The company’s economic model
The model is appropriate for decision
making but there is uncertainty in
extrapolating overall survival after
transplant and the cure fraction used.
The company presented an economic
model in two parts: an initial decision
tree to determine if patients were in
remission after induction therapy, and
whether they had a stem cell transplant
or not, and then subsequent partitioned
survival models. The model had a 30-year
time horizon. This was assumed to be
a lifetime time horizon because patients
in the model were 60 to 75 years, as in
Study 301. To extrapolate beyond the trial
period, the company modelled parametric
curves separately by treatment group.
Overall survival and relapse-free survival
outcomes were modelled separately for
three groups based on data from Study
301: people in remission who had a stem
cell transplant, people in remission who
did not have a transplant and people who
were not in remission. For people in the
liposomal cytarabine–daunorubicin group
who were in remission and had a stem cell
transplant, the company chose a Gompertz
distribution to extrapolate overall survival.
This was based on clinical plausibility and
because it was the best fit to the trial data.
The committee considered that, although
the Gompertz distribution produced a
plateau, which would be expected after
transplant, the plateau seemed overly
optimistic. The committee agreed that the
Study 301 data were not mature enough
to justify this extrapolation, particularly
with the amount of censoring. At the
first committee meeting, the committee
noted that the modelled curve for the
comparator group did not reach a plateau
and stated that, after around 2 years,
general population mortality rates would
be applied for most people in the liposomal
cytarabine–daunorubicin group in its base-
case model because these rates were used
when the modelled mortality rates would
otherwise be lower. The ERG explored
several parametric curves for extrapolating
6 | 2019 | hospitalpharmacyeurope.com
overall survival after transplant for the
liposomal cytarabine–daunorubicin group.
It noted that the choice of curve had
a large effect on the predicted benefit and
therefore the cost-effectiveness results. So,
the ERG used a model averaging approach
to address the uncertainty. The committee
considered that this approach did not
address the clinical implausibility of the
extrapolation. The committee stated that
it would prefer to see a cure model for the
whole population, whether or not they
had a stem cell transplant. The committee
agreed that a plateau, or ‘cure’, should be
accounted for in the standard cytarabine
and daunorubicin survival extrapolation.
It also agreed that it would prefer to see
overall survival analysis based on a more
mature data cut to make the long-term
extrapolation more reliable. In response
to consultation, the company presented
statistical cure model extrapolations
for the whole population. However, the
company did not use these models in
the cost- effectiveness results because it
stated that the cure model for the whole
liposomal cytarabine–daunorubicin
group overestimated survival compared
with the Kaplan–Meier data and gave
overly favourable cost-effectiveness
results. Instead, the company used cure
models for overall survival after stem
cell transplant, which it stated matched
the updated Kaplan–Meier data well. The
cure models were based on the original
trial data (December 2015 data cut)
because the company only had a limited
dataset of updated individual patient
level data, which did not include event-
free status. The committee agreed that
it would have preferred to have seen the
whole population modelled together. The
company manually set a cure fraction
of 20% in the standard cytarabine and
daunorubicin group. The ERG noted that
this figure seemed to have been taken from
a visual inspection of the Kaplan–Meier
curve and that a 25% cure fraction could
also be considered as a plausible upper
limit. The company presented