Type: Package
Title: Non-Parametric Recruitment Prediction for Randomized Clinical Trials
Version: 0.2.0
Description: Accurate prediction of subject recruitment for Randomized Clinical Trials (RCT) remains an ongoing challenge. Many previous prediction models rely on parametric assumptions. We present functions for non-parametric RCT recruitment prediction under several scenarios.
License: MIT + file LICENSE
Depends: R (≥ 4.2.0)
Imports: lubridate, methods, Rcpp
LinkingTo: Rcpp
Encoding: UTF-8
LazyData: true
RoxygenNote: 7.3.2
URL: https://github.com/imalagaris/RCTRecruit
BugReports: https://github.com/imalagaris/RCTRecruit/issues
Suggests: knitr, magrittr, testthat (≥ 3.0.0), withr
Config/testthat/edition: 3
NeedsCompilation: yes
Packaged: 2025-04-21 18:13:56 UTC; ymala
Author: Ioannis Malagaris ORCID iD [aut, cre, cph], Alejandro Villasante-Tezanos [aut], Christopher Kurinec [aut], Xiaoying Yu [aut]
Maintainer: Ioannis Malagaris <iomalaga@utmb.edu>
Repository: CRAN
Date/Publication: 2025-04-21 18:30:02 UTC

RCTRecruit: Non-Parametric Recruitment Prediction for Randomized Clinical Trials

Description

Accurate prediction of subject recruitment for Randomized Clinical Trials (RCT) remains an ongoing challenge. Many previous prediction models rely on parametric assumptions. We present functions for non-parametric RCT recruitment prediction under several scenarios.

Details

Package: RCTRecruit
Type: Package
Version: 0.2.0
Date: 2025-04-21
License: MIT + file LICENSE
Functions:
GetDistance Euclidean distance between predicted and actual recruitment
GetWeekPredCI Calculate median recruitment with 95% CI for the next 104 weeks (two years)
LoadData Load recruitment data.
plot.RCTPredCI Plots RCTPredCI object
Time2Nsubjects Simulate the number of weeks needed to recruit a given number of subjects
Datasets:
gripsYR1 Daily recruitment data for the 1st year of the GRIPS study
gripsYR2 Daily recruitment data for the 2nd year of the GRIPS study
gripsYR2Weekly Weekly recruitment data for the 2nd year of the GRIPS study

Author(s)

Maintainer: Ioannis Malagaris iomalaga@utmb.edu (ORCID) [copyright holder]

Authors:

References

  1. Villasante-Tezanos A, Kuo Y, Kurinec C, Li Y, Yu X (2024). "A non-parametric approach to predict the recruitment for randomized clinical trials: an example in elderly inpatient settings." BMC medical research methodology, 24, 189. ISSN 1471-2288, https://pubmed.ncbi.nlm.nih.gov/39210285/.

  2. Gajewski BJ, Simon SD, Carlson SE (2008). "Predicting accrual in clinical trials with Bayesian posterior predictive distributions." Statistics in medicine, 27, 2328-40. ISSN 0277-6715, https://pubmed.ncbi.nlm.nih.gov/17979152/.

  3. Jiang Y, Simon S, Mayo MS, Gajewski BJ (2015). "Modeling and validating Bayesian accrual models on clinical data and simulations using adaptive priors." Statistics in medicine, 34, 613-29. ISSN 1097-0258, https://pubmed.ncbi.nlm.nih.gov/25376910/.

See Also

Useful links:

Packages:

Other Links: GetDistance(), GetWeekPredCI(), LoadData(), Time2Nsubjects(), gripsYR1, gripsYR2, gripsYR2Weekly, plot.RCTPredCI()

Euclidean distance between predicted and actual recruitment

Description

Euclidean distance between predicted and actual recruitment

Usage

GetDistance(
  target,
  nSim = 10000L,
  fillGaps = FALSE,
  cauchyWt = FALSE,
  efficiencyFactor = 1
)

Arguments

target

A vector with the actual recruitment by week

nSim

Number of simulations to run (default = 1e4L). Accepted values are in the
range of 1 to 10,000.

