An orphan disease is defined in the EU as a disorder affecting less than 1 in 2 000 individuals. The concept of ultra-orphan has been proposed for diseases with a prevalence of less than 1:50 000. Drugs for ultra-orphan diseases are amongst the most expensive medicines on a cost-per-patient basis. The extremely high prices have prompted initiatives to evaluate cost-effectiveness and cost-utility in EU-member states. The objective of this review was to evaluate the quality of cost-effectiveness and cost-utility studies on ultra-orphan drugs. We searched 2 databases and the reference lists of relevant systematic reviews. Studies reporting on full economic evaluations, or at least aiming at such evaluation, were eligible for inclusion. Quality was assessed with the use of the Consensus on Health Economic Criteria (CHEC)-list. Two-hundred-fifty-one studies were identified. Of these, 16 fitted our inclusion criteria. A study on enzyme replacement and substrate reduction therapies for lysosomal storage disorders did not perform a full economic evaluation due to the high drug costs and the lack of a measurable effect on either clinical or health-related quality of life outcomes. Likewise, a cost-effectiveness analysis of laronidase for mucopolysaccharidosis type 1 was considered unfeasible due to lack of clinical effectiveness data, while in the same study a crude model was used to estimate cost-utility of enzyme replacement therapy (ERT) for Fabry disease. Three additional studies, one on ERT for Fabry disease, one on ERT for Gaucher disease and one on eculizumab for paroxysmal nocturnal haemoglobinuria, used an approach that was too simplistic to lead to a realistic estimate of the incremental cost-effectiveness (ICER) or cost-utility ratio (ICUR). In all other studies (N = 11) more sophisticated pharmacoeconomic models were used to estimate cost-effectiveness and cost-utility of the specific drug, mostly ERT or drugs indicated for pulmonary arterial hypertension (PAH). Seven studies used a Markov-state-transition model. Other models used were patient-level simulation models (N = 3) and decision trees (N = 1). Only 4 studies adopted a societal perspective. All but 2 studies discounted costs and effects appropriately. Drugs for metabolic diseases appeared to be significantly less cost-effective than drugs indicated for PAH, with ICERs ranging from €43 532 (Gaucher disease) to €3 282 252 (Fabry disease). Quality of studies using a Markov-state-transition or patient-level simulation model is in general good with 14-19 points on the CHEC-list. We therefore conclude that economic evaluations of ultra-orphan drugs are feasible if pharmacoeconomic modelling is used. Considering the need for modelling of several disease states and the small patient groups, a Markov-state-transition model seems to be most suitable type of model. However, it should be realised that ultra-orphan drugs will usually not meet the conventional criteria for cost-effectiveness. Nevertheless, ultra-orphan drugs are often reimbursed. Further discussion on the use of economic evaluations and their consequences in case of ultra-orphan drugs is therefore warranted.