This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
EDIC is a long-term study in people with type 1 diabetes mellitus. The registry describes the current 2022–2027 cycle as an assessment of advanced microvascular complications, major cardiovascular disease events or early subclinical disease, changes in physical and cognitive dysfunction, liver fibrosis and steatosis, and obstructive sleep apnea.
| Feature | EDIC |
|---|---|
| Study acronym | EDIC |
| Full title | Epidemiology of Diabetes Interventions and Complications (EDIC) |
| Condition | Type 1 Diabetes Mellitus |
| Status | Active, not recruiting |
| Enrollment | 1,441 |
| Study start | 1994-04 |
| Primary completion date | 2027-06 |
| Lead sponsor | George Washington University |
| Sponsor type | Other |
| ClinicalTrials.gov | NCT00360893 |
2. Clinical Question
The current registry-defined cycle asks a broad longitudinal clinical question: how do advanced complications and functional outcomes associated with type 1 diabetes evolve across several organ systems and domains during the 2022–2027 assessment period?
Population
Participants in the EDIC study with type 1 diabetes mellitus. The registry reports enrollment of 1,441.
Study focus
Longitudinal assessment of microvascular complications, cardiovascular outcomes, physical and cognitive function, liver disease, and sleep.
Assessment cycle
The registered primary endpoint is the 2022–2027 Cycle, with different assessment schedules across the endpoint domains.
Primary question
How do advanced complications and functional outcomes identified in the registry change or occur during the specified assessment periods?
3. Trial Design
EDIC is a long-running observational and longitudinal assessment program in people with type 1 diabetes mellitus as represented by the current registry record. Its current status is active, not recruiting. The study began in April 1994 and has a primary completion date of June 2027.
4. Endpoints
The ClinicalTrials.gov record identifies one registered primary endpoint for the current cycle: the 2022–2027 Cycle. Its description covers several clinically distinct outcome domains, with different time frames for assessment.
| Endpoint domain | Registry time frame | Registry description |
|---|---|---|
| Eyes | Yrs 32-33, 2025-26 | Advanced microvascular complications including retinopathy. |
| Kidney | Annually | Advanced microvascular complications including kidney disease. |
| Nerves | Yrs 29-31, 33-34, 2022-26 | Advanced microvascular complications including neuropathy. |
| CVD events | Continuous, adjudicated annually | Major CVD events or early subclinical cardiovascular disease. |
| Physical | Yrs 29-31, 2022-24 | Changes in physical dysfunction. |
| Cognitive | Yrs 32-33, 2025-26 | Changes in cognitive dysfunction. |
| Liver | Yrs 29-31, 2022-24 | Fibrosis and steatosis. |
| Sleep | Yrs 29-31, 2022-24 | Obstructive sleep apnea. |
These domains are statistically heterogeneous. Some concern complications that can be classified at repeated assessments, cardiovascular events are described as occurring continuously with annual adjudication, and physical and cognitive outcomes concern changes over time. A single statistical method would therefore not necessarily be appropriate for every component of the registered endpoint.
5. Statistical Methodology
Why endpoint type determines the analysis
A major statistical feature of EDIC is the combination of several outcome structures within one registered cycle. A complication assessed repeatedly over time can be analysed differently from a cardiovascular event that occurs at an identifiable event time, while a measure of physical or cognitive dysfunction may be naturally represented as a longitudinal continuous outcome.
| Endpoint structure | Typical statistical framework | What the method estimates |
|---|---|---|
| Time to CVD event | Survival analysis such as Kaplan-Meier estimation and Cox regression | Event-free survival over time and relative event hazard |
| Repeated complication assessment | Longitudinal or repeated-measures models, depending on the outcome scale | Change or occurrence patterns across repeated assessments |
| Binary complication status | Logistic regression or longitudinal binary models when observations repeat | Association between participant characteristics and the probability of an outcome |
| Continuous physical or cognitive measure | Linear mixed-effects or related longitudinal models | Average change over time while accounting for repeated observations |
| Liver fibrosis or steatosis classification | Regression or longitudinal models appropriate to the recorded measurement | Differences or changes in liver-related outcomes over time |
| Obstructive sleep apnea | Outcome-specific regression or longitudinal methods | Occurrence or change in sleep-related outcome measures |
Time-to-event analysis
For a cardiovascular endpoint in which the time from observation to a major CVD event is known, survival analysis is a natural framework. It preserves information about when an event occurs rather than reducing follow-up to a simple yes/no indicator.
The survival function represents the probability that the event time T exceeds a specified time t. In a longitudinal study, participants who have not experienced an event by their last observed time may contribute censored follow-up.
