Title: | Biomarker Prognostic Enrichment Tool |
---|---|
Description: | Prognostic Enrichment is a clinical trial strategy of evaluating an intervention in a patient population with a higher rate of the unwanted event than the broader patient population (R. Temple (2010) <DOI:10.1038/clpt.2010.233>). A higher event rate translates to a lower sample size for the clinical trial, which can have both practical and ethical advantages. This package is a tool to help evaluate biomarkers for prognostic enrichment of clinical trials. |
Authors: | Jeremy Roth [cre, aut], Kathleen F. Kerr [aut], Kehao Zhu [aut] |
Maintainer: | Jeremy Roth <[email protected]> |
License: | GPL (>= 2) |
Version: | 0.2.2 |
Built: | 2024-11-27 06:28:58 UTC |
Source: | CRAN |
Simulated cohort data containing demographic variables, marker values and cancer outcome.
dcaData
dcaData
A data frame with 500 rows and 6 variables:
Age: Age in years.
Female: Indicator for female gender.
Smokes: Indicator for smoking status.
Marker1: simulated biomarker.
Marker2: simulated biomarker.
Cancer: Indicator for cancer.
Evaluating biomarkers for prognostic enrichment of clinical trials using real data
enrichment_analysis(formula, data, family = binomial(link = logit), reduction.under.treatment, cost.screening = NULL, cost.keeping = NULL, do.bootstrap = TRUE, n.bootstrap = 1000, smooth.roc = FALSE, power = 0.9, alpha = 0.025, alternative = c("one.sided", "two.sided"), selected.biomarker.quantiles = seq(from = 0, to = 0.95, by = 0.05))
enrichment_analysis(formula, data, family = binomial(link = logit), reduction.under.treatment, cost.screening = NULL, cost.keeping = NULL, do.bootstrap = TRUE, n.bootstrap = 1000, smooth.roc = FALSE, power = 0.9, alpha = 0.025, alternative = c("one.sided", "two.sided"), selected.biomarker.quantiles = seq(from = 0, to = 0.95, by = 0.05))
formula |
Object of class "formula", in the form "outcome ~ predictors", where the outcome is a binary indicator with a value of 1 in cases and a value of 0 in controls. |
data |
Data frame containing the outcome and predictors specified in the “formula” argument. Observations with a missing value of the outcome or of any predictor are dropped. |
family |
Character object or call to the family() function specifying the link function that is passed to 'glm' to estimate a risk score when more than one predictor is specified. Defaults to binomial(link = "logit"), which yields logistic regression. |
reduction.under.treatment |
Number between 0 and 1 indicating the percent reduction in event rate under treatment that the trial should be able to detect with the specified power |
cost.screening |
Number indicating the cost of screening a patient to determine trial eligibility, This argument is optional; if both the “cost.screening” and “cost.keeping” arguments are specified, then the total cost of the trial based on each screening threshold is estimated and returned. |
cost.keeping |
Number indicating the cost of retaining a patient in the trial after enrolling. This argument is optional; if both the “cost.screening” and “cost.keeping” arguments are specified, then the total cost of the trial based on each screening threshold is estimated and returned. |
do.bootstrap |
Logical indicating whether bootstrap 95% confidence intervals should be computed. Defaults to TRUE. |
n.bootstrap |
Number of bootstrap samples to draw, if “do.bootstrap” is set to TRUE. Defaults to 1000. |
smooth.roc |
Logical indicating the “smooth” argument passed to the roc() function from the ‘pROC’ package when a single biomarker is given. Defaults to FALSE. |
power |
Number between 0 and 1 giving the power the trial should have to reject the null hypothesis that there is no treatment effect. Defaults to 0.9. |
alpha |
Number between 0 and 1 giving the type I error rate for testing the null hypothesis that there is no treatment effect. Defaults to 0.025. |
alternative |
Character specifying whether the alternative hypothesis is one-sided with a higher event rate in the treatment group (“one.sided”) or two-sided (“two.sided”). Defaults to “one.sided”. |
selected.biomarker.quantiles |
Numeric vector specifying the quantiles of the biomarker measured in controls that will be used to screen trial participants. Defaults to 0, 0.05, ..., 0.95. All entries must be between at least 0 and less than 1. |
A list with components
estimates: A data frame with the following summary measures for each biomarker threshold that is used to screen trial participants: ‘selected.biomarker.quantiles’: quantiles of observed biomarker values used for screening. ‘biomarker.screening.thresholds’: the values of the biomarker corresponding to the quantiles, ‘event.rate’: post-screening event rate, ‘NNS’: The estimated number of patients needed to screen to identify one patient eligible for the trial, ‘SS’: The sample size in a clinical trial enrolling only patients whose biomarker-based disease risk is above the level used for screening, ‘N.screen’: The total number of individuals whose biomarker values are screened to determine whether they should be enrolled in the trial, ‘N.screen.increase.percentage’: Percentage in N.screen relative to a trail that does not based on the biomarker. ‘total.cost’: The estimated total cost of running the trial if the biomarker were used for prognostic enrichment (if cost.screening and cost.keeping are specified), ‘cost.reduction.percentage’: The reduction in total cost relative to a trial that does not screen based on the biomarker.
