This study has some limitations which will be addressed in the following.
Within this study, we extrapolated 16-weeks-clinical trial data to a five year time horizon which is associated with some uncertainties. Furthermore, the study from Prantera et al. [14
], which was used for the long-term remission maintenance rates, only lasts for one year. However, to our knowledge, no other data exists, which gives information about the long-term effects of Mezavant versus Asacol treatment. Initial remission rates of Mezavant and Asacol were taken from a single study [9
] to ensure internal validity (e.g. consistent endpoints or homogeneity of included patients). Nevertheless, other clinical trials (Schroeder et al. [26
]; Sninsky et al. [27
]) are documented which also investigated the efficacy of Asacol. However, the reported remission rates are in line with the findings of the study of Kamm et al. [9
], although other clinical endpoints were set and a six-week-period was chosen as a trial length. All in all, these studies support the conservative assumption included in the present study regarding the inclusion of a remission rate of 61.0% for Asacol-patients receiving an increased dose of mesalazine.
It also might have been reasonable from the clinical trial data to assume an even lower remission rate due to the fact that patients who initially received Asacol and did not achieve a remission in the first eight weeks were henceforth treated with an increased dose of Mezavant. Since the remission rate of Mezavant in the first eight weeks was higher than the rate of Asacol (40.5% versus 32.6%), it could have been also assumed that an increased dosage of Asacol would not have had the same positive effect as an increased dose of Mezavant. Hence, an alteration of the model inputs reflecting the other clinical studies for Asacol might have been even more favorable to Mezavant.
Within the evaluation, it was assumed that patients are fully adherent. Higher adherence is associated with a reduced probability of relapses as many published findings suggest [28
]. One aspect which influences adherence is the frequency of dosing. For instance, Kane et al. [28
] reports adherence rates of maintenance therapy in UC-patients by dosing regimen: 39% for twice daily, 27% for three times daily and 6% for four times daily. Furthermore, the authors examined the relationship between adherence and treatment efficacy and found that adherent patients had an 89% chance of maintaining remission in comparison to patients who were non-adherent (39%). With respect to the two medications in this study, Mezavant is dosed once daily whereas Asacol is dosed three times daily for acute treatment and two times daily for maintenance. Therefore, it might be assumed that the difference between the remission rates of Mezavant and Asacol might have been even higher if adherence were considered within the model.
In addition, it is indicated that UC-patients have an increased risk of developing colorectal cancer (CRC). This risk, however, may be reduced by the adherent use of 5-ASA [30
]. Therefore, the inclusion of the increased risk of CRC due to non-adherence might also have as well increased the cost effectiveness of Mezavant in comparison to Asacol.
Further uncertainty is connected to the inclusion of surgery associated mortality data from the UK, which medical system is not directly comparable to the German system. However, due to the lack of sufficient German data, it seemed reasonable to include data from another developed European country. Furthermore, the percentage of patients in hospital who needed to undergo a surgery was taken from the German Federal Statistical Office (10.9%). However, this figure refers to the overall risk of surgery in hospital for all UC-patients and is not specific for the UC-population we are focusing on. Therefore, we have checked the importance of this figure by conducting a one-way sensitivity analysis (variation: 5-30%). The results change only slightly (incremental costs over a 5-year-period +/- 20 Euro). This shows that this specific figure has not a great impact on the results of this model.
Further uncertainties are connected to the structure of the model. Treatment for frequently relapsing patients can include oral immunosuppressants or TNF-alpha-inhibiting agents which were not included in the model. However, TNF do not play an important role in the treatment of UC-patients in Germany yet [21
]; furthermore, there is no valid data with respect to transition probabilities which could have been included in the model. The inclusion of TNF therapy would probably not have an effect on the original aim of the study (comparison of Asacol and Mezavant) because the population receiving such therapy is very small and both drugs were given the same treatment pathways in the model. Further uncertainty is connected with assumption that patients who suffer a severe relapse/mesalazine failure need hospitalization. On the other hand, some patients may also require an immediate surgery due to a severe relapse. However, no valid data exists for a more precise approach and the model structure includes no surgical or post-surgical complications. Nevertheless, post-surgical complications could likely occur in patients with a long active disease history. Due to lack of robust data with respect to further health states and transition probabilities, the model was not designed to assess post-surgical complications. Thus, only robust and conservative transition probabilities were included into the model calculations.
The resource use and costs which were included in the model are also connected with some uncertainties. The data source used for collection of resource usage and costs [21
] is not set as a longitudinal collection of data but on a cross-sectional study design and therefore represents rather a snapshot at a specific time point. However, the study is the most eligible one in the German context as it includes inpatient as well as outpatient resource use, is up to date (2007) and addresses the perspective of the SHI. According to the current German guidelines [22
] the drug prices of Mezavant and Asacol were taken from July 2009. To check whether the prices have changed after July 2009 or at the beginning of 2010 (and thereby the relative results of the analysis), the SHI-drug prices were re-calculated using current drug price. The last price change occurred in April 2010. After subtracting discounts and co-payments by patients, the drugs price of Asacol is currently 0.39 Euro per 400 mg and 1.00 Euro per 1,200 mg of Mezavant [23
]. Hence, Mezavant still is less costly for the SHI than Asacol. If these new prices were put into the model, the results would have even been more favorable to Mezavant. In general, due to the specialty of the reimbursement system in Germany, the results of this cost-effectiveness-model are not directly transferable to other foreign settings.
There is an urgent need for more research regarding the health care of patients suffering from UC. As stated above, long-term data which gives information on the long-term effects of specific UC-therapies are missing to date. Therefore, in models efficacy data from clinical trials needs to be extrapolated which is associated with uncertainties. Furthermore, utility data for UC-patients is very rare. The only utilities which were eligible for this cost effectiveness model came from two studies that were only available in an abstract format [19
]. Therefore, there is an urgent need to collect utility data for patients with UC, also for different country settings.
To our knowledge, there exists no cost effectiveness model for the German setting yet, which compares mesalazines in the treatment of UC. Hence, it is not possible to compare the results of this study with other German publications. This indicates a need for further health economic research in this area to identify cost effective options in the treatment of UC-patients in Germany.
In summary, the analysis shows that over a five year as well as a lifelong time horizon, Mezavant is less expensive and results in higher gains of QALYs compared with Asacol. The analysis and its results are reasonably robust as the PSA shows.