Repeated measurements
When the same participant contributes physical, cognitive, or other measurements at multiple time points, those observations are correlated. Treating every observation as independent would generally understate that within-person dependence. Longitudinal models account for the repeated structure explicitly.
Here, Yij represents an outcome for participant i at observation j, β1 represents an average time trend, and bi represents participant-level heterogeneity. The exact model would depend on the scale and distribution of the recorded outcome.
Adjudicated cardiovascular events
The registry specifies that CVD events are assessed continuously and adjudicated annually. This distinction is important: the occurrence process and the adjudication process are related but different components of the data pipeline. Statistical analysis should preserve the event timing while using the adjudicated outcome definition consistently.
6. Planned Analysis
No formal statistical analyses are posted to ClinicalTrials.gov for the current EDIC record. The registry does, however, identify the outcomes to be measured and the periods during which those assessments occur.
Microvascular complications
For eyes, kidney, and nerves, the registry focuses on advanced microvascular complications. Depending on how each outcome is recorded, a statistical analysis would typically distinguish between the occurrence of a complication, its severity or classification, and its change over repeated assessments.
Cardiovascular disease events
For CVD events, the registry specifies continuous observation with annual adjudication. A typical analysis would use the event time as the fundamental unit of analysis. Kaplan-Meier methods can describe the cumulative event-free experience, while a Cox model can quantify relative differences in event hazard when its assumptions are appropriate.
Physical and cognitive dysfunction
The physical and cognitive domains are described as changes in dysfunction. When a quantitative measure is observed repeatedly, a longitudinal model can estimate average change while accounting for correlations among measurements from the same participant. If the outcome is categorical rather than continuous, the corresponding longitudinal binary or ordinal framework would be more appropriate.
Liver outcomes
The liver domain addresses fibrosis and steatosis. These outcomes can involve categorical, continuous, or otherwise structured measurements. The appropriate statistical model depends on the exact measurement scale and its repeated-observation structure.
Sleep outcomes
The sleep domain addresses obstructive sleep apnea. A typical analysis would first define the outcome precisely and then use a regression framework appropriate to whether the endpoint is binary, continuous, ordinal, or time-to-event.
7. Statistical Methods Explained
Why would cardiovascular events use survival analysis?
Because the registry specifies continuous CVD-event assessment, the timing of an event contains information. Two participants who both experience an event are not necessarily statistically equivalent if one experiences it substantially earlier than the other. Survival analysis retains this timing information and can also accommodate participants whose event has not occurred during their observed follow-up.
What does a hazard represent?
A hazard represents an instantaneous event rate among participants who remain event-free immediately before a given time. A hazard ratio compares those instantaneous rates between groups or according to a covariate. It is not the same quantity as a probability or a risk ratio.
Why are repeated observations not treated as independent?
Measurements from the same person tend to be related because they come from the same underlying individual. A standard regression that assumes every observation is independent can therefore produce inappropriate standard errors. Longitudinal methods account for within-participant correlation.
Why can the analysis differ between eyes, kidney, nerves, and sleep?
The clinical domains do not necessarily have the same statistical scale or observation process. A binary complication indicator, a continuous functional score, and a time-to-event outcome contain different types of information. Matching the model to the endpoint preserves the information most relevant to the clinical question.
What does censoring mean in a time-to-event analysis?
Censoring occurs when the event time is not observed during the available follow-up, even though the participant contributed information before follow-up ended. Standard survival methods use that partial information rather than simply treating the participant as if no event could ever occur.
Why does annual adjudication matter?
Adjudication provides a structured process for determining whether a reported cardiovascular event meets the study's outcome definition. From a statistical perspective, consistent outcome classification is important because misclassification can alter estimated event rates and treatment or exposure associations.
8. Longitudinal Structure and Follow-Up
The EDIC record spans multiple calendar periods and uses endpoint-specific assessment schedules. This is important statistically because a participant does not necessarily contribute the same type or number of observations to every endpoint.
Study start
The EDIC study begins its registered period of follow-up in April 1994.
Multiple current-cycle assessments
The registry specifies physical, liver, and sleep assessments during years 29-31, corresponding to 2022-24.
Extended current-cycle assessments
Nerve assessments are specified for years 29-31 and 33-34, covering 2022-26. CVD events are assessed continuously with annual adjudication.
Later current-cycle assessments
Eyes and cognitive outcomes are specified for years 32-33, corresponding to 2025-26.
Primary completion
The registry lists June 2027 as the primary completion date.
