We have demonstrated that it is feasible to meet the practical challenges of undertaking a risk sharing scheme within the NHS. This also illustrates the potential of the NHS as a 'population laboratory' and further supports the development of research networks . We have collected baseline and annual review data for up to three years on over 5000 patients recruited from 70 centres across the UK, and continued collection of these data for the remaining seven years of the scheme should allow identification of a price at which the NHS can provide DMTs in a cost-effective manner. This is important because the debate about the cost-effectiveness of DMTs continues, despite the literature on their effectiveness in certain patients, and their wide international use . We are not aware of any other published evidence showing the practicality of implementing risk sharing schemes.
The MS Risk Sharimg Scheme Monitoring Study has several advantages for measuring cost-effectiveness in usual clinical practice. Because our cohort includes the majority of patients prescribed DMTs over the recruitment period in the vast majority of prescribing centres, and since we have not altered usual clinical practice in these centres, the external validity of our study is very high. This cohort of patients represents the vast majority of those prescribed DMTs according to ABN guidelines over the recruitment period, and the data collected show the expected demographic characteristics in terms of age, sex and MS type. The DMT prescribing pattern in the cohort reflects clinical practice in the UK, though the considerable variation found between centres requires further investigation. The study already has follow up as long as any published RCT and will be the largest cohort of DMT treated patients with long term follow up. The ScHARR MS model predicts long term EDSS progression under DMT treatment using natural history data from a cohort of Canadian MS patients followed up for 25 years, together with a reanalysis of RCT data on the impact of treatments on progression [2, 6]. There are no concurrent control patients, and it has been argued that a long term RCT would have been more appropriate than the MS Risk Sharing Scheme . However, it would have been unacceptable to many patients and clinicians to randomise to no treatment (since benefit on relapse rates has already been proven). Randomising between DMT products was not acceptable to the pharmaceutical companies who were parties to the Risk Sharing Scheme. The study will also enable accurate estimation of rates of switching and stopping DMTs as well as the costs and quality of life associated with different EDSS states which it is essential to know to accurately estimate the cost-effectiveness of the DMTs. We have ensured a complete transmission of all entered data at the time of the analysis reported here, locking the database on 31st January 2006. In these data between 12.0% (for those in the study at least one year) and 28.2% (for those in the study at least 3 years) of interim EDSS review scores were missing. However, the cost-effectiveness analysis linked to pricing, proposed in the original Health Service Circular , will not use these interim scores. The analysis depends only on the baseline score and the last confirmed score collected before analysis. Recording final review scores will therefore be a priority for future data collection. Cost-effectiveness may differ between subgroups of patients at different stages of the disease and subgroup analyses will be important to address this.
Over the first three years, we have found small but statistically significant progression of disease, as measured by mean EDSS changes, in MS patients on DMTs. This is consistent with previous, short term, RCTs where the small changes observed have made it difficult to assess the effect of DMTs on disease progression . However, the small changes in mean EDSS between baseline and each annual review mask considerable variation in individual patients. At annual reviews EDSS scores have decreased almost as often as increased, compared to baseline. This variation could be due to a number of factors. With EDSS, as with any outcome measure, there is some intra- and inter-observer variation, though when we restricted our analysis to patients whose EDSS assessment was made by the same clinician at baseline and annual review, the findings were unaffected. It is also possible that the patients whose EDSS scores have decreased were still recovering from relapse at the time of baseline assessment. A third possibility is that the lower EDSS scores represent true (if temporary) regressions in disability in these patients. Without a control group, it is hard to say whether this level of regression is due to the effect of DMTs, or whether it occurs in the disease's natural history, though by the end of the ten year follow up, there will be a much clearer picture of overall progression on these drugs. The annual relapse rate at first annual review was reduced by 54% compared to baseline. However, differences in the estimation of relapse rate at baseline and at annual review and the possibility of a regression to the mean caused by patients with temporarily high relapse rates being entered into the scheme mean that caution is required in interpreting this finding. Nevertheless, the reduction observed is in line with that seen in previous RCTs . The results from data source verification, the inter-rater reliability study and low rates of missing data are supportive of data quality, as is the EDSS distribution pattern, which is as would be expected from previous cohort studies .
Our data indicate possible heterogeneity in disease progression in specific population subgroups. In particular, older age at disease onset may be associated with faster rate of progression in disability. This heterogeneity is consistent with other recently published reports [11, 12], and has implications for the future planned analyses of the data to determine potential price adjustments. The natural history cohort used in the analysis should be adjusted from the Canadian study population to match the characteristics of patients entering the Risk Sharing Scheme. It is also important that the natural history data is validated, since there is evidence that the incidence of MS is changing and that the natural history may be becoming more benign [11, 13]. To ensure a level playing field in the assessment of cost-effectiveness of the different DMTs, case-mix adjustment will be necessary. Further complexity is added by patients switching and stopping drugs, and this needs to be addressed in the analysis. Additionally, the cost effectiveness may depend on characteristics that influence responsiveness such as neutralising antibodies and relapse frequency.
We have shown that the practical difficulties in establishing the Risk Sharing Scheme can be overcome, and the scientific challenges we have identified can be addressed through appropriate methods of analysis. However, perhaps the most important threat to the NHS's ability to determine the cost-effectiveness of DMTs comes from the tensions inherent in such a scheme. These arise from the differing interests of the Department of Health, the pharmaceutical companies, researchers and patients. The commercial risks associated with the scheme led to the involvement of the companies as well as the Department of Health in the scheme's governance. The scheme was retendered but the ScHARR consortium decided not to apply as they were not happy with the proposed arrangements for data access and publication rights, and the scheme is now being undertaken by a contract research organisation. The data that is being collected by the contract research organisation for use in future publications may not be the same as the data we have presented here. The data from the scheme is owned by the MS Trust.