| Title: | Two-/Three-Stage Designs for Phase 1&2 Clinical Trials |
|---|---|
| Description: | Calculate optimal Zhong's two-/three-stage Phase II designs (see Zhong (2012) <doi:10.1016/j.cct.2012.07.006>). Generate Target Toxicity decision table for Phase I dose-finding (two-/three-stage). This package also allows users to run dose-finding simulations based on customized decision table. |
| Authors: | Wenchuan Guo, Jianan Hui, Bob Zhong |
| Maintainer: | Wenchuan Guo <[email protected]> |
| License: | GPL-2 |
| Version: | 1.1-8 |
| Built: | 2025-03-10 03:48:14 UTC |
| Source: | https://github.com/cran/tsdf |
adjust Zhong's 2-/3-stage design for over-/under-running
adj.two(n1, r1, s1, n2, alpha1, alpha2, beta, pc, pe, ...)adj.two(n1, r1, s1, n2, alpha1, alpha2, beta, pc, pe, ...)
n1 |
sample size at stage 1. |
r1 |
inefficacy boundary at stage 1. |
s1 |
efficacy boundary at stage 1. if no early stopping for efficacy, |
n2 |
sample size at stage 2. |
alpha1 |
left-side overall type I error. |
alpha2 |
right-side overall type I error. |
beta |
type II error. |
pc |
a numeric vector of response rate. should be a vector with length 1 or 2. |
pe |
alternative hypothesis. |
... |
not used argument. |
To be added
An object of class "opt.design" is a list containing:
bdry |
rejection regions |
error |
true type 1/2 errors |
n |
sample size at each stage |
complete |
complete list of feasible designs |
alpha1 |
input; left-side type 1 error |
alpha2 |
input; right-side type 1 error |
beta |
input; type 2 error |
pc |
input; a vector of response rate. |
pe |
input; a vector of alternative response rate |
sf |
input; the alpha-spending function used |
stage |
input; two- or three- stage design is used |
Wenchuan Guo <[email protected]>, Jianan Hui <[email protected]>
n1 <- 22 r1 <- 6 s1 <- 22 n2 <- 24 pc <- 0.4 pe <- pc + 0.15 alpha1 <- 0.3 alpha2 <- 0.1 beta <- 0.2 out <- adj.two(n1, r1, s1, n2, alpha1, alpha2, beta, pc, pe)n1 <- 22 r1 <- 6 s1 <- 22 n2 <- 24 pc <- 0.4 pe <- pc + 0.15 alpha1 <- 0.3 alpha2 <- 0.1 beta <- 0.2 out <- adj.two(n1, r1, s1, n2, alpha1, alpha2, beta, pc, pe)
Run dose-finding simulations based on a customized decision table.
dec.sim(truep, decTable, start.level = 1, nsim = 1000)dec.sim(truep, decTable, start.level = 1, nsim = 1000)
truep |
a vector of length k (the number of doses being considered in the trial), with values equal to the true probabilities of toxicity at the dose levels. |
decTable |
a customized decision table. (same format as output of |
start.level |
starting dose level. Defaults to 1, i.e. the lowest dose level. |
nsim |
the number of simulation trials. Defaults to 1000. |
Assume there are $d$ dose levels to be studied. Denote the cumulative number of patients treated and cumulative number of DLTs at the current dose level are $n_i$ and $m_i$, respectively. $n_max$ is the maximum number of patients permitted to be treated at each dose level. The procedure is as follows
Step 1 : Update cumulative number of DLTs $m_i$ and total number of patients $n_i$ treated at the current dose and use the decision table to make a decision: if decision is “S" –> step 2; if decision is “D" or “DU” –> step 3; if decision is “E” –> step 4
Step 2 : If $n_i = n_max$, declare dose i as the MTD; otherwise, update $m_i$ and $n_i$ with additional cohort of patients and go to Step 1.