estimates.min.total.cost: The row of the estimates data frame corresponding the screening strategy that results in the lowest total cost.
bootstrap.CIs: 95% bootstrap-based CIs for reported summary measures (if do.bootstrap=TRUE).
simulation: A logical indicating whether data were simulated.
bootstrap.CIs: 95% bootstrap-based CIs for reported summary measures (if do.bootstrap=TRUE).
biomarker: Biomarker from the given dataset, either the single biomarker specified or the predicted values from logistic regression if multiple biomarkers are specified).
response: Response variable specified in the dataset.
enrichment_simulation
, plot_enrichment_summaries
data(dcaData) ## using a single biomarker in the dataset analysis.single.marker <- enrichment_analysis(Cancer ~ Marker1, data=dcaData, reduction.under.treatment=0.3, cost.screening=100, cost.keeping=1000) head(analysis.single.marker$estimates) head(analysis.single.marker$bootstrap.CIs) ## combining two biomarkers in the dataset analysis.two.markers <- enrichment_analysis(Cancer ~ Marker1 + Marker2, data=dcaData, reduction.under.treatment=0.3, cost.screening=100, cost.keeping=1000) head(analysis.two.markers$estimates) head(analysis.two.markers$bootstrap.CIs)
data(dcaData) ## using a single biomarker in the dataset analysis.single.marker <- enrichment_analysis(Cancer ~ Marker1, data=dcaData, reduction.under.treatment=0.3, cost.screening=100, cost.keeping=1000) head(analysis.single.marker$estimates) head(analysis.single.marker$bootstrap.CIs) ## combining two biomarkers in the dataset analysis.two.markers <- enrichment_analysis(Cancer ~ Marker1 + Marker2, data=dcaData, reduction.under.treatment=0.3, cost.screening=100, cost.keeping=1000) head(analysis.two.markers$estimates) head(analysis.two.markers$bootstrap.CIs)
Evaluating biomarkers for prognostic enrichment of clinical trials using simulated data
enrichment_simulation(baseline.event.rate, reduction.under.treatment, estimated.auc, roc.type = NULL, cost.screening = NULL, cost.keeping = NULL, simulation.sample.size = 5e+05, alternative = c("one.sided", "two.sided"), power = 0.9, alpha = 0.025, selected.biomarker.quantiles = seq(from = 0, to = 95, by = 5))
enrichment_simulation(baseline.event.rate, reduction.under.treatment, estimated.auc, roc.type = NULL, cost.screening = NULL, cost.keeping = NULL, simulation.sample.size = 5e+05, alternative = c("one.sided", "two.sided"), power = 0.9, alpha = 0.025, selected.biomarker.quantiles = seq(from = 0, to = 95, by = 5))
baseline.event.rate |
A number between 0 and 1 indicating the prevalence of the event in the study population. |
reduction.under.treatment |
A number between 0 and 1 indicating the percent reduction in event rate under treatment that the trial should be able to detect with the specified power. |
estimated.auc |
A numeric vector, with each entry between 0.5 and 1, that specifies the AUC for each biomarker to use in simulations. |
roc.type |
A character vector with the same length as the estimated.auc argument. Each entry must be one of "symmetric", "right.shifted", or "left.shifted", which describes the general shape of the ROC curve to use for simulated data. Defaults to "symmetric" for each biomarker. |
cost.screening |
A positive number indicating the cost of screening a patient to determine trial eligibility, This argument is optional; if both cost.screening and cost.keeping are specified, then then the total cost of the trial based on each screening threshold is estimated and returned. |
cost.keeping |
A positive number indicating the cost of retaining a patient in the trial after enrolling. This argument is optional; if both cost.screening and cost.keeping are specified, then then the total cost of the trial based on each screening threshold is estimated and returned. |
simulation.sample.size |
A positive number giving the sample size to use for simulated data. Defaults to 500,000 (to help see trends). |
alternative |
Character specifying whether the alternative hypothesis is one-sided (“one.sided”) with a higher outcome probability in the treatment group or two-sided (“two.sided”). Defaults to “one.sided”. |
power |
Number between 0 and 1 giving the power the trial should have to reject the null hypothesis that there is no treatment effect. Defaults to 0.9. |
alpha |
Number between 0 and 1 giving the type I error rate for testing the null hypothesis that there is no treatment effect. Defaults to 0.025. |
selected.biomarker.quantiles |
Numeric vector specifying the quantiles of the biomarker measured in controls that will be used to screen trial participants. Defaults to 0, 5, ..., 95. All entries must be between at least 0 and less than 001. |
A list with components
estimates: A data frame with the following summary measures for each biomarker threshold that is used to screen trial participants: ‘selected.biomarker.quantiles’: quantiles of observed biomarker values used for screening. ‘biomarker.screening.thresholds’: the values of the biomarker corresponding to the quantiles, ‘event.rate’: post-screening event rate, ‘NNS’: The estimated number of patients needed to screen to identify one patient eligible for the trial, ‘SS’: The sample size in a clinical trial enrolling only patients whose biomarker-based disease risk is above the level used for screening, ‘N.screen’: The total number of individuals whose biomarker values are screened to determine whether they should be enrolled in the trial, ‘N.screen.increase.percentage’: Percentage in N.screen relative to a trail that does not based on the biomarker. ‘total.cost’: The estimated total cost of running the trial if the biomarker were used for prognostic enrichment (if cost.screening and cost.keeping are specified), ‘cost.reduction.percentage’: The reduction in total cost relative to a trial that does not screen based on the biomarker. ‘Biomarker’: label for the biomarker.
