A systematic review and economic model of the effectiveness and ...

106
Economic model
IR-MPH was assumed to apply to all drug
treatments. However, owing to the variation in
programmes, this analysis can tell us little about
combination therapy in practice in the UK.
Adverse events
The model assumes that patients who fail to
tolerate treatment during the titration period will
discontinue that treatment if dose modification
does not address the problem, and that any sideeffects
will dissipate upon treatment cessation.
Hence the main consequence of adverse events in
the model is discontinuation of treatment, and
intolerable side-effects are not associated with an
additional utility decrement. The cost and utility
estimates for responders (i.e. patients continuing
on treatment) take account of the minor sideeffects
that commonly accompany treatment for
ADHD. However, it must be noted that they do
not discriminate between treatments in this
respect. As noted in the clinical effectiveness
review in Chapter 4, there is little trial evidence to
discriminate between the active therapies in terms
of tolerable side-effects. Of the events assessed,
insomnia may be more common with stimulant
therapy in comparison with ATX, so this must be
considered when interpreting the model results.
Another important factor is the lack of data
regarding long-term adverse events beyond the
period observed in the clinical trials.
In the clinical trials, patients withdrew from
treatment for many reasons, including lack of
efficacy and intolerable adverse events. The
reasons given differed between trials, and in some
cases the reason for withdrawal was not stated. As
a result, the withdrawal rates for the model were
calculated to include all reported withdrawals,
regardless of the reason given. This approach was
chosen to maintain consistency between trials with
differing definitions of withdrawal, and it is also a
conservative approach from the point of view of
NICE. This is because a higher rate of withdrawal
will increase the cost-effectiveness ratios associated
with active treatment, and so a treatment that
appears cost-effective using this broad definition
of withdrawal would appear even more costeffective
under a more precise definition.
However, it must be noted that in four of the 10
trials shown in Table 87, 63,83,90,94 a proportion of
withdrawals were attributed to non-response. This
demonstrates that although the approach taken
may be conservative and consistent in calculating
withdrawal rates, it may include some double
counting of non-responders. The degree of double
counting is difficult to quantify because, as noted
in the clinical effectiveness review in Chapter 4,
none of the four trials calculated response in an
ITT analysis.
The probability of withdrawal was calculated in the
same way as the response rates, hence the same
model applies, as shown in Appendix 9. Owing to
the lack of reported data on withdrawal, one set of
rates was calculated including all studies that
provided an estimate of response rate, regardless
of definition of response. Table 87 details the data
used to calculate withdrawal rates for the
economic model. These data apply to all analyses,
regardless of definition of response.
In the secondary analysis including combination
therapy, it was assumed that withdrawal would
only be induced by pharmacotherapy, and so the
same withdrawal rates are applied as to drug
monotherapy.
As noted earlier in this chapter, few data are
available on adverse events associated with longterm
use of pharmacotherapy for ADHD.
Therefore, the adverse events reflected in the
model are limited to those observed during the
treatment phase of the included clinical trials.
Compliance
Compliance can be thought of as the ability of
patients to take the required number of doses of
medication, or of the ability of patients to take
pills within the correct time frame; these are dosetaking
and dose-timing compliance, respectively.
This section of the report refers to dose-taking
compliance. If patients take fewer pills than
prescribed by their doctor, they will be receiving a
lower dose of medication. In their submissions,
Janssen-Cilag and Celltech both put forward the
argument that compliance to a midday dose of
treatment will be lower than compliance to an
early morning dose. They argue that double-blind,
double-dummy trials do not capture the effect of
improved compliance to the once-daily extendedrelease
formulations of MPH as both comparator
groups must take more than one pill per day (the
extra pills being placebo in the ER-MPH groups).
In explanation, in a double-blind, double-dummy
trial, patients are blinded to treatment and receive
the same number of doses of medication per day
in each arm. Dummy pills (placebo) are used to
prevent patients identifying the treatment to
which they have been allocated by counting the
number of doses they receive. This means that in a
double-blind, double-dummy trial of IR-MPH TID
versus ER-MPH12, patients in both arms would
receive three pills each day. In the IR-MPH arm,
all the pills would contain an active dose of IR-