Step 3 : If the current dose level is the highest dose level, then: if n_i < n_max, update $m_i$ and $n_i$ with additional cohort of patients and go to Step 1; otherwise, stop the trial and declare that the MTD is higher than the highest dose level (inconclusive); If the current dose is not the lowest dose, then: if $n_i-1 < n_max$, update $m_i-1$ and $n_i-1$ with additional cohort of patients and set the current dose level to be the next lower dose level, and go to Step 1; otherwise, stop the trial and declare the next lower dose level is the MTD; Additionally, if the decision is “DU”, record this dose level as DU and never treat additional patients at the current dose level again.
Step 4 : If the current dose level is the highest dose level, then: if $n_i < n_max$, update $m_i$ and $n_i$ with additional cohort of patients and go to Step 1; otherwise, stop the trial and declare that the MTD is higher than the highest dose level (inconclusive); If the next higher dose level is of status DU, then: if $n_i < n_max$, update $m_i$ and $n_i$ with additional cohort of patients and go to step 1; otherwise stop, the current dose level is MTD; Otherwise: if $n_i+1 < n_max$, update $m_i+1$ and $n_i+1$ with additional cohort of patients, set the current dose level to be next higher dose level, and go to step 1; else, the current dose level is the MTD.
the functions summary.dec.sim is used to obtain and print a summary table of the results (recommended). An object of class "dec.sim" is a list containing:
mtd |
a vector of dose levels giving the recommended maximum tolerated dose (MTD) at the end of the trial. |
mtd.prob |
a vector of length |
over.prob |
a vector of length |
n.patients |
the average number of patients dosed at each level. |
dlt |
the average number of DLTs experienced at each dose level. |
truep |
input; true probabilities of toxicity. |
start.level |
input; starting dose level. |
nsim |
input; number of simulated trails. |
Wenchuan Guo <[email protected]>
truep <- c(0.3, 0.45, 0.5, 0.6) res <- dec.table(0.6,0.4,0.2,0.3,c(3,3,3)) out <- dec.sim(truep, res$table, start.level = 2, nsim = 1000) summary(out, pt = 0.3)truep <- c(0.3, 0.45, 0.5, 0.6) res <- dec.table(0.6,0.4,0.2,0.3,c(3,3,3)) out <- dec.sim(truep, res$table, start.level = 2, nsim = 1000) summary(out, pt = 0.3)
Generate three stage dose finding decision table
dec.table(alpha.l, alpha.r, alpha.u, pt, n, sf.param = 4, pe.par = 0.25, ...)dec.table(alpha.l, alpha.r, alpha.u, pt, n, sf.param = 4, pe.par = 0.25, ...)
alpha.l |
left-side overall type 1 error. Control the upper bound of dose escalation. |
alpha.r |
right-side overall type 1 error. Control the lower bound of dose de-escalatition. |
alpha.u |
right-side overall type 1 error. This also controls the lower bound of dose de-escalatition, but it is used to find lower bound for "DU". |
pt |
a numeric vector of target toxicity. Should be a vector with 1 or 2(when the target is an interval). |
n |
a vector of sample size at each stage. |
sf.param |
a single real value specifying the gamma parameter for which Hwang-Shih-DeCani spending is to be computed; allowable range is [-40, 40]. Increasing this parameter implies that more error is spent early stage and less is available in late stage. Default to 4. |
pe.par |
alternative hypothesis that used to calculate power/type 2 error. The alternative is set to be |
... |
not used argument. |
Alpha-spending method is added to two-/three-stage designs. dec.table supports Hwang-Shih-DeCani spending function.