9. Analysis of the Registered Endpoint Structure
The phrase 2022–2027 Cycle encompasses multiple clinically meaningful domains rather than a single scalar endpoint. Statistically, this creates an important distinction between the registry's umbrella endpoint description and the individual variables that would be analysed within each domain.
| Statistical issue | Why it matters for EDIC |
|---|---|
| Different observation times | Eyes, kidney, nerves, CVD, physical, cognitive, liver, and sleep outcomes have different registered assessment schedules. |
| Repeated observations | Several domains are assessed across multiple years, creating within-participant correlation. |
| Event timing | CVD events are described as continuous and therefore naturally retain information about when events occur. |
| Outcome heterogeneity | Complications, dysfunction, fibrosis, steatosis, and sleep apnea may require different statistical scales and models. |
| Adjudication | CVD outcomes undergo annual adjudication, creating a defined outcome-classification process. |
This structure makes EDIC particularly useful for understanding why a complex longitudinal study should not be reduced to one generic statistical test. The statistical model is part of the definition of the estimand: whether the question concerns event occurrence, event timing, average change, prevalence, or another outcome property determines what constitutes an appropriate analysis.
10. Missing Data and Censoring
The registry specifies long-term follow-up with assessments occurring at different times. In longitudinal clinical research, incomplete observations can arise for several reasons, including missed visits, incomplete measurements, or the end of available follow-up.
The appropriate treatment depends on the endpoint and the missingness mechanism. For time-to-event outcomes, censoring methods can use information accumulated before the last observed time. For repeated continuous outcomes, mixed-effects models can often use incomplete longitudinal records under explicit assumptions about the missing-data process.
11. Multiplicity and Multiple Endpoint Domains
The current registered endpoint encompasses eight clinical domains: eyes, kidney, nerves, CVD events, physical function, cognition, liver outcomes, and sleep. These domains raise a statistical distinction between describing multiple outcomes and making formal confirmatory claims across multiple hypotheses.
If several hypotheses are tested formally, the number and relationship of those hypotheses can affect the overall probability of false-positive findings. Procedures such as hierarchical testing, multiplicity adjustment, or false-discovery-rate control can be considered depending on the scientific objective.
However, the registry record does not post a multiplicity strategy for the current cycle. The appropriate interpretation of any future collection of statistical tests would therefore require the study's prespecified analysis plan and the intended role of each endpoint.
12. Subgroup and Covariate Analysis
The current registry information identifies the clinical domains and assessment schedules but does not post subgroup estimates, covariate-adjusted effect estimates, interaction tests, or a prespecified subgroup analysis framework.
In a longitudinal study, covariate adjustment can serve several purposes. Baseline variables may improve precision, account for important sources of variation, or describe associations between participant characteristics and outcomes. The statistical interpretation depends on whether a variable is being used for prediction, adjustment, description, or a causal analysis.
Prediction
A model may estimate which participant characteristics are associated with future outcomes without claiming that changing those characteristics would cause the outcome to change.
Adjustment
Covariates can account for measured differences when estimating an association between another variable and an outcome.
Interaction
An interaction term evaluates whether an association differs across levels of another variable rather than simply testing each subgroup separately.
Longitudinal modeling
Repeated measurements require attention to both covariate structure and the dependence among observations from the same participant.
13. Interim Analysis, Bayesian Methods, and Non-Inferiority
The current ClinicalTrials.gov record does not report a prespecified interim efficacy analysis, Bayesian statistical framework, or non-inferiority margin for the 2022–2027 cycle.
These design concepts are important to distinguish from ordinary longitudinal follow-up. An interim analysis is a formal statistical look at accumulating evidence under a predefined monitoring framework. Bayesian analysis requires an explicit prior distribution and likelihood-based updating. Non-inferiority analysis requires a prespecified margin defining the largest clinically acceptable loss of efficacy.
None of those elements is identified in the current registry endpoint information, so they are not incorporated into the planned analysis description for EDIC here.
14. Safety and Clinical Outcomes
The current registry endpoint description focuses on advanced microvascular complications, cardiovascular disease events or early subclinical disease, physical and cognitive dysfunction, liver fibrosis and steatosis, and obstructive sleep apnea.
No serious-adverse-event counts by study arm are reported in the current record information. Accordingly, the statistical focus of the current cycle is on the registered clinical outcome domains rather than on a treatment-arm safety comparison.