fillGaps

Whether to fill recruitment gaps in the data (default = FALSE).
Recruitment gaps are defined as any full week (Monday through Sunday) with
no dates recorded in the loaded data. If at least one date is present within
a given week, that week will not be considered a gap in recruitment.

cauchyWt

Whether to use Cauchy weights for sampling. If FALSE (default),
binomial weights are used.

efficiencyFactor

An efficiency coefficient to apply to the recruitment rate (default = 1).
If the efficiency of the recruitment process is expected to match
the provided data, this value should be set to 1. If the recruitment
process is expected to be slower, this value should less than 1. Finally,
if the recruitment process is expected to proceed faster, this value should be
greater than 1. Accepted values range from 0.1 to 2:

  • 0.1: Indicates that the recruitment rate is expected to be 10% of the original rate.

  • 2.0: Indicates that the recruitment rate is expected to be double the original rate.

Value

An object of RCTDist class with four elements.

  1. dist: A numeric vector with length equal to nSim containing the simulated
    Euclidean distance.

  2. CI: A numeric vector with the median and the 95% CI Euclidean distance.

  3. call.: The call (deparsed) that created the object

  4. cargs: A list with the arguments of the call that created the object
    including the default arguments

See Also

Other Links: GetWeekPredCI(), LoadData(), RCTRecruit-package, Time2Nsubjects(), gripsYR1, gripsYR2, gripsYR2Weekly, plot.RCTPredCI()

Examples

LoadData(gripsYR1, ScreenDt, Enrolled)
(res <- GetDistance(gripsYR2Weekly$enrolled))
str(res)

Calculate median recruitment with 95% CI for the next 104 weeks (two years)

Description

Calculate median recruitment with 95% CI for the next 104 weeks (two years)

Usage

GetWeekPredCI(
  nSim = 10000L,
  fillGaps = FALSE,
  cauchyWt = FALSE,
  efficiencyFactor = 1
)

Arguments

nSim

Number of simulations to run (default = 1e4L). Accepted values are in the
range of 1 to 10,000.

fillGaps

Whether to fill recruitment gaps in the data (default = FALSE).
Recruitment gaps are defined as any full week (Monday through Sunday) with
no dates recorded in the loaded data. If at least one date is present within
a given week, that week will not be considered a gap in recruitment.

cauchyWt

Whether to use Cauchy weights for sampling. If FALSE (default),
binomial weights are used.

efficiencyFactor

An efficiency coefficient to apply to the recruitment rate (default = 1).
If the efficiency of the recruitment process is expected to match
the provided data, this value should be set to 1. If the recruitment
process is expected to be slower, this value should less than 1. Finally,
if the recruitment process is expected to proceed faster, this value should be
greater than 1. Accepted values range from 0.1 to 2:

  • 0.1: Indicates that the recruitment rate is expected to be 10% of the original rate.

  • 2.0: Indicates that the recruitment rate is expected to be double the original rate.

Value

An object of RCTPredCI class with 5 elements.

  1. predCI: An 104x3 matrix with the 2.5%, 50% and 97.5% weekly percentiles

  2. pargs:
    An environment which contains objects and functions used to construct
    the plot with base::plot(). For internal use only.

  3. call.: The call (deparsed) that created the object

  4. cargs: A list with the arguments of the call that created the object
    including the default arguments

See Also

Other Links: GetDistance(), LoadData(), RCTRecruit-package, Time2Nsubjects(), gripsYR1, gripsYR2, gripsYR2Weekly, plot.RCTPredCI()

Examples

LoadData(gripsYR1, ScreenDt, Enrolled)
(res <- GetWeekPredCI(fillGaps = TRUE, efficiencyFactor = 1.5))
scenarios <- list(
  sc1 = GetWeekPredCI(),
  sc2 = GetWeekPredCI(cauchyWt = TRUE),
  sc3 = GetWeekPredCI(fillGaps = TRUE),
  sc4 = GetWeekPredCI(fillGaps = TRUE, efficiencyFactor = 1.2)
)
maxY <- sapply(scenarios, \(x) x$pargs$maxY) |> max()
defaultGraphicParams <- par(no.readonly = TRUE)
graphics::par(mfrow = c(2, 2), cex.main = 1)
for (x in scenarios) plot(x, yMax = maxY, main = x$call.)
do.call(par, defaultGraphicParams)

Load recruitment data.