An object of class "dec.table" is a list containing:
table |
the generated decision table. |
alpha.two |
a vector of true type 1 error for two-tailed test. |
alpha.one |
a vector of true type 1 error for right-tailed test. |
beta |
a single value of true type 2 error(depends on alternative). |
E |
a vector of "E" bound. |
D |
a vector of "D" bound. |
DU |
a vector of "DU" bound. |
pt |
input; a vector of target toxicity |
n |
input; a vector with sample size at each stage. |
sf.param |
input; the alpha-spending function parameter used. |
Wenchuan Guo <[email protected]>
alpha.l <- 0.6 alpha.r <- 0.4 alpha.u <- 0.1 pt <- 0.3 # print out decision table for a 3+3+3 design n <- rep(3, 3) dec.table(alpha.l, alpha.r, alpha.u, pt, n)$table # 3+3 design n <- rep(3, 2) dec.table(alpha.l, alpha.r, alpha.u, pt, n)$tablealpha.l <- 0.6 alpha.r <- 0.4 alpha.u <- 0.1 pt <- 0.3 # print out decision table for a 3+3+3 design n <- rep(3, 3) dec.table(alpha.l, alpha.r, alpha.u, pt, n)$table # 3+3 design n <- rep(3, 2) dec.table(alpha.l, alpha.r, alpha.u, pt, n)$table
calculate optimal 2-/3-stage design given by Bob Zhong
opt.design( alpha1, alpha2, beta, pc, pe, stage = 2, stop.eff = FALSE, frac_n1 = NULL, frac_n2 = NULL, sf.param = NULL, show = FALSE, nmax = 100, n.choice = 1, ... )opt.design( alpha1, alpha2, beta, pc, pe, stage = 2, stop.eff = FALSE, frac_n1 = NULL, frac_n2 = NULL, sf.param = NULL, show = FALSE, nmax = 100, n.choice = 1, ... )
alpha1 |
left-side overall type I error. |
alpha2 |
right-side overall type I error. |
beta |
type II error |
pc |
a numeric vector of response rate. should be a vector with length 1 or 2. |
pe |
alternative hypothesis. |
stage |
2 or 3. default to 2 (2-stage design). |
stop.eff |
logical flag. default to |
frac_n1 |
proportion of n1. for 2-stage design, default to |
frac_n2 |
proportion of n2. Used for 3-stage design. default to |
sf.param |
a single real value specifying the gamma parameter for which Hwang-Shih-DeCani spending is to be computed; allowable range is [-40, 40]. Increasing this parameter implies that more error is spent early stage and less is available in late stage. For two-stage designs, default to |
show |
logical. If |
nmax |
maximum sample size. default to 100. |
n.choice |
stop criterion for the search of feasible designs. stop if number of designs is more than |
... |
not used argument. |
The two-stage design setup is: n1 patients are treated in the first stage. At the end of the first stage, either the trial continues to the second stage or inefficacy is concluded and the trial is stopped (early termination), depending on the number of responses observed at the first stage. If the trial does continue to the second stage, additional n2 patients are treated. Three-stage design is an extension of two-stage design where one stage is added between Stage 1 and 2. The left-side rejection region is response <= r_i for i = 1, 2, or 3 and right-side rejection region is response > s. Alpha-spending method is added to two-/three-stage designs. opt.design supports Hwang-Shih-DeCani spending function. You can change the definition of HSD function to use a different spending function.
An object of class "opt.design" is a list containing:
bdry |
rejection regions |
error |
true type 1/2 errors |
n |
sample size at each stage |
complete |
complete list of feasible designs |
alpha1 |
input; left-side type 1 error |
alpha2 |
input; right-side type 1 error |
beta |
input; type 2 error |
pc |
input; a vector of response rate. |
pe |
input; a vector of alternative response rate |
sf |
input; the alpha-spending function used |
stage |
input; two- or three- stage design is used |
Wenchuan Guo <[email protected]>, Jianan Hui <[email protected]>
Zhong. (2012) Single-arm Phase IIA clinical trials with go/no-go decisions. Contemporary Clinical Trials, 33, 1272–1279.
