Suppose we are planning a drug development program testing the superiority of an experimental treatment over a control treatment. Our drug development program consists of an exploratory phase II trial which is, in case of promising results, followed by a confirmatory phase III trial.
The drugdevelopR package enables us to optimally plan such programs using a utility-maximizing approach. To get a brief introduction, we presented a very basic example on how the package works in Introduction to planning phase II and phase III trials with drugdevelopR. In the basic setting, only one phase II and one phase III trial were conducted. Contrary to this, we now investigate what happens when multi-arm trials are conducted, where several different treatments or several doses of the same treatment are tested at the same time. The drugdevelopR package enables multi-arm program with two treatment arms and one control arm.
Suppose we are developing a new tumor treatment, exper. The
patient variable that we want to investigate is how long the patient
survives without further progression of the tumor (progression-free
survival). This is a time-to-event outcome variable. We want to test two
different doses of, exper1 and exper2. Therefore, we
will use the function optimal_multiarm, which calculates
optimal sample sizes and threshold decisions values for time-to-event
outcomes when multi-arm trials are conducted.
Within our drug development program, we will compare our two experimental treatments exper1 and exper2 to the control treatment contro. The treatment effect measure is given by \(\theta = −\log(HR)\), where the hazard ratio \(HR\) is the ratio of hazards of each treatment and the control.
After installing the package according to the installation instructions, we can load it using the following code:
As, the parameters in the multi-arm setting differ slightly from the other cases (bias adjustment and multitrial), we will explain them in more detail. Contrary to the basic setting, the treatment effect are always assumed to be fixed, so the option to use prior distributions is not available. Nevertheless, some options to adapt the program to your specific needs are also available in this setting (see More parameters), however, skipping phase II and choosing different treatment effects in phase II and III are not possible.
The following parameters are different to the basic setting, even if the names are the same.
hr1 is our hazard ratio for the first dose
exper1. As already explained above, we assume that our
experimental treatment exper1 leads to unfavorable events
occurring only 75% of times compared to the control treatment
contro. Therefore, we set hr1 = 0.75. Analogously,
hr2 is the hazard ratio for our second dose
exper2. For our example, we set hr2 = 0.8.ec is the control arm event rate for phase II and III,
how often the unfavorable event occurs in the control group. We assume
that 60% of patients in the group contro have an unfavorable
event, thus, we set ec = 0.6.n2min and n2max specify the minimal and
maximal number of participants for the phase II trial. The package will
search for the optimal sample size within this region. For now, we want
the program to search for the optimal sample size in the interval
between 20 and 400 participants. In addition, we will tell the program
to search this region in steps of ten participants at a time by setting
stepn2 = 10. This is different to optimal_tte,
where the region of events (and not the sample size region) is
searched.One parameter that is not in the basic setting is the following parameter:
strategy allows the user to choose
between three strategies. One can select strategy = 1 in
which only the best promising candidate of the two doses proceeds to
phase III, strategy = 2 in which all promising candidates
proceed to phase III or strategy = 3, in which both
strategies are explored. For our example we want to explore both
settings, so we use strategy = 3.The other parameters are identical to the basic setting. Let’s plug them into the function:
res <- optimal_multiarm(hr1 = 0.75, hr2 = 0.80, ec = 0.6, # define assumed true HRs and control arm event rate
n2min = 100, n2max = 200, stepn2 = 10, # define optimization set for n2
hrgomin = 0.76, hrgomax = 0.9, stephrgo = 0.02, # define optimization set for HRgo
alpha = 0.025, beta = 0.1, # drug development planning parameters
c2 = 0.75, c3 = 1, c02 = 100, c03 = 150, # define fixed and variable costs for phase II and III
b1 = 1000, b2 = 2000, b3 = 3000, # gains for each effect size category
strategy = 3,
num_cl = 3)After setting all these input parameters and running the function, let’s take a look at the output of the program.
