General Information
Conflict of Interest
Request Clinical Trials
Associated Trial(s): What type of data are you looking for?: Individual Participant-Level Data, which includes Full CSR and all supporting documentationRequest Clinical Trials
Data Request Status
Status: OngoingResearch Proposal
Project Title: A Study on Changes in Clinical Risk Associated with Disease Progression in Type 2 Diabetes Patients Using the CANVAS Test
Scientific Abstract:
Background
Numerous studies have been conducted to prevent cardiovascular events in patients with T2DM. However, clinical trial analyses are based on the trial period, meaning subjects with differing durations since onset are analyzed on the same timeline.
If hyperglycemia damages the vascular endothelium, thereby increasing cardiovascular risk, then the disease time--that is, the period during which the patient's vascular endothelium has been damaged--is expected to be an important factor in predicting cardiovascular risk. However, strictly controlling the onset of diabetes is difficult, and considering the disease time is not straightforward.
Objective
By employing analytical methods such as SReFT to conduct risk assessments that account for each individual disease time, we aim to establish a novel method for evaluating clinical risk in patients with T2DM.
Study Design
The objective is to estimate the time to disease onset in patients, performing analysis using SReFT with the acquired data. Clinical risk is evaluated by conducting Cox analyses adjusting the disease time.
Participants
all participants (a total of 4,330 indivisuals)
Primary and Secondary Outcome Measure(s)
For model construction of SReFT analysis: HbA1c, Blood pressure, Fasting lipids, renal functions, beta cell functions, QOL scores,
For estimating event risk with disease time: (primary)MACE, renal functions, (secondary)beta cell functions
Statistical Analysis
Data processing and analysis: R, Rstudio, SAS
Analysis of SReFT: NONMEM (ICON plc)
Analysis of Machine Learning: Open-source libraries
Brief Project Background and Statement of Project Significance:
It is known that type 2 diabetes patients have a 2- to 3-fold increased risk of cardiovascular events such as myocardial infarction and stroke due to prolonged exposure of the vascular endothelium to hyperglycemia. Since these clinical risks, including cardiovascular complications, significantly impact patient prognosis, numerous studies have been conducted to prevent them. However, these analyses were based on the trial period. Although age was an inclusion criterion for patients, there was no criterion for duration of disease. Consequently, subjects with differing lengths of time since diabetes onset were analyzed on the same timeline. If hyperglycemia damages the vascular endothelium, thereby increasing cardiovascular risk, then the duration of disease--that is, the period during which the patient's vascular endothelium has been damaged--is expected to be an important factor in predicting cardiovascular risk. Indeed, the median duration of diabetes among subjects enrolled in the CANVAS trial was approximately 14 years, suggesting a strong potential for such effects. However, strictly controlling the onset of diabetes is difficult, and considering the individual duration of diabetes for each person is not straightforward.
Therefore, we focused on SReFT (Statistical estimation of fragmented time course), a method for estimating the long-term course of chronic diseases. This novel method, developed by one of the researchers, Hisaka et al., estimates disease progression over decades by calculating each individual's disease time (the time elapsed since true disease onset) from subject-specific data from clinical trials lasting less than ten years. SReFT has been applied to analyze Alzheimer's disease, COPD, and Parkinson's disease, yielding excellent results in each case [1,2,4]. Recent analyses also suggest potential common disease progression pathways between diabetes and heart failure. Therefore, we focused on the CANVAS trial as a clinical study that includes a cohort of subjects with a history of both diseases. By confirming the reproducibility obtained from previous SReFT analyses and advancing the identification of new common disease axes, we believe it will become possible to individually assess clinical risks for cardiovascular diseases.
Specific Aims of the Project:
The objective of this study is to establish a method for evaluating clinical risks associated with disease progression in patients with type 2 diabetes, particularly cardiovascular event risk such as heart failure, from a novel perspective that accounts for disease duration. Conventional CANVAS trial analyses have assessed risk based on the observation period following trial enrollment. However, the time elapsed since diabetes onset (true disease duration) varies significantly among subjects, potentially failing to adequately reflect disease progression and the accumulation of vascular endothelial damage.