alpha1 <- 0.15 alpha2 <- 0.10 beta <- 0.15 pc <- 0.25 pe <- pc + 0.20 # calculate optimal two-stage design without using alpha-spending opt.design(alpha1, alpha2, beta, pc, pe, stage=2) ## Not run: # calculate optimal two-stage design with Pocock-like spending function opt.design(alpha1, alpha2, beta, pc, pt, stage = 2, sf.param = 1) # calculate optimal three-stage design with =O’Brien-Fleming like spending function opt.design(alpha1, alpha2, beta, pc, pt, stage = 3, sf.param = -4) ## End(Not run)alpha1 <- 0.15 alpha2 <- 0.10 beta <- 0.15 pc <- 0.25 pe <- pc + 0.20 # calculate optimal two-stage design without using alpha-spending opt.design(alpha1, alpha2, beta, pc, pe, stage=2) ## Not run: # calculate optimal two-stage design with Pocock-like spending function opt.design(alpha1, alpha2, beta, pc, pt, stage = 2, sf.param = 1) # calculate optimal three-stage design with =O’Brien-Fleming like spending function opt.design(alpha1, alpha2, beta, pc, pt, stage = 3, sf.param = -4) ## End(Not run)
Three plots are currently available: a plot of true toxicity at each dose level (type = "s"); a bar plot of the probability of selecting as the MTD for each dose level (type = "prob"); a bar plot of the average number of patients treated at each dose level (type = "np"); a bar plot of the average number of patients experienced DLT at each dose level (type = "dlt") and type = "all" generates all above plots.
## S3 method for class 'dec.sim' plot( x, pt, s = 1, type = c("all", "s", "prob", "np", "dlt"), label = TRUE, col = "cornflowerblue", text.col = "darkblue", cex = 1, ... )## S3 method for class 'dec.sim' plot( x, pt, s = 1, type = c("all", "s", "prob", "np", "dlt"), label = TRUE, col = "cornflowerblue", text.col = "darkblue", cex = 1, ... )
x |
an object of class |
pt |
a vector with target toxicity for each scenario. |
s |
scenario to be plotted. Defaults to 1. |
type |
plot type. See descriptions above. |
label |
a logical value indicating if values are shown on plot. |
col |
graphical parameter |
text.col |
plotting color of text shown. |
cex |
graphical parameter |
... |
arguments to be passed to |
# generate decision table dt <- dec.table(0.6,0.4,0.2,0.3,c(3,3,3)) # simulate trials from test data test.file <- system.file("extdata", "testS.csv", package = "tsdf") out <- sl.sim(dt$table, test.file) plot(out, pt=rep(0.3,2), s=1, type="all") plot(out, pt=rep(0.3,2), s=2, type="prob") plot(out, pt=rep(0.3,2), s=1, type="np") plot(out, pt=rep(0.3,2), s=2, type="dlt")# generate decision table dt <- dec.table(0.6,0.4,0.2,0.3,c(3,3,3)) # simulate trials from test data test.file <- system.file("extdata", "testS.csv", package = "tsdf") out <- sl.sim(dt$table, test.file) plot(out, pt=rep(0.3,2), s=1, type="all") plot(out, pt=rep(0.3,2), s=2, type="prob") plot(out, pt=rep(0.3,2), s=1, type="np") plot(out, pt=rep(0.3,2), s=2, type="dlt")
plot method for class "dec.table"
## S3 method for class 'dec.table' plot(x, ...)## S3 method for class 'dec.table' plot(x, ...)
x |
an object of class |
... |
Not used argument. |
plot.dec.table prints the decision boundarys.
truep <- c(0.3, 0.45, 0.5, 0.6) out <- dec.table(0.6,0.4,0.2,0.3,c(3,3,3)) plot(out)truep <- c(0.3, 0.45, 0.5, 0.6) out <- dec.table(0.6,0.4,0.2,0.3,c(3,3,3)) plot(out)
print method for class "dec.table"
## S3 method for class 'dec.table' print(x, ...)## S3 method for class 'dec.table' print(x, ...)
x |
an object of class |
... |
Not used argument. |
print.dec.table prints the decision table with legend keys.
