In ddinart/GUIP1: Graphical User Interface for Phase I Designs

Graphical user interface for phase I R package GUIP1

D. Dinart(1), J. Fraisse(2), D. Tosi(2), A. Mauguen(3), Y. Laghzali(2), C. Touraine(2), M.C. Le Deley(4), C. Bellera(1), C. Mollevi(2)
(1) Inserm CIC1401, Module Epidemiologie clinique, Institut Bergonie, Bordeaux, France
(2) Institut du Cancer Montpellier (ICM), Montpellier, France
(3) Memorial Sloan Kettering Cancer Center, New York, USA
(4) Centre Oscar Lambret, Lille, France

The first objective of GUIP1 package is to simplify the use of model-based designs used in phase I trial.

GUIP1 package allows the user to:

• Test the operating characteristics of the underlying dose-toxicity model
• Manage in real-time a phase I trial

Package dependencies

To use GUIP1 package the following tools have to be installed:

• R packages: dfcrm, bcrm, tcltk, tcltk2, tkrplot, gridExtra, sp, stringr, openxlsx, magrittr, ggplot2, rjags, R2WinBUGS
• Just Another Gibbs Sampler JAGS

Implemented designs

Currently, five model-guided adaptive (MGA) designs are implemented in GUIP1 package:

• Continual Reassessment Method with Bayes estimate CRMB
• Continual Reassessment Method with maximum likelihood estimate CRML
• TIme-To-Event Continual Reassessment Method TITE-CRM)
• Escalation With Overdose Control EWOC
• TIme-To-Event Escalation With Overdose Control TITE-EWOC

GUIP1 application

library(GUIP1)

### Launch GUIP1
GUIP1()
#select one of the 5 MGA designs

Once the model selected, the user is invited to choose between either the study of the operating characteristics of the underlying model via simulations or the management of a trial.

To do this, it's needed to click either on interactive option or simulator option. In the following, an example with CRMB design is provided.

Simulation

Input parameters

To run simulations, all fields must be completed. The following input parameters are required:

• Number of dose levels
• Target DLT rate
• Sample size
• Number of simulations
• Starting dose level
• Seed

• Model:

  • Logistic or Empiric for CRMB, CRML and TITE-CRM
  • Hyperbolic tangent, logistic, power or 2-parameter logistic for EWOC
  • Hyperbolic tangent, logistic, power for TITE-EWOC
  • Parameters related to model if required

  • Uncertainty prior distribution:

  • Only normal distribution with mean=0 is available for CRMB, CRML and TITE-CRM

  • Gamma, uniform, Lognormal and Bivariate Lognormal distributions are available for EWOC
  • Gamma and uniform distribtuions are available for TITE-EWOC
  • Prior alpha value: First value of the uncertainty variable (if required)
  • Dose skipping constraint: To apply restrictions on dose skipping level
  • Observation window: Periode where DLT is evaluated (if applicable)
  • Pointest: Parameter to controle overdose escalation (if applicable)
  • Study name
  • Probabilities of toxicity:
  • True: probabilities theoretically observed
  • Prior: Guess probabilies

Outputs

Once the simulations run, it's possible to visualize the output in the Results tab.

The user can retrieve:

• A summary of the entered inputs
• the MTD distribution (recommendation percentage of each dose level as being the MTD)
• the average number of patients treated by dose level
• the average number of toxicities observed by dose level

All of these results can be either export in an excel file or being saved into R file.

Trial management

Input parameters

Fig5. and Fig6. are dedicated to input parameters needed to manage a trial. Overall, the same parameters than those presented in simulation section are required except those related to simulation (seed etc.). Depending on the design chosen, all parameters required can be displayed on the same tab. The following input parameters are required:

• Number of dose levels
• Target DLT rate
• Halfwidth parameter (if available): A help to define the interval between the prior probabilities

• Model:

  • Logistic or Empiric for CRMB, CRML and TITE-CRM
  • Hyperbolic tangent, logistic, power or 2-parameter logistic for EWOC
  • Hyperbolic tangent, logistic, power for TITE-EWOC
  • Parameters related to model if required
  • Prior guess of MTD (dose level)

