New-Onset Clinical Trial Simulation Tool
This quantitative clinical trial simulation tool helps optimize clinical trial design of therapies for new-onset T1D patients using C-peptide, HbA1C and insulin.
General information about the C-peptide clinical trial simulation tool, please read and acknowledge below prior to accessing the tool.
For a DAP guide and help with account setup, click here.
A quantitative clinical trial simulation tool to help optimize clinical trial design of therapies for new-onset T1D patients using an univariate model of C-peptide.
Important: while this model has been submitted to the FDA’s fit-for-purpose pathway for review and comment, it has not yet been endorsed by the FDA for any purpose.
Enrichment strategies for clinical trial populations can dramatically increase the power of a trial, relative to similarly designed unenriched trials, potentially facilitating smaller, shorter, and more efficient trials.
A cloud-based graphical user interface (GUI) has been developed, which allows a user-friendly experience for scientists and clinicians of all backgrounds to perform simulations based on the model.
A model-based clinical trial simulation tool describing the progression of new-onset T1D centered on a drug-disease-trial model that measures C-peptide AUC during the 2-hour MMTT to predict C-peptide decline over time. Models for this measure incorporate relevant sources of variability, including age at diagnosis, length of time since diagnosis, sex, BMI, and baseline values for C-peptide as measured through 2-hour MMTT.
T1DC has analyzed participant-level placebo data from 23 studies that have enrolled participants with new-onset T1D. One study (SEARCH) is an observational study, 21 studies are clinical trials or intervention studies, and one study (TrialNet-16) assesses longitudinal data from patients enrolled in several other studies.
A summary of key points for the model are described below:
- A C-peptide model was developed using the 2-hour C-peptide AUC from the mixed meal tolerance test
- Data were log-transformed to prevent negative predictions
- A sigmoidal Emax model best described the trajectory of C-peptide progression
- Significant covariates that are reflected in the clinical trial simulation tool include baseline BMI Z-score, age and C-peptide
- Visual predictive checks (VPCs), with a training dataset utilizing 80% of the data and a validation dataset using the other 20% of the data, demonstrate good fit with predicted and observed data falling well within the 95% confidence intervals
- Explore population-level trial enrichment approaches using C-peptide as an endpoint.
- Determination of potential appropriate trial durations based on pre-specified trial population
Note: the GUI is a pilot version based on a model subject to revision based on comments from the pending FDA review of T1DC’s fit-for-purpose submission. You can read more about the fit-for-purpose initiative here.
Insulin and HbA1c Univariate Model Tabs are included for users’ interest; however, these models will not be submitted to FDA. Given the interdependence of these variables caution with interpretation of results based on these univariate models is advised. These parameters will be included in a multi-variate model currently under development for a future FDA submission.
A summary of key points for the HbA1c and Insulin models is described below:
HbA1c Placebo Model:
- HbA1c data were log-transformed.
- The model was trained and validated using an 80/20 data split.
- Stepwise covariate model building was used to test the following covariates: baseline (HbA1c, age, BMI, disease duration), sex, race, ethnicity, and HLA genotypes.
- HbA1c was best described by a sigmoidal Emax equation with an exponential function to describe the honeymoon phase dip in HbA1c.
- Baseline age was the only predictor of HbA1c.
- Model performance was guided by standard goodness-of-fit measures and visual predictive checks. Visual predictive checks on the training and validation dataset showed good performance
Insulin Placebo Model:
- Weight-adjusted total daily insulin values were modeled as averages (3-7 days) and square root-transformed.
- The model was trained and validated using an 80/20 data split.
- Stepwise covariate model building was used to test the following covariates: baseline (insulin, age, BMI, disease duration), sex, race, ethnicity, and HLA genotypes.
- Insulin use was best described by a quadratic function with additive inter-individual and residual unexplained variability.
- Baseline age, disease duration and HbA1c were predictors on exogenous insulin use.
- Model performance was guided by standard goodness-of-fit measures and visual predictive checks. Visual predictive checks on the training and validation dataset showed good performance.
To request access to the CTS tool, click here.
View a recording of the tool’s debut webinar here.
User feedback is invited. Contact t1dcadmin@c-path.org for comments or more information.
*Please note, you will have access to the clinical trial simulation tool in order to query the data, however you will not have direct access to the data.
The CTST has not yet been reviewed by FDA and has therefore not been accepted by the FDA for use in assessing clinical endpoints in human clinical trials. If the calculated number of patients is large (e.g., >5,000), the waiting time may be long. Please consider adjusting the simulation settings.
T1DC would like to acknowledge:
The T1DC would like to thank all those who have contributed data to TOMI-T1D. This pilot tool was prepared by C-Path’s T1DC and does not necessarily reflect the opinion or views of any individual organization that contributed data.
