Donepezil, rivastigmine, galantamine and memantine for ...

130
Economic analysis
Results are sensitive to a range of alternative
inputs, particularly in relation to effectiveness,
health state utility data and cost inputs for longer
term care. Large differences in cost per QALY are
seen where a 1-point shift (in both directions) for
effectiveness data (change in ADAS-cog) is
considered. Given that the benefit of the
intervention is seen as delaying the time to FTC
(or reducing time in FTC), results are sensitive to
changes in the assumptions/costs associated with
FTC, that is, where it is assumed that all
institutional costs fall on the NHS and PSS and
that the majority of patients requiring FTC are
institutionalised, the cost per QALY is in the
region of £50,000–69,000.
Where the impact of mortality, including a zero
rate, is reduced, there is an impact on costeffectiveness
results but the differences are not
dramatic, and the cost per QALY remains in the
region of £61,000–76,000.
Owing to relatively small incremental QALY gains,
the cost per QALY is sensitive to small changes in
the 5-year costs; for example, where the entry costs
(6-month drug and monitoring cost associated with
treatment) are disregarded the cost per QALY falls
to between £63,000 and £71,000, a drop in the
region of £20,000 per QALY. This point is
supported (in the other direction) by the increases
seen in the cost per QALY estimates when it is
assumed that monitoring costs are increased by
one further additional outpatient appointment per
year, or by one additional outpatient and two
additional GP visits per year, with cost per QALY
increases of around 10 and 20%, respectively.
The cost-effectiveness findings are sensitive to
changes in the difference in health state values
between the health states of pre-FTC and FTC.
The base case shows a mean difference of 0.26 (on
the 0–1 QALY scale), and with (1) an assumption
of a 0.55 mean difference it is seen that the cost
per QALY estimates fall by over 50%, (2) an
assumption of a 0.16 mean difference it is seen
that the cost per QALY estimates increase by over
50%.
Where the patient group is assumed to have a
higher prevalence of EPS and psychotic symptoms
(20% versus a baseline of 6.2 and 10%,
respectively), small changes in the cost-effectiveness
findings are seen. Where we assume a benefit of
treatment on the presence of psychotic symptoms
(a 20% reduction in the number of patients
showing psychotic symptoms), this does have an
effect on the cost-effectiveness, with the cost per
QALY falling; however, only where it is assumed
that the majority of the patient group have
psychotic symptoms and that the drug treatment
has an effect on the prevalence of symptoms does
the cost-effectiveness fall dramatically.
Limitations of the SHTAC
cost-effectiveness model
The rationale for the model structure has been
discussed, and limitations with the use of the
predictive risk equation for progression of AD to
FTC 131 have been highlighted. The fact that the
model is provided to offer additional information
to the NICE Appraisal Committee, to enable them
to consider independent analysis on the costeffectiveness
of donepezil, rivastigmine and
galantamine in the context of the same modelling
methodology and input values for costs and
outcomes has also been highlighted.
The model is simplified and does not include
consideration of dropouts from treatment (as seen
in all trials for these products); this is based on a
need to keep the model simple, and it is felt that it
is a conservative assumption that favours the
intervention. Where dropouts are factored into
this modelling approach, additional costs would be
expected on the treatment cohort (compared with
usual care), with no additional benefits accruing,
therefore the estimated cost-effectiveness would be
expected to be greater with consideration of
dropouts within the disease modelling process.
However, it is accepted that it is a limitation of the
model.
Good-quality input data for costs or outcomes
(health state utilities) have not been identified;
however, this appears to be a common problem
across all reported cost-effectiveness modelling
studies. Furthermore, there is uncertainty over
parameter values depicting the proportions of
FTC patients by setting (community/institution)
and the proportions that are publicly funded when
in an institutional setting. Once again these are
common limitations in the literature available to
inform on modelling disease progression and costeffectiveness
in AD, but these are accepted as
limitations in the SHTAC model.
Numerous sources of data have been relied upon
to define our cohort of AD patients, and clinical
opinion, from a Southampton AD treatment
centre, was sought to estimate the additional NHS
resources associated with management and
monitoring of AD patients on treatment. These
are accepted to be limitations with the SHTAC
modelling approach.