  • Uncertainty prior distribution:

  • Only normal distribution with mean=0 is available for CRMB, CRML and TITE-CRM

  • Gamma, uniform, Lognormal and Bivariate Lognormal distributions are available for EWOC
  • Gamma and uniform distribtuions are available for TITE-EWOC
  • Observation window: Periode where DLT is evaluated (if applicable)
  • Pointest: Parameter to controle overdose escalation (if applicable)
  • Dose skipping constraint: To apply restrictions on dose skipping level (if available)
  • Stopping rules:
  • Maximun number of patients to be included in the trial
  • Maximum number of patient to be entered at a dose level (if available)
  • Number of patients related to the stopping rules
  • Study name
  • Prior probabilities of toxicity (may be completed automatically with some designs but the user is free to change the estimates)

How to include a patient

The Include tab contains a short summary of trial inputs and the current state of inclusions and recommendations.

The dose level recommended for the next patient corresponds to dose level displayed in estimated MTD. This is the dose level estimated as being closest to the targeted DLT rate. From this tab, the user can either include a patient with New patient button or complete information for a patient with Pending patient button or modify information already entered with Modif data patient button. The last 2 buttons are not available for all designs. Basically, Modif data patient button is reserved for TIme-To-Event designs.

To include a patient it's needed to click on the New patient button displayed in Include tab and entered information from previous patient. The information required are:

• Dose level assigned
• Toxicity response (DLT y/n)
• Pending allows to include a patient even with a unknown toxicity response (if available)
• Follow-up time (if applicable)
• Delay time (if applicable)

Pending patient or patient with evolving status

The two options Pending patient and Modif data patient work the same way. The user is invited to select the patient concerned and in the first case provide the toxicity response (Y/N). In the second case, the user can modify (update) all information provided about a patient. This option is particularly handy for TITE designs where patients entered in a staggered fashion.

Result outputs

Results tab is visible from the first inclusion. In this tab, the user will be able to visualize outputs at patient and dose level, as well as graphics describing dose-toxicity and patient-toxicity relastionship and export his results in excel and pdf files.

By clicking on Patient summary results a table summarizing the data entered for each patient:

• ID: order of arrival
• Presence of toxicity (y/n)
• Dose level assigned
• Follow-up time (if applicable)
• Delay time (if applicable)

Depending on the selected design, the information provided by Dose level summary results are quite different, not all of the following information are displayed. The following is given for each dose level:

• Prior probabilities of toxicity
• Number of patient treated
• Number of toxicities
• Posterior probabilities of toxicity
• Posterior mean probabilities of toxicity
• Posterior probability quantiles of toxicity
• Confidence interval of posterior probabilities of toxicity

Fig 13. is a graphic describing the relationship between the patient number and the presence or not of toxicity. Presence of toxicity is represented by a red circle.

Fig 14., Fig 15. and Fig 16. are graphics describing dose-toxiciy relationship either with the help of curves and confidence intervals or boxplots. The graphics proposed vary with the selected design.

references

O'Quigley, J., Pepe, M., and Fisher, L. (1990). Continual reassessment method: a practical design for phase 1 clinical trials in cancer. Biometrics 46, 33-48.

O'Quigley, J., and Shen, L.Z. (1996). Continual reassessment method: a likelihood approach. Biometrics 52, 673-684.

Babb, J., Rogatko, A., and Zacks, S. (1998). Cancer phase I clinical trials: efficient dose escalation with overdose control. Stat. Med. 17, 1103-1120.

Cheung, Y.K., and Chappell, R. (2000). Sequential designs for phase I clinical trials with late-onset toxicities. Biometrics 56, 1177-1182.

Mauguen, A., Le Deley, M.C., and Zohar, S. (2011). Dose-finding approach for dose escalation with overdose control considering incomplete observations. Stat. Med. 30, 1584-1594.

ddinart/GUIP1 documentation built on Nov. 4, 2019, 10:21 a.m.