This study employs SReFT (Statistical estimation of fragmented time course) to estimate each patient's disease time from longitudinal biomarker data such as HbA1c, blood pressure, and QOL indicators. This estimated disease duration is then used as a correction factor in analyses such as Cox proportional hazards models. Specific hypotheses to be tested include: "Accounting for estimated disease time allows more accurate explanation and prediction of cardiovascular event risk compared to conventional analyses," "Patients with longer disease time exhibit a stepwise increase in clinical risk," and "The cardiovascular risk-reducing effect of canagliflozin may vary depending on the expected disease time." This aims to develop a risk assessment model for diabetes patients that will contribute to personalized medicine.
Study Design: Individual trial analysis
What is the purpose of the analysis being proposed? Please select all that apply.: New research question to examine treatment effectiveness on secondary endpoints and/or within subgroup populations Develop or refine statistical methods Research on clinical prediction or risk prediction
Software Used: Python, R, RStudio
Data Source and Inclusion/Exclusion Criteria to be used to define the patient sample for your study:
Explicit inclusion and exclusion criteria: None
Use of data other than those requested through the YODA Project: None
Primary and Secondary Outcome Measure(s) and how they will be categorized/defined for your study:
Primary Outcome Measure:
The primary endpoint of this study is the risk of cardiovascular events (such as myocardial infarction and stroke) in patients with type 2 diabetes. These events align with the major clinical risk factors defined in the CANVAS trial.
Definition and Categorization Method:
These outcomes adopt a time axis based on "disease time (time elapsed since true disease onset)" and analyze and evaluate changes in event occurrence risk over this time period using the Cox proportional hazards model.
Secondary Outcome Measure(s):
We will also secondarily examine "clinical risks" beyond major cardiovascular risks. This includes other important clinical complications collected within the trial and disease progression based on biomarker trends.
Categorization/Definition:
Similar to the primary outcome, the secondary outcomes will be adjusted for disease time estimated by SReFT, and risks will be analyzed and evaluated for each stage of disease progression. Furthermore, by including the interaction between the treatment intervention (canagliflozin) and disease time in the analysis, differences in treatment effects according to the stage of the disease will be clarified.
Main Predictor/Independent Variable and how it will be categorized/defined for your study:
The primary independent variables in this study are the "Estimated Disease Time" calculated using the novel analytical method SReFT and the "treatment intervention" in the CANVAS trial.
First, we define "Estimated Disease Time (EDT)". The cardiovascular risk in type 2 diabetes is primarily determined by the time elapsed since the onset of the actual disease. Since subjects in the CANVAS trial had diverse disease durations (median approximately 14 years), it is necessary to uniformly assess each subject's stage of disease progression rather than using the trial start date as the baseline. This study employs the SReFT method developed by Hisaka et al. To evaluate the effects of canagliflozin treatment on the biomarkers assessed in the CANVAS trial: reduction in major adverse cardiovascular events (MACE), albuminuria (based on urinary albumin/creatinine ratio), and β-cell function (based on the homeostasis model assessment). Additionally, data such as HbA1c, blood pressure, fasting lipids, and QOL scores were analyzed using SReFT to estimate the EDT for each subject. EDT is treated as a continuous variable and serves as the reference axis for capturing disease progression over decades.
Next, "Treatment Intervention" is defined as a binary categorical variable, classified into the canagliflozin treatment group or the placebo group based on randomization in the CANVAS trial.
These variables are clearly defined and calculated/managed within the YODA platform. They are presented in a format directly comparable to the final academic paper report, ensuring transparency and reproducibility of the analysis.
Other Variables of Interest that will be used in your analysis and how they will be categorized/defined for your study:
In this study, the "other variables of interest" used besides the primary independent variables (EDT and treatment intervention) are the biomarkers and background factors explicitly stated in the source as follows.
First, they are the longitudinal biomarker sets used as input values in the SReFT analysis. These include the reduction in major adverse cardiovascular events (MACE) with canagliflozin treatment evaluated in the CANVAS trial, its effects on albuminuria (based on urinary albumin-to-creatinine ratio) and beta-cell function (based on the homeostasis model assessment), as well as HbA1c, blood pressure, fasting lipids, and QOL scores. These variables are treated as continuous variables and defined as foundational data for estimating each subject's EDT from fragmented clinical data that varies per subject.
Second is "age" as a subject background factor. The "age" explicitly stated as an inclusion criterion in the CANVAS trial is considered background information for assessing disease progression and clinical risk.