print(dec.table(0.6,0.4,0.2,0.3,c(3,3,3)))print(dec.table(0.6,0.4,0.2,0.3,c(3,3,3)))
print method for class "opt.design"
## S3 method for class 'opt.design' print(x, ...)## S3 method for class 'opt.design' print(x, ...)
x |
an object of class |
... |
not used argument. |
alpha1 <- 0.20 alpha2 <- 0.1 beta <- 0.20 pc <- 0.5 pt <- pc + 0.2 out <- opt.design(alpha1, alpha2, beta, pc, pt, stage = 2, sf.param = 1) print(out)alpha1 <- 0.20 alpha2 <- 0.1 beta <- 0.20 pc <- 0.5 pt <- pc + 0.2 out <- opt.design(alpha1, alpha2, beta, pc, pt, stage = 2, sf.param = 1) print(out)
Run dose-finding simulations based on a customized decision table for a list of scenarios.
sl.sim(decTable, file, header = TRUE, sep = ",", ...)sl.sim(decTable, file, header = TRUE, sep = ",", ...)
decTable |
A customized decision table. (same format as output of |
file |
The name of the file which the data are to be read from. See details in |
header |
A logical value indicating whether the file contains the names of the variables as its first line. Default is |
sep |
The field separator character. Default is |
... |
arguments to be passed to |
In each line of the input file, the parameters must be ordered in accordance as follows: pt, start.level, nsim, truep. See details in read.table. The algorithm for dose-finding is described in dec.sim.
The function summary is used to obtain and print a summary table of the results. An object of class "dec.sim" (1 scenario) or "sl.sim" (more than 1 scenarios)is a list containing:
MTD |
A vector of dose levels giving the recommended maximum tolerated dose (MTD) at the end of the trial. |
n.patients |
The average number of patients dosed at each level. |
truep |
input; true probabilities of toxicity. |
start.level |
input; starting dose level. |
nsim |
input; number of simulated trails. |
Wenchuan Guo <[email protected]>
dt <- dec.table(0.6,0.4,0.2,0.3,c(3,3,3)) test.file <- system.file("extdata", "testS.csv", package = "tsdf") # use a customized decision table table.file <- system.file("extdata", "decTable.csv", package = "tsdf") dec <- read.table(table.file, sep=",", col.names=c(3,4,8,10), row.names = 1, check.names = FALSE) out1 <- sl.sim(dt$table, test.file) out2 <- sl.sim(dec, test.file)dt <- dec.table(0.6,0.4,0.2,0.3,c(3,3,3)) test.file <- system.file("extdata", "testS.csv", package = "tsdf") # use a customized decision table table.file <- system.file("extdata", "decTable.csv", package = "tsdf") dec <- read.table(table.file, sep=",", col.names=c(3,4,8,10), row.names = 1, check.names = FALSE) out1 <- sl.sim(dt$table, test.file) out2 <- sl.sim(dec, test.file)
summary method for class "dec.sim".
## S3 method for class 'dec.sim' summary(object, pt, ...)## S3 method for class 'dec.sim' summary(object, pt, ...)
object |
an object of class |
pt |
target toxicity for each scenario. |
... |
Not used argument. |
summary is used for formating important statistics for dose-finding simulation. Giving the target toxicity, it returns the probability of selecting current dose level as the MTD and over the MTD, probability of selecting the true MTD, probability of subjects treated at or below the true MTD, etc. The MTD is defined as the highest dose level such that the toxicity probability is less than target toxicity probability, if target is less than the smallest probability, then the lowest dose level is set as MTD. For example, if target is 0.3 and true toxicity for five doses are 0.1, 0.25, 0.35, 0.40, then MTD is dose 2.
test.file <- system.file("extdata", "testS.csv", package = "tsdf") dt <- dec.table(0.6,0.4,0.2,0.3,c(3,3,3)) out <- sl.sim(dt$table, test.file) pt <- c(0.3, 0.4) summary(out, pt)test.file <- system.file("extdata", "testS.csv", package = "tsdf") dt <- dec.table(0.6,0.4,0.2,0.3,c(3,3,3)) out <- sl.sim(dt$table, test.file) pt <- c(0.3, 0.4) summary(out, pt)