In zhengguoxu/bootf2: Simulation and Comparison of Dissolution Profiles

knitr::opts_chunk$set(
  echo = TRUE,
  collapse = TRUE,
  comment = "#",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

lcran <- "https://CRAN.R-project.org/package=bootf2"

Installation{-}

• To install the released version from CRAN:

install.packages("bootf2", repos = "https://cloud.r-project.org/")

• To install the development version from Github, please install Rtools for building R packages from sources if you use Windows, then

# Need devtools. if you don't have it, install.packages("devtools") 
devtools::install_github("zhengguoxu/bootf2")

Introduction

The package bootf2 was developed to compare the dissolution profiles using bootstrap $f_2$ method, as recommended recently by several regulatory guidelines [@EMA-2018-09-QA.MSD.DISSO; @Davit-2013-03-BA; @Lum-2019-05-WS; @Mandula-2019-05-WS]. Several additional functions were included later for the simulation of the dissolution profiles.

Currently, there are four main functions in the package:

1. sim.dp() to simulate dissolution profiles using mathematical models or multivariate normal distribution. See vignette Simulation of Dissolution Profiles on CRAN for details.
2. calcf2() to calculate similarity factor $f_2$ according to different regulatory rules. See vignette Calculating Similarity Factor $f_2$ on CRAN for details.
3. sim.dp.byf2() to find a dissolution profile that, when compared to a given reference profile, has $f_2$ value equal to the predefined target $f_2$. See vignette Simulation of Dissolution Profiles with Predefined Target $f_2$ on CRAN for details.
4. bootf2() to estimate the confidence intervals of $f_2$s using bootstrap method. See vignette Confidence Intervals of f2 Using Bootstrap Method on CRAN for details.

In addition to the vignettes for the main functions, some common topics such as regulation rules are discussed in the vignette Introduction to bootf2 on CRAN.

The most basic usage is given below as a brief demonstration.

Examples

Function sim.dp()

The complete list of arguments are shown below. Read the function manual with ?sim.dp for more details.

dat <- sim.dp(tp, dp, dp.cv, model = c("Weibull", "first-order"),
              model.par = NULL, seed = NULL, n.units = 12L, product,
              max.disso = 100, ascending = FALSE, message = FALSE,
              time.unit = c("min", "h"), plot = TRUE,
              plot.max.unit = 36L, empirical = TRUE, cv.tol = 1e-6)

For the most basic use, the function can be run without any user provided arguments, e.g., sim.dp(). In such case, 12 units of individual dissolution profiles will be simulated using Weibull model with a typical sampling time points of 5, 10, 15, 20, 30, 45, and 60 min. A seed number will be randomly generated, if not provided by the user, and included in the output for reproducibility purpose.

library(bootf2)
# simulation. simple as that. 
d.ref <- sim.dp(seed = 1234)

The output of sim.dp() is a list of at least 3 components:

1. sim.summary: A data frame with summary statistics of all individual dissolution profiles.

print(d.ref$sim.summary)

1. sim.disso: A data frame with all individual dissolution profiles.

print(d.ref$sim.disso)

1. sim.info: A data frame with information of the simulation.

print(d.ref$sim.info)

Depending on the argument settings, there might be two additional components:

1. model.par.ind: A data frame of individual model parameters that are used to simulate the individual dissolution profiles if mathematical models are chosen for the simulation.

print(d.ref$model.par.ind)

1. sim.plot: A plot if plot = TRUE.

print(d.ref$sim.plot)

Simple case like this might be useful in situations such as testing other programs where data with certain format is needed. In general, to have better controlled outcomes, argument tp, model, and model.par should be provided.

Function calcf2()

The complete list of arguments are shown below. Read the function manual with ?calcf2 for more details. In addition, refer to the vignette Introduction to bootf2 on CRAN for detailed discussion on different regulatory requirements regarding to the applicability of $f_2$.

calcf2(test, ref, path.in, file.in, path.out, file.out,
       regulation = c("EMA", "FDA", "WHO", "Canada", "ANVISA"),
       cv.rule = TRUE, message = FALSE, min.points = 3L,
       f2.type = c("est.f2", "exp.f2", "bc.f2", "vc.exp.f2",
                   "vc.bc.f2", "all"), both.TR.85 = FALSE,
       digits = 2L, time.unit = c("min", "h"),  plot = TRUE,
       plot.start.time = 0, plot.max.unit = 24L)

The minimum required arguments are test and ref. Data can also be read from an Excel file. For interactive use, such as the examples below, the test and ref should be data frames with the time as the first column and individual profiles as the rest columns. The sim.disso data frame in the output of sim.dp() comes with the correct format, as shown above. This is the base function used by function bootf2().

