OBJECTIVES: To address the practical and methodological issues associated with using low-quality evidence outcomes in health economic modelling. METHODS: A cost-effectiveness model for disease-modifying drugs (DMDs) in multiple sclerosis (MS) in The Netherlands was used to assess how to deal with low-quality evidence in health economic modelling. The model adopted a 10-year time horizon and a societal perspective. A Markov model was constructed based on EDSS staging in MS, including relapse. The central focus was on disease progression — instead of relapse — which appeared to be the driver of the cost-effectiveness outcomes. The main data source was a recent Cochrane review estimating relative efficacy and acceptability of DMDs in relapse-remitting MS. Other data sources included additional published literature, clinical trials, and official price/tariff lists. RESULTS: The analysis based on the Cochrane review data showed that interferon beta-1a-R (Rebif) is cost-effective over interferon beta-1a-A (Avonex) (dominant) and interferon beta-1b (€27,654/QALY), but that interferon beta-1a-R is not cost-effective over glatiramer acetate. However, for disease progression, the level of evidence is considered very low (level 1) for all drugs, except interferon beta-1a-R (moderate - level 3), implying unreliable effectiveness outcomes which, consequently, can result in unreliable cost-effectiveness outcomes. Two reasonable alternative approaches may be to exclude very low evidence from the cost-effectiveness analysis or assume placebo efficacy. Alternative analyses, including placebo efficacy for disease progression for drugs of which the evidence is labelled very low by Cochrane (all except interferon beta-1a-R), strongly impacted outcomes: interferon beta-1a-R was cost-effective over interferon beta-1a-A (dominant), interferon beta-1b (€6,265), and glatiramer acetate (dominant). CONCLUSIONS: Inclusion of very low-quality evidence in health economic modelling may lead to unreliable cost-effectiveness conclusions. However, a gold standard is lacking for handling levels of clinical evidence in health economic models. One alternative, presented here, would be to assume placebo efficacy in such cases.