res
#> Optimization result where only the most promising candidate continues:
#> Utility: 66.88
#> Sample size:
#> phase II: 140, phase III: 393, total: 533
#> Probability to go to phase III: 0.75
#> Total cost:
#> phase II: 205, phase III: 505, cost constraint: Inf
#> Fixed cost:
#> phase II: 100, phase III: 150
#> Variable cost per patient:
#> phase II: 0.75, phase III: 1
#> Effect size categories (expected gains):
#> small: 1 (1000), medium: 0.95 (2000), large: 0.85 (3000)
#> Success probability: 0.38
#> Success probability for a trial with:
#> two arms in phase III: 0.38, three arms in phase III: 0
#> Significance level: 0.025
#> Targeted power: 0.9
#> Decision rule threshold: 0.82 [HRgo]
#> Assumed true effects [HR]:
#> treatment 1: 0.75, treatment 2: 0.8
#> Control arm event rate: 0.6
#>
#> Optimization result where all promising candidates continue:
#> Utility: 56.6
#> Sample size:
#> phase II: 100, phase III: 481, total: 581
#> Probability to go to phase III: 0.68
#> Total cost:
#> phase II: 175, phase III: 583, cost constraint: Inf
#> Fixed cost:
#> phase II: 100, phase III: 150
#> Variable cost per patient:
#> phase II: 0.75, phase III: 1
#> Effect size categories (expected gains):
#> small: 1 (1000), medium: 0.95 (2000), large: 0.85 (3000)
#> Success probability: 0.4
#> Success probability for a trial with:
#> two arms in phase III: 0.18, three arms in phase III: 0.22
#> Significance level: 0.025
#> Targeted power: 0.9
#> Decision rule threshold: 0.78 [HRgo]
#> Assumed true effects [HR]:
#> treatment 1: 0.75, treatment 2: 0.8
#> Control arm event rate: 0.6The program returns a data frame, where the output for all implemented strategies is listed. For strategy 1 (only the best promising proceeds to phase III) we get:
res[1,]$n2 is the optimal number of participants for
phase II and res$n3 the resulting number of events for
phase III. We see that the optimal scenario requires 140 participants in
phase II and 393 participants in phase III.res[1,]$HRgo is the optimal threshold value for the
go/no-go decision rule. We see that we need a hazard ratio of less than
0.82 in phase II in order to proceed to phase III.res[1,]$u is the expected utility of the program for
the optimal sample size and threshold value. In our case it amounts to
66.88, i.e. we have an expected utility of 6 688 000$.The results for strategy 2 (all promising proceed to phase III) are:
res[2,]$n2 is the optimal number of participants for
phase II and res$n3 the resulting number of events for
phase III. We see that the optimal scenario requires 100 participants in
phase II and 481 participants in phase III.res[2,]$HRgo is the optimal threshold value for the
go/no-go decision rule. We see that we need a hazard ratio of less than
0.78 in phase II in order to proceed to phase III.res[2,]$u is the expected utility of the program for
the optimal sample size and threshold value. In our case it amounts to
56.6, i.e. we have an expected utility of 5 660 000$.Furthermore, consider the return values sProg,
sProg2 and sProg3. Contrary to the basic
setting, sProg2 is the probability of a successful program
with only one treatment in phase III (i.e. two arms) and
sProg3 is the probability of a successful program with two
treatments in phase III (i.e. three arms). Hence, for strategy 1 (only
the most promising treatment in phase II proceeds to phase III)
sProg = sProg2 = 0.38. With strategy 2 (all promising
treatments proceed to phase III), the success probability
sProg is the sum of the probability that both treatments go
to phase III sProg3 = 0.22 and the probability that only
one treatment proceeds to phase III sProg2 = 0.18.
In this article, we presented an example where multi-arm trials are
conducted. Note that this example is not restricted to time-to-event
endpoints but can also be applied to binary and normally distributed
endpoints by using the functions optimal_multiarm_binary
and optimal_multiarm_normal.
For more information on how to use the package, see: