €3,258,240,100 for bottom-up. Healthcare costs represent 0.03% and 1.03% of total Dutch health care expenditure, respectively. CONCLUSION: Our results show large differences between the two methods. The top-down figure probably is an underestimation of true costs due to incomplete registrations. On the other hand, the bottom-up patient group may not be representative of the Dutch BPD population because of the in-en exclusion criteria used in the trial, which exclude the very mild and the very severe cases. In conclusion, we recommend to assess COI and prevalence in a combined design. First, prevalence in the general population is assessed. Subsequently, those subjects diagnosed with the disease under study should be followed, receiving care as usual, in order to determine COI. This is the only way to match bottom-up patient group and total population.