Description

This function must be called before any other function in this package. LoadData
checks the input data and stores the results internally for the session.
Calling this function more than once in the same session will overwrite the
previously created internal data.

Usage

LoadData(data, date, enrolled)

Arguments

data

Main dataset containing at least two columns:

  • A date column with the calendar date of the screening

  • A enrolled column with the number of subjects enrolled

If the entries cover a period longer than 1 year, only the entries within
one year prior to the latest date will be retained.

date

The name (symbol or string) of the column in the dataset with the calendar
dates of active screening. All active calendar dates should be included, even
if the recruitment for that date is 0. Only dates with truly non-active
recruitment should be omitted. The date column must be:

  • an object inheriting from class the Date class

  • or a character vector with a valid date format.

enrolled

The name (symbol or string) of the column in the dataset with the number of
subjects recruited on the corresponding calendar date. It must be a numeric vector.

Value

This function does not return any value. It runs several tests and proccesses
the data and stores internally the results. It prints a message to the console
if the data is successfully loaded or an error message if there is an issue
with the input data. Once the dataset is loaded, the following functions can
be used:

See Also

Other Links: GetDistance(), GetWeekPredCI(), RCTRecruit-package, Time2Nsubjects(), gripsYR1, gripsYR2, gripsYR2Weekly, plot.RCTPredCI()

Examples

# Load using names as symbols
LoadData(gripsYR1, ScreenDt, Enrolled)

# Load using names as strings
LoadData(gripsYR1, "ScreenDt", "Enrolled")

# Load using base pipe operator
gripsYR1 |> LoadData(ScreenDt, Enrolled)

# Load using magrittr pipe operator
if (base::requireNamespace("magrittr", quietly = TRUE)) {
  library(magrittr)
  gripsYR1 %>% LoadData(ScreenDt, Enrolled) 
}

Simulate the number of weeks needed to recruit a given number of subjects

Description

Simulate the number of weeks needed to recruit a given number of subjects

Usage

Time2Nsubjects(
  nSub = 50L,
  nSim = 10000L,
  fillGaps = FALSE,
  cauchyWt = FALSE,
  efficiencyFactor = 1
)

Arguments

nSub

Number of subjects to recruit (default = 50L)

nSim

Number of simulations to run (default = 1e4L). Accepted values are in the
range of 1 to 10,000.

fillGaps

Whether to fill recruitment gaps in the data (default = FALSE).
Recruitment gaps are defined as any full week (Monday through Sunday) with
no dates recorded in the loaded data. If at least one date is present within
a given week, that week will not be considered a gap in recruitment.

cauchyWt

Whether to use Cauchy weights for sampling. If FALSE (default),
binomial weights are used.

efficiencyFactor

An efficiency coefficient to apply to the recruitment rate (default = 1).
If the efficiency of the recruitment process is expected to match
the provided data, this value should be set to 1. If the recruitment
process is expected to be slower, this value should less than 1. Finally,
if the recruitment process is expected to proceed faster, this value should be
greater than 1. Accepted values range from 0.1 to 2:

  • 0.1: Indicates that the recruitment rate is expected to be 10% of the original rate.

  • 2.0: Indicates that the recruitment rate is expected to be double the original rate.

Value

An object of RCTNWeeks class with four elements.

  1. weeks is an integer vector with length equal to nSim containing the simulation results.

