The use of modelling to evaluate new drugs for patients with a chronic condition: the case of antibodies against tumour necrosis factor in rheumatoid arthritis.

OBJECTIVES: To address the structural issues relating to mortality and quality of life (QoL) effects and to identify data on the general pattern of QoL of rheumatoid arthritis (RA) patients through a restructured and enhanced version of the Birmingham Preliminary Model (BPM). DATA SOURCES: Electronic databases and a postal survey of current UK rheumatological practice. REVIEW METHODS: The focus for this report was to evaluate two new drugs, etanercept and infliximab [antibodies against tumour necrosis factor (anti-TNFs)], for use in the treatment of RA using the Birmingham Rheumatoid Arthritis Model (BRAM). Having carried out a rapid systematic review of physician surveys of current disease-modifying antirheumatic drugs (DMARDs) usage patterns in adult patients with RA and a postal survey of consultant rheumatologists working in the UK, the drug sequences were then identified for the model. A series of analyses were then run using the model. The issue of specifying the correct comparison in the analysis being undertaken was investigated using two separate analyses: the situation of comparing anti-TNFs with placebo, and the comparison of a sequence using anti-TNFs with a sequence that represents current practice in the UK. RESULTS: Results from the survey of rheumatologists highlighted the fact that RA has different manifestations and responds to different agents in different patients, all of which makes any summary of practice difficult to achieve and open to the criticism of being an oversimplification. However, the findings generally agree with other surveys and trends observed, such as the increasing acceptance of methotrexate as first line drug of choice in patients with RA, especially if the disease is of an aggressive nature. The newer anti-TNF agents have also begun to be incorporated into use. The incremental cost-effectiveness ratios resulting from the use of an inappropriate comparator of placebo were consistently lower than in the base case where appropriate comparator drugs sequences are used. The focus of the BRAM on a drug sequence helped to avoid the incremental cost-effectiveness of new treatments appearing lower than they really are when inappropriate comparators are used. To test the effect on the analysis results of using the disease-modifying antirheumatic sequence that represents current UK practice, the BRAM was run for the strategies representing current UK practice. The results were not very different from the base-case results. CONCLUSIONS: The main achievement of this work was to bring about a more realistic modelling of real-life clinical pathways and events, as it has developed from the BPM to the BRAM. This has been brought about by overcoming structural and data limitations. In addition, the modelling approach reflected in the BRAM is applicable to other chronic conditions, especially those where a sequential approach to therapeutic options exists. The model has been successfully restructured so that different sequences of treatment can readily be considered, including the sequence that best represents current clinical practice in the UK. In addition, the flexibility inherent in using a modelling approach to consider these health policy questions has been demonstrated. One of the key uncertainties that can now be explored concerns the impact of new drugs on disease progression. Current evidence on this is weak, but should new agents demonstrate such a benefit then the BRAM may be a suitable vehicle through which to investigate the costs and full effects. Inevitably, there remain problems and limitations with the BRAM, but these are almost entirely data limitations. As data on these issues become available the BRAM provides a convenient tool through which reanalysis might be undertaken.