Date format: yyyy-mm-dd

*File upload format can be found in the 'About' tab

Weibull prior

Gompertz prior

Log-logistic prior

Log-normal prior


Generate Report

Overview

This App can be used to predict a milestone event time (date of certain number of events occurs) for clinical trial monitoring.

Data

  • The Milestone Prediction App supports .csv, .tsv, .txt, .xlsx, .xls, .sas7bdat files.
  • The uploaded data can only contain the following columns in the order they are specified below.
    1. Column 1: Days of patient enrollment since first patient enrolled (i.e. date of patient enrolled minus the date of first patient enrolled). Note, the first patient enrolled for the study would have 0 day.
    2. Column 2: Failure time variable. Typically, this is the time from randomization to the last known status of the event of interest (in days).
    3. Column 3: Event indicator (1=event, 0=censored). There should be at least 1 event in the uploaded dataset.

Disclaimer

Mayo Clinic does not save any of the data or output.

Bayesian Prior

  • Assuming the underlying failure time from the historical control is exponentially distributed with rate parameter λe
  • The prior distributions for parameters taking values on the entire real line is set to be normally
  • distributed and Gamma prior distributions will be used for positive valued parameters.
  • The parametric distributions used for prediction are the following

  • Mean for the prior parameters are defaulted with the following

  • The variance for each of the parameters is defaulted to 10 * max(|Θ|; 1) where Θ is the prior mean of the parameter.
Download User Guide

Purpose of this app

For clinical trials which use time-to-event endpoints, the analysis timing is typically based on the number of events observed. It is important for resource planning to predict the time at which a certain number of events will occur. This app predicts the time to a certain number of events occurs based on the current observed information.

Input dataset

The input dataset needs to conform to the file format. Please see “About” tab in the app for details.

How to use the app

Step 1: Input all the necessary data

Step 2: Upload the observed data

Use the upload function to select the file to use.

See “About” tab for data upload requirements.

After the file is uploaded, the “Data View” tab will display the information uploaded.

Step 3: Perform prediction

After confirming that the dataset is correct, go to the “Calculate Milestone” tab and click the “Run Milestone Prediction”.

The progress of the prediction can be seen in the lower right corner of the monitor.

Step 4: Review the prediction result

  1. The prediction results are shown in this figure. The solid dot indicates the predicted milestone date with the corresponding confidence interval. We provide 14 different predictions based on different underlying distribution and computation methods.
  2. Our recommended method is shown in green.
  3. The predicted results are also shown in the table to provide more detailed information.

A report can be downloaded by clicking the “Generate report” button.

An Example

Suppose there is a study “AAA” which is a randomized phase 3 study designed to detect a target hazard ratio of 0.6. The historical data indicates that the control arm had a 3-year disease-free survival (DFS) rate of 75%. The study design calls for a sample size of 350 patients per arm (700 patients in total) to result in 165 events so the study can be powered at 90% level (type I error = 0.025, 1-sided). Suppose the first patient was enrolled in January 1, 2015 and we would like to predict when the 165th event will occur when we are 3.5 years (1270 days) into the trial.

Here is the setup that I can use for the prediction.

The data is what is expected based on visual inspection.

We can then run the prediction (go to Calculate Milestone" tab then click “Run Milestone Prediction”).

The predication can be carried out fairly quickly (less than 1 min). Based on the recommended method, Bayesian Predictive Synthesis (MSPE), the final analysis is predicted to occur around June 2020 with a confidence interval of (November 2019, January 2021).

DISCLAIMER: The content on the site is NOT medical advice. Although some content may be provided by medical professionals, users acknowledge that access or use of the content does not create a provider-patient relationship and does not constitute medical advice, treatment, diagnosis or services of any kind. The information is provided for educational purposes only and as such is not a substitute for professional medical attention and treatment by medical professionals. Users are solely responsible and accept all liability resulting from use of the content and any related services or products.