  2. CI shows the median and the 95% CI.

  3. call.: The call (deparsed) that created the object

  4. cargs: A list with the arguments of the call that created the object
    including the default arguments

See Also

Other Links: GetDistance(), GetWeekPredCI(), LoadData(), RCTRecruit-package, gripsYR1, gripsYR2, gripsYR2Weekly, plot.RCTPredCI()

Examples

LoadData(gripsYR1, ScreenDt, Enrolled)
(res <- Time2Nsubjects())
str(res)

Daily recruitment data for the 1st year of the GRIPS study

Description

Modified recruitment data from the first year of the GRIPS study. This dataset includes the number or participants recruited for each calendar date during active recruitment periods.

Usage

gripsYR1

Format

A data frame with 159 observations of 2 variables

[,1] ScreenDt character Calendar date of the screening process for recruitment in the study
[,2] Enrolled integer Number of new subjects enrolled in the study on that date

Source

Villasante-Tezanos A, Kuo Y, Kurinec C, Li Y, Yu X (2024). "A non-parametric approach to predict the recruitment for randomized clinical trials: an example in elderly inpatient settings." BMC medical research methodology, 24, 189. ISSN 1471-2288, https://pubmed.ncbi.nlm.nih.gov/39210285/.

See Also

Other Links: GetDistance(), GetWeekPredCI(), LoadData(), RCTRecruit-package, Time2Nsubjects(), gripsYR2, gripsYR2Weekly, plot.RCTPredCI()

Daily recruitment data for the 2nd year of the GRIPS study

Description

Modified recruitment data from the second year of the GRIPS study. This dataset includes the number or participants recruited for each calendar date during active recruitment periods.

Usage

gripsYR2

Format

A data frame with 292 observations of 2 variables

[,1] ScreenDt character Calendar date of the screening process for recruitment in the study
[,2] Enrolled integer Number of new subjects enrolled in the study on that date

Source

Villasante-Tezanos A, Kuo Y, Kurinec C, Li Y, Yu X (2024). "A non-parametric approach to predict the recruitment for randomized clinical trials: an example in elderly inpatient settings." BMC medical research methodology, 24, 189. ISSN 1471-2288, https://pubmed.ncbi.nlm.nih.gov/39210285/.

See Also

Other Links: GetDistance(), GetWeekPredCI(), LoadData(), RCTRecruit-package, Time2Nsubjects(), gripsYR1, gripsYR2Weekly, plot.RCTPredCI()

Weekly recruitment data for the 2nd year of the GRIPS study

Description

Modified recruitment data from the second year of the GRIPS study, aggregated by calendar week.

Usage

gripsYR2Weekly

Format

A data frame with 52 observations of 4 variables

[,1] week double Calendar week
[,2] year double Calendar year
[,3] enrolled integer Number of people enrolled that week
[,4] activeDays integer Number of days in that week when recruitment was active

Source

Villasante-Tezanos A, Kuo Y, Kurinec C, Li Y, Yu X (2024). "A non-parametric approach to predict the recruitment for randomized clinical trials: an example in elderly inpatient settings." BMC medical research methodology, 24, 189. ISSN 1471-2288, https://pubmed.ncbi.nlm.nih.gov/39210285/.

See Also

Other Links: GetDistance(), GetWeekPredCI(), LoadData(), RCTRecruit-package, Time2Nsubjects(), gripsYR1, gripsYR2, plot.RCTPredCI()

Plots RCTPredCI object

Description

Plots RCTPredCI object

Usage

## S3 method for class 'RCTPredCI'
plot(x, yMax = NULL, includeYR2 = FALSE, xGrid = TRUE, yGrid = TRUE, ...)

Arguments

x

RCTPredCI object

yMax

It sets upper value for the y-axis. It is useful when several figures are plotted together

includeYR2

Whether to plot predictions for 104 weeks. By default (default = FALSE) it only plots the first 52 weeks.

xGrid

Whether to plot vertical grid lines (default = TRUE)

yGrid

Whether to plot horizontal grid lines (default = TRUE)

...

Additional arguments to be passed as in base::plot() function

See Also

Other Links: GetDistance(), GetWeekPredCI(), LoadData(), RCTRecruit-package, Time2Nsubjects(), gripsYR1, gripsYR2, gripsYR2Weekly

mirror server hosted at Truenetwork, Russian Federation.