CureCheck: An R Shiny Application for Evaluating Cure Model Appropriateness — KIDS26 Team 9

CureCheck is a guided, point-and-click Shiny app for running the cureAssess R package. Deciding whether a cure model is appropriate for a survival dataset currently requires writing R code and knowing the cure-model literature, and that gate keeps the method away from the clinicians and analysts who need it. The app wraps the published, peer-review-backed cureAssess workflow so that a non-statistician can go from "here is my dataset" to a plain-language verdict plus a downloadable HTML report — without writing a line of R. This repository contains code for a functional app prototype created during the St. Jude BioHackathon 2026, Wednesday 16 – Friday 18 September 2026.

Under construction: adding a guided manual walking through using the Shiny app.

Project Profile

Field	Detail
Project name	CureCheck — an interactive Shiny app for the cureAssess R package
Question / problem	To make checking cure model appropriateness easily accessible. Checking that a cure model is appropriate is an important step to ensure that the model used is valid for exploration and inference.
Data and inputs	CSV format with columns for time and event status.
App capabilities	The app allows users to explore qualitative and quantitative metrics to asses cure model appropriateness, using published literature and methods.
Expected package dependencies	shiny, bslib, DT, rmarkdown, survival, flexsurv, flexsurvcure, survminer, ggplot2, dplyr.
Current deployment	Local only — shiny::runApp() R/RStudio. A future improvement is to create a publicly available web app.
Team lead	Durbadal Ghosh, PhD (@Durbadal0)
Team members	Geethanjalee Mudunkotuwa, PhD (@GeethanjaleeM); Sharon Freshour, PhD; Rashid Mehmood, PhD

Why cure model appropriateness matters

Ordinary survival analysis assumes everyone eventually has the event. Modern therapy broke that assumption: some patients are effectively cured and never relapse. A mixture cure model splits the population in two, S(t) = (1 − p) + p · Su(t), where (1 − p) is the cure fraction and Su(t) is survival among the uncured. On a Kaplan-Meier curve, a cure fraction looks like a plateau that flattens out above zero.

That model carries two extra assumptions: a non-zero cured fraction exists, and follow-up runs long enough to identify it (formally, τ_F0 ≤ τ_G — follow-up must extend past the time by which all susceptible patients would have had the event). If the second fails, late censored patients could be cured or uncured-and-not-yet-relapsed, and the two are observationally indistinguishable. Othus et al. re-analysed six SWOG trials at two follow-up times and found cure-model estimates of mean survival shifted materially — and the direction of the shift was not predictable, so it cannot be corrected after the fact.

A plateau is not proof. Heavy censoring alone manufactures a plateau that is a censoring artifact, not cure. That is exactly what the diagnostics are for. The manuscript's three-step check is ① expert judgment, ② visual assessment, ③ quantitative assessment. cureAssess automates ② and ③. Step ① is a conversation with a clinician, so the app must present it as something the user confirms, never something the software decides.

What the app will do

Under construction: Add screenshot of tabs from the app itself. Briefly explain.

Repository layout

KIDS26-Team9/
├── README.md                  
├── LICENSE.md                 
├── app/                       
│   ├── app.R                  
│   └── R/              
├── cureAssess/
├── data-raw/              
├── data/examples/             
├── docs/                      
└── report/                    

Getting started

The user can clone the repo locally to their desired location.

git clone https://github.com/stjude-biohackathon/KIDS26-Team9.git
cd KIDS26-Team9

# 1. R >= 4.5.3 is required.
R.version.string

# 2. App and package dependencies.
install.packages(c(
  "shiny", "bslib", "DT", "rmarkdown",
  "survival", "flexsurv", "flexsurvcure", "survminer", "ggplot2", "dplyr"
))

# 3. Install the vendored cureAssess package (version 0.1.0).
install.packages("cureAssess/cureAssess_0.1.0.tar.gz", repos = NULL, type = "source")
# ...or, for development without installing:
# devtools::load_all("cureAssess")

# 4.  Run the app.
shiny::runApp("app")

About the cureAssess package

cureAssess 0.1.0 — Assessing Cure Model Appropriateness for Survival Data. Author, maintainer and copyright holder: Geethanjalee Mudunkotuwa. Author: Durbadal Ghosh. MIT licensed. Upstream: https://github.com/GeethanjaleeM/cureAssess.

A companion tutorial manuscript, "A Tutorial for Evaluating Cure Model Appropriateness" (Mudunkotuwa, Ghosh, Triplett, Selukar), is in preparation and available on arXiv. The app makes the workflow described in Figure 1 of this manuscript into an interactive workflow.

References

• Maller RA, Zhou S (1992). Estimating the proportion of immunes in a censored sample. Biometrika, 79(4), 731-739. doi:10.1093/biomet/79.4.731
• Maller RA, Zhou S (1994). Testing for sufficient follow-up and outliers in survival data. Journal of the American Statistical Association, 89(428), 1499-1506. doi:10.1080/01621459.1994.10476889
• Shen P-S (2000). Testing for sufficient follow-up in survival data. Statistics & Probability Letters, 49(4), 313-322. doi:10.1016/S0167-7152(00)00063-8
• Selukar S, Othus M (2023). RECeUS: Ratio estimation of censored uncured subjects, a different approach for assessing cure model appropriateness in studies with long-term survivors. Statistics in Medicine, 42(3), 209-227. doi:10.1002/sim.9610

Data and ethics note

Public and simulated data only. Nothing restricted, nothing identifiable, and nothing from St. Jude clinical datasets enters this repository or appears on screen during the demo.