simulation: Logical indicating whether data were simulated (always TRUE for the plot_enrichment_summaries
function).
enrichment_analysis
, plot_enrichment_summaries
## three biomarkers with symmetric ROC curves simulation.three.markers <- enrichment_simulation(baseline.event.rate=0.2, reduction.under.treatment=0.3, estimated.auc=c(0.72, 0.82, 0.85), roc.type=c("symmetric", "symmetric", "symmetric"), cost.screening=1, cost.keeping=10, simulation.sample.size=1e+5) head(simulation.three.markers$estimates)
## three biomarkers with symmetric ROC curves simulation.three.markers <- enrichment_simulation(baseline.event.rate=0.2, reduction.under.treatment=0.3, estimated.auc=c(0.72, 0.82, 0.85), roc.type=c("symmetric", "symmetric", "symmetric"), cost.screening=1, cost.keeping=10, simulation.sample.size=1e+5) head(simulation.three.markers$estimates)
Plot summaries of prognostic enrichment of clinical trials estimated by the enrichment_analysis
and enrichment_simulation
functions.
plot_enrichment_summaries(x, text.size.x.axis = 10, text.size.y.axis = 10, text.size.plot.title = 10, text.size.axis.ticks = 10, annotate.no.screening.cost = FALSE, smooth.roc = TRUE)
plot_enrichment_summaries(x, text.size.x.axis = 10, text.size.y.axis = 10, text.size.plot.title = 10, text.size.axis.ticks = 10, annotate.no.screening.cost = FALSE, smooth.roc = TRUE)
x |
Object returned by either the |
text.size.x.axis |
Size of text for the x-axis of plots. Defaults to 10. |
text.size.y.axis |
Size of text for the y-axis of plots. Defaults to 10. |
text.size.plot.title |
Size of text for the plot titles. Defaults to 10. |
text.size.axis.ticks |
Size of axis tick marks for plots. Defaults to 10. |
annotate.no.screening.cost |
Logical indicating whether to annotate the relative to total cost curve at the point where no biomarker screening occurs. Defaults to FALSE |
smooth.roc |
Logical indicating whether the ROC curves (plotting with the roc() function in the ‘pROC’ package) should be smoothed. Defaults to TRUE. |
A grid of either 4 or 6 plots, summarizing the results of either the enrichment_analysis
or the enrichment_simulation
function.
enrichment_analysis
, enrichment_simulation
data(dcaData) # one marker analysis.single.marker <- enrichment_analysis(Cancer ~ Marker1, data=dcaData, reduction.under.treatment=0.3, cost.screening=100, cost.keeping=1000) plot_enrichment_summaries(analysis.single.marker) # two markers analysis.two.markers <- enrichment_analysis(Cancer ~ Marker1 + Marker2, data=dcaData, reduction.under.treatment=0.3, cost.screening=100, cost.keeping=1000) plot_enrichment_summaries(analysis.two.markers)
data(dcaData) # one marker analysis.single.marker <- enrichment_analysis(Cancer ~ Marker1, data=dcaData, reduction.under.treatment=0.3, cost.screening=100, cost.keeping=1000) plot_enrichment_summaries(analysis.single.marker) # two markers analysis.two.markers <- enrichment_analysis(Cancer ~ Marker1 + Marker2, data=dcaData, reduction.under.treatment=0.3, cost.screening=100, cost.keeping=1000) plot_enrichment_summaries(analysis.two.markers)