# simulate a test data
d.test <- sim.dp(seed = 100, plot = FALSE, message = TRUE)

# calculate f2 with default settings
tmp.f2 <- calcf2(d.test$sim.disso, d.ref$sim.disso, message = TRUE)

print(tmp.f2)

When the conditions to apply $f_2$ are not fulfilled, the function will stop and, depending on the details of non-compliance of regulatory rules, show different error messages.

# simulate reference profile with CV% criterion not fulfilled  
d.ref2 <- sim.dp(seed = 456)

# output with error message
calcf2(d.test$sim.disso, d.ref2$sim.disso, message = TRUE)

Function sim.dp.byf2()

The complete list of arguments are shown below. Read the function manual with ?sim.dp.byf2 for more details.

dat <- sim.dp.byf2(tp, dp, target.f2, seed = NULL, min.points = 3L,
                   regulation = c("EMA", "FDA", "WHO", "Canada", "ANVISA"),
                   model = c("Weibull", "first-order"), digits = 2L,
                   max.disso = 100, message = FALSE, both.TR.85 = FALSE,
                   time.unit = c("min", "h"), plot = TRUE, sim.dp.out,
                   sim.target = c("ref.pop", "ref.median", "ref.mean"),
                   model.par.cv = 50, fix.fmax.cv = 0, random.factor = 3)

Given any dissolution profile dp at time points tp, and target $f_2$ value (e.g., target.f2 = 55), this function will find another dissolution profile such that when the newly simulated profile is compared to the dp, the calculated $f_2$ will be equal to the target $f_2$. If target.f2 is provided as a range, such as target.f2 = c(54.95, 55.04), then the calculated $f_2$ with simulated profile will be within this range.

# mean dissolution profile for tp
tp <- c(5, 10, 15, 20, 30, 45, 60)
dp <- c(51, 66, 75, 81, 88, 92, 95)

# find another profile with target f2 = 60
d.t2 <- sim.dp.byf2(tp, dp, target.f2 = 60, seed = 123, message = TRUE)

The model parameters in the output are more useful in simulation studies since they can be used as initial model parameter input to the function sim.dp() to simulate a large population of individual dissolution profiles that have the known population $f_2$ value when compared to target dissolution profile.

Function bootf2()

The complete list of arguments are shown below. Read the function manual with ?bootf2 for more details.

result <- bootf2(test, ref, path.in, file.in, path.out, file.out,
                 n.boots = 10000L, seed = 306L, digits = 2L, alpha = 0.05,
                 regulation = c("EMA", "FDA", "WHO", "Canada", "ANVISA"),
                 min.points = 1L, both.TR.85 = FALSE, print.report = TRUE,
                 report.style = c("concise",  "intermediate", "detailed"),
                 f2.type = c("all", "est.f2", "exp.f2", "bc.f2",
                             "vc.exp.f2", "vc.bc.f2"),
                 ci.type = c("all", "normal", "basic", "percentile",
                             "bca.jackknife", "bca.boot"),
                 quantile.type = c("all", as.character(1:9), "boot"),
                 jackknife.type = c("all", "nt+nr", "nt*nr", "nt=nr"),
                 time.unit = c("min", "h"), output.to.screen = FALSE,
                 sim.data.out = FALSE)

The minimum required arguments are dissolution profiles of test and ref. The function can output many different 90% confidence intervals for several $f_2$ estimators. With default settings, the function prints all confidence intervals for all $f_2$ estimators, and the result will be save in a text file.

# get test and reference data set with correct format
test <- d.test$sim.disso
ref  <- d.ref$sim.disso

# use most default settings (output all) but small number of bootstrap
# to have shorter run time for the example. default n.boots = 10000L
t_vs_r <- bootf2(test, ref, n.boots = 100L, print.report = FALSE,
                 output.to.screen = TRUE)

The output of the bootf2() is a list containing: 1. boot.ci: A data frame of bootstrap $f_2$ confidence intervals. 1. boot.f2: A data frame of all individual $f_2$ values for all bootstrap data sets. This can be used to make plots for visual presentation. 1. boot.info: A data frame with detailed information of bootstrap for reproducibility purpose, such as all arguments used in the function, time points used for calculation of \eqn{f_2}{f2}, and the number of NAs. 1. boot.summary: A data frame with descriptive statistics of the bootstrap $f_2$.

And depending on the function settings, it might contains boot.t and boot.r, lists of all individual bootstrap data sets for the test and reference products.

Disclaimer

Despite the best efforts the author has put into, the package is offered without any guarantee of accuracy and absolutely no warranty. Validation of the package, especially when it is used in regulatory field, is the responsibility of the users. The author accept absolutely no liability for any financial loss or risk to public health resulting from the use of this package.

References

zhengguoxu/bootf2 documentation built on Dec. 23, 2021, 9:19 p.m.