From the reviews of the literature and the economic evaluations examined here, it appears that the cost-effectiveness studies on the use of donepezil in AD patients have reported a variety of conclusions with studies reporting either cost savings over time (usually from a societal perspective) or cost neutrality as well as incremental costs, alongside with benefits in cognitive function (MMSE) associated with the treatment. Likewise, the results of the cost-effectiveness studies of rivastigmine appear inconclusive, being either cost savings or cost incurring. Further, while most of those studies reported that rivastigmine treatment delays the cognitive decline in AD patients over time, this is based on the assumption that a delay in cognitive decline translates into meaningful long-term patient outcomes. The cost-effectiveness studies of galantamine were rather similar in methods, and reported a delay in time to full-time care need and to institutionalisation associated with a cost saving or cost incurring profile. In moderately severe to severe AD patients, memantine was reported to be the dominant option (i.e. more effective and less costly) in comparison with usual care.
The heterogeneity and unreliability in the studies’ conclusions about the costeffectiveness of AD medications mainly stems from the lack of robustness in the assumptions used to model disease progression and in the final outcomes used to consider patient benefits.
A key assumption in most models estimating the cost-effectiveness of AD medications (donepezil, rivastigmine, memantine and some models for galantamine) is that the severity of the AD states (i.e. the mild, moderate and severe disease stages) and the progression between the states are defined with MMSE scores. There are however concerns about the use of cognition alone to model disease progression over time, since other factors (such as the functional ability) may also be determinant in the progression of the disease or the need for institutionalisation. By contrast, in Loveman et al. and in most models for galantamine, disease progression, though limited to two health states (pre-FTC and FTC), was defined according to both cognitive and noncognitive criteria.
In a preceding chapter of this short HTA, it was reported that treatments with ChEIs for mild to moderately severe AD or with memantine for moderately severe to severe AD showed statistically significant benefits for some outcome measures in clinical trials (e.g. change in a cognitive scale score). However, the link from those short-term clinical trial outcomes with longer-term, more final patient-related outcomes as used in most economic evaluations (i.e. delay in disease progression, reduction in institutionalisation) is not straightforward and appears to be lacking in the literature so far. Such link is nevertheless assumed in most economic evaluations of AD medication. In this short HTA, we could only identify one high-quality RCT demonstrating a significant reduction in the rate of institutionalisation with non-pharmaceutical treatments (caregiver support) of AD patients’ caregivers. In the AD2000 trial, the risk of institutionalisation in persons taking donepezil (5 or 10 mg) was not found to be different from the risk of institutionalisation in persons taking placebo over a 3-year follow-up period. Also, in their recent study, IQWiG reports that no interpretable data are available on the AD medications goal of delay to institutionalisation.
Another common problem across all reported cost-effectiveness modelling studies is the lack of good quality input data, e.g. health state utilities, transition probabilities...
Numerous sources of data had thus to be combined from disparate studies to feed the models and many assumptions had to be made to palliate this paucity of data.