Intention-to-Diagnose (ITD) is an analysis principle in which all study participants are evaluated within their initially randomized diagnostic pathway, irrespective of whether they actually received that intervention, deviated from protocol, or were lost to follow-up. This strategy preserves the prognostic balance achieved by randomization, preventing selection bias that would arise if subjects who crossed over or failed to complete the assigned diagnostic procedure were excluded or reclassified.
Glossary
Intention-to-Diagnose
An analysis strategy including all subjects in their originally assigned diagnostic groups regardless of protocol deviations, preserving the benefits of randomization.
CLINICAL TRIAL METHODOLOGY
What is Intention-to-Diagnose?
An analysis strategy that preserves the benefits of randomization by including all subjects in their originally assigned diagnostic groups, regardless of protocol deviations or non-compliance.
In contrast to a per-protocol analysis, which only evaluates subjects who fully adhered to the assigned diagnostic workflow, ITD provides a pragmatic, real-world estimate of effectiveness rather than idealized efficacy. By maintaining the integrity of the original randomized groups, ITD analysis yields a more conservative and generalizable assessment of a diagnostic test's clinical impact, making it the preferred primary analysis for regulatory submissions and pivotal trials.
INTENTION-TO-DIAGNOSE
Core Characteristics of ITD Analysis
The Intention-to-Diagnose (ITD) principle is a cornerstone of pragmatic clinical trial analysis. It mandates that all subjects be analyzed according to their original randomized group assignment, regardless of protocol deviations, non-compliance, or withdrawal, thereby preserving the prognostic balance created by randomization.
01
Preservation of Randomization
The primary rationale for ITD is to maintain the baseline prognostic balance achieved by randomization. Excluding non-compliant subjects introduces selection bias, as the reasons for deviation are often linked to prognosis. By analyzing all subjects as originally assigned, the comparison remains an unbiased estimate of the effectiveness of a diagnostic strategy in real-world clinical practice, rather than just its efficacy under perfect conditions.
02
Pragmatic vs. Explanatory Contrast
ITD analysis is the standard for pragmatic trials measuring effectiveness (does it work in practice?), while per-protocol (PP) analysis suits explanatory trials measuring efficacy (can it work ideally?).
- ITD: Estimates the policy of using the diagnostic test.
- PP: Estimates the biological effect of the test result itself.
- Result: ITD often yields a more conservative, conservative estimate of the treatment effect, reflecting dilution due to non-adherence.
03
Handling of Missing Data
A critical challenge in ITD is the management of missing outcome data from subjects lost to follow-up. Simple exclusion (complete-case analysis) breaks the ITD principle. Robust methods include:
- Multiple Imputation (MI): Creates plausible values based on observed data.
- Inverse Probability Weighting (IPW): Weights complete cases to represent missing ones.
- Sensitivity Analysis: Tests if conclusions hold under different assumptions about the missing data mechanism (Missing Not at Random).
04
Regulatory Preference & ICH E9
Regulatory bodies like the FDA and EMA strongly prefer ITT (Intention-to-Treat) for confirmatory trials. The ICH E9 (R1) addendum explicitly frames ITT as a defining principle for estimating treatment effects. For diagnostic AI submissions, a pivotal trial analyzed by ITD demonstrates that the software's clinical utility is robust against real-world workflow non-compliance, such as a clinician ignoring the AI's recommendation.
05
Non-Inferiority & ITD Bias
In non-inferiority (NI) trials, ITD analysis is particularly sensitive. Non-compliance and diagnostic crossover tend to make the experimental and control arms look more similar, artificially inflating the chance of concluding non-inferiority. Consequently, regulators often require both ITD and PP analyses to reach concordant conclusions in an NI study. A conservative ITD approach is essential to avoid falsely claiming a new diagnostic AI is 'just as good' as the standard of care.
06
Estimand Framework Application
The ICH E9(R1) estimand framework precisely defines the treatment effect of interest. For an ITD analysis, the estimand attributes intercurrent events (like protocol deviations) using a 'treatment policy' strategy. This means the occurrence of the intercurrent event is considered part of the diagnostic strategy itself. The clinical question becomes: 'What is the effect of assigning the AI diagnostic tool, regardless of whether clinicians follow its advice?'