These variables are utilized for EDT via SReFT and as adjustment factors in subsequent Cox analyses. They may also be adopted as variables to enhance predictive accuracy in machine learning algorithms such as neural networks and random forests, which are being explored as non-parametric analysis methods.
Statistical Analysis Plan:
This study quantitatively evaluates changes in clinical risk associated with disease progression in type 2 diabetes using individual patient data (IPD) from 4,330 subjects in the CANVAS trial (NCT01032629). Analyses combine survival time analysis (Cox proportional hazards model) with advanced nonlinear analysis (SReFT and machine learning).
The uniqueness of this study lies in using the "EDT" for each subject--the time elapsed since true disease onset--as the reference axis for analysis, rather than time since trial initiation. Analysis will employ the SReFT method developed by Hisaka et al. Input variables include longitudinal biomarker data such as HbA1c, blood pressure, fasting lipids, renal function (eGFR, urinary albumin/creatinine ratio, etc.), β-cell function (HOMA-B, proinsulin/insulin ratio, etc.), and QOL indicators (information recorded by subjects themselves in the Subject diary). These fragmented biomarker trends are integrated to identify the stage of the disease progression process, spanning decades, for each subject, and to calculate the "EDT". This analysis primarily uses NONMEM software.
Subsequently, a Cox proportional hazards model is constructed to assess cardiovascular event risk, using the estimated "EDT" as the reference axis. In the model specification, the primary independent variables are "EDT" and "treatment intervention (canagliflozin vs. placebo)". Including an interaction term between "EDT" and "treatment intervention" in the model allows us to examine how the cardiovascular event suppression effect of canagliflozin varies depending on the stage of diabetes progression (length of disease time).
Additionally, to confirm model robustness and capture more complex nonlinear relationships, we will incorporate machine learning (ML) methods that do not assume specific models. Using neural networks and random forests, along with libraries such as Python's scikit-learn, we will conduct multifaceted analyses of the impact of EDT and other covariates on clinical risk.
All analyses will be conducted within the secure remote platform provided by the YODA project. The software to be used includes R, RStudio, SAS, NONMEM, and Python (scikit-learn).
Narrative Summary: This study applies patient-specific data from the CANVAS trial using SReFT, a model analysis, to predict long-term disease progression in diabetic patients. SReFT is a form of time alignment technology that can statistically link trends in multiple individual biomarkers over short observation periods to reconstruct long-term trajectories. The CANVAS trial evaluated the reduction in major adverse cardiovascular events (MACE) with canagliflozin treatment, as well as its effects on albuminuria and beta-cell function, according to the duration of trial observation. We will perform SReFT analysis using the acquired data to estimate patient disease duration, conduct event risk analysis, and employed Cox analysis including interactions to predict long-term progression of diabetes and assess clinical risks, like individual cardiovascular risk.
Project Timeline:
Expected Project Start Date: Data Request Approval Date
Analysis Completion Date: March 2028
Manuscript Preparation and Initial Publication Submission Date: September 2028
Date to Report Results to the YODA Project: March 2029
Dissemination Plan: The findings from this research will be presented at domestic and international conferences in the fields of diabetes and clinical pharmacology. Additionally, they will be published as academic papers in English-language journals within the same field (Clinical Pharmacology & Therapeutics, British Journalof Clinincal Pharmacolofy, Journal of Pharmacological Sciences, etc.)
Bibliography:
[1] T. Ishida, K. Tokuda, A. Hisaka, M. Honma, S. Kijima, H. Takatoku, T. Iwatsubo, T. Moritoyo, and H. Suzuki, Clin. Pharmacol. Ther. 105, 436 (2019).
[2] S. Kawamatsu, R. Jin, S. Araki, H. Yoshioka, H. Sato, Y. Sato, and A. Hisaka, J. Clin. Med. 9, 2676 (2020).
[3] Jin R, Yoshioka H, Sato H, Hisaka A. Data-driven disease progression model of Parkinson’s Disease and effect of sex and genetic variants. CPT: Pharmacometrics & Systems Pharmacology, 13(4):649-659, 2024.
[4] Sano Y, Jin R, Yoshioka H, Nakazato Y, Sato H, Hisaka A. Clinical & Translational Science, 18(10):e70351 (2025).
[5] Yoshioka H, Jin R, Hisaka A, Suzuki H. Pharmacol Ther. 108655 (2024).