15. Why This Trial Matters Statistically
EDIC is a useful statistical teaching case because its current registry endpoint combines long-term follow-up, repeated assessments, event-time information, and multiple organ-system outcomes within one study program.
| Concept | How it appears in EDIC |
|---|---|
| Longitudinal analysis | Multiple assessment periods spanning the current 2022–2027 cycle. |
| Repeated measurements | Several domains are assessed during multiple specified years. |
| Time-to-event analysis | CVD events are followed continuously. |
| Event adjudication | CVD events are adjudicated annually. |
| Endpoint heterogeneity | The cycle covers eyes, kidney, nerves, CVD, physical, cognitive, liver, and sleep outcomes. |
| Censoring | Longitudinal follow-up naturally creates incomplete event histories when observation ends before an event. |
| Missing data | Repeated clinical assessments can produce incomplete longitudinal records that require appropriate handling. |
| Multiplicity | Multiple clinical domains create a distinction between descriptive analyses and formal multiple-hypothesis testing. |
| Model selection | The endpoint scale and observation process determine whether survival, regression, or longitudinal methods are appropriate. |
16. Clinical Interpretation vs Statistical Interpretation
Clinical interpretation
The registry's current cycle follows several important manifestations and consequences of type 1 diabetes, including advanced microvascular complications, cardiovascular outcomes, physical and cognitive dysfunction, liver disease, and obstructive sleep apnea.
Statistical interpretation
The breadth of these outcomes means that each endpoint must be analysed according to its measurement scale, timing, repeated-observation structure, and event process. The same statistical model is not automatically suitable for every domain.
17. Important Limitations and Interpretation Issues
- Multiple outcome domains: the registered cycle contains several distinct clinical domains rather than one homogeneous endpoint.
- Different assessment schedules: the registry specifies different time frames for eyes, kidney, nerves, CVD, physical, cognitive, liver, and sleep outcomes.
- Longitudinal dependence: repeated observations from the same participant are correlated and should not ordinarily be analysed as independent observations.
- Event timing: CVD events are followed continuously, so reducing them to a simple ever/never indicator could discard information about event timing.
- Adjudication: annual adjudication is part of the registered CVD-event process and should be reflected in the definition of the analysed endpoint.
- Missing observations: long-term follow-up can generate incomplete longitudinal records, and the appropriate treatment depends on the endpoint and missing-data mechanism.
- Multiplicity: multiple endpoint domains can generate multiple statistical comparisons, making the distinction between exploratory and confirmatory inference important.
- Current results availability: the ClinicalTrials.gov record does not post formal statistical analyses for the registered primary endpoint.
- Ongoing study: the registry lists the study as active, not recruiting, with primary completion in June 2027.
18. What a Future Statistical Results Table Should Contain
For a study with this structure, a useful statistical results presentation would separate the outcome domains rather than combining them into one overall effect measure.
| Domain | Core statistical question | Useful result type |
|---|---|---|
| Eyes | How does retinopathy status or severity change over follow-up? | Outcome-specific longitudinal or regression estimate |
| Kidney | How does kidney disease develop or change over annual assessments? | Longitudinal complication or progression estimate |
| Nerves | How does neuropathy change across the registered assessment periods? | Repeated-measures or event-based estimate |
| CVD events | When do major cardiovascular events occur? | Event-free survival and hazard-based estimates |
| Physical | How does physical dysfunction change? | Longitudinal change estimate |
| Cognitive | How does cognitive dysfunction change? | Longitudinal change estimate |
| Liver | How do fibrosis and steatosis change or occur? | Outcome-specific regression or longitudinal estimate |
| Sleep | How does obstructive sleep apnea occur or change? | Outcome-specific regression or longitudinal estimate |
This framework also clarifies why an individual estimate should always be accompanied by its estimand, analysis population, observation period, uncertainty measure, and statistical model. Without those elements, a numerical result can be difficult to interpret even when the underlying measurement is clinically important.
19. Record Summary
EDIC is a long-term study of type 1 diabetes mellitus with enrollment of 1,441 and a study period beginning in April 1994. The current registry endpoint, the 2022–2027 Cycle, spans eight clinical domains: eyes, kidney, nerves, CVD events, physical function, cognition, liver outcomes, and sleep. The assessment schedules range from annual kidney evaluation to continuous cardiovascular-event observation with annual adjudication and specified multi-year windows for the other domains.
Statistically, the central lesson is the importance of matching the analysis to the endpoint. Cardiovascular events naturally support time-to-event methods, while repeated physical or cognitive measurements call for longitudinal methods that account for within-participant correlation. Complication outcomes may require regression or longitudinal approaches according to their measurement scale. Multiple endpoint domains also make multiplicity and the distinction between descriptive and confirmatory inference important.
20. Sources
- ClinicalTrials.gov: NCT00360893 — Epidemiology of Diabetes Interventions and Complications (EDIC).
Continue through the Clinical Biostats clinical-trial library
Explore additional clinical trial analyses and statistical resources covering study design, endpoints, longitudinal methods, survival analysis, and clinical-trial interpretation.