ANALYSIS POPULATION STRATEGIES
Intention-to-Diagnose vs. Per-Protocol Analysis
Comparison of subject inclusion criteria and statistical properties for the two primary analysis populations in diagnostic accuracy trials
| Feature | Intention-to-Diagnose (ITD) | Per-Protocol (PP) | As-Treated (AT) |
|---|---|---|---|
Primary principle | Analyze all subjects as originally randomized regardless of protocol adherence | Analyze only subjects who completed the study without major protocol violations | Analyze subjects based on the diagnostic intervention actually received |
Preserves randomization | |||
Reflects real-world clinical practice | |||
Risk of selection bias | |||
Estimates treatment efficacy under ideal conditions | |||
Handles non-adherence | Subjects remain in assigned group | Non-adherent subjects excluded | Subjects reclassified by actual intervention received |
Regulatory preference for superiority trials | |||
Regulatory preference for non-inferiority trials | Recommended as co-primary | Often required as co-primary | |
Missing data approach | Imputation required (e.g., multiple imputation, worst-case scenario) | Complete case analysis or simple imputation | Complete case analysis or simple imputation |
Statistical power impact | May be reduced due to dilution from non-adherence | May be inflated due to selection of compliant subjects | Unpredictable; depends on crossover patterns |
Generalizability of findings | High; mirrors clinical deployment population | Low; restricted to ideal protocol followers | Low; confounded by selection factors |
Risk of attrition bias | Mitigated by imputation strategies | High; differential dropout between arms biases results | High; reasons for switching interventions confound outcomes |
CLINICAL TRIAL METHODOLOGY
Frequently Asked Questions
Clarifying the statistical foundations of rigorous diagnostic AI validation, these answers address common queries about the Intention-to-Diagnose principle and its role in preserving randomization benefits.
The Intention-to-Diagnose (ITD) principle is an analysis strategy where all study subjects are evaluated in their originally assigned diagnostic groups, regardless of protocol deviations, non-compliance, or withdrawal. This approach preserves the benefits of randomization by preventing selection bias that would arise if subjects were excluded based on post-assignment events. In a clinical validation study for an AI diagnostic tool, ITD mandates that a patient randomized to receive the AI-assisted diagnosis is analyzed in that group even if the clinician overrides the AI recommendation or the software encounters a technical failure. The primary goal is to estimate the real-world effectiveness of a diagnostic strategy rather than its efficacy under perfect laboratory conditions. This mirrors the Intention-to-Treat (ITT) principle in therapeutic trials and is critical for maintaining the statistical integrity of pivotal trials intended for regulatory submission.
Enabling Efficiency, Speed & Accuracy
Intelligent Analysis, Decision & Execution
We build AI systems for teams that need search across company data, workflow automation across tools, or AI features inside products and internal software.
Talk to Us
Search across company data
Give teams answers from docs, tickets, runbooks, and product data with sources and permissions.
Useful when people spend too long searching or get different answers from different systems.
Enterprise searchRAGPermissions
Read more
Automate internal workflows
Use AI to route work, draft outputs, trigger actions, and keep approvals and logs in place.
Useful when repetitive work moves across multiple tools and teams.
AI agentsWorkflow automationGovernance
Read more
Add AI to products and internal tools
Build assistants, guided actions, or decision support into the software your team or customers already use.
Useful when AI needs to be part of the product, not a separate tool.
AI integrationDecision supportModel routing
Read more
About the author
Prasad Kumkar
CEO & MD, Inference Systems
Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.
His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.
LinkedIn
Limited slotsGet a Free AI Consultation
Partnered with leading AI, data, and software stack.
How We Work
Custom AI workflows for your Business
One-fit-all AI don't work for modern businesses. At Inferensys, we aim to understand your business & custom requirements; which we use to define most efficient agentic workflows, the data, and the tools for your business.
01
Review the use case
We understand the task, the users, and where AI can actually help.
Read more02
Pick the right approach
We define what needs search, automation, or product integration.
Read more03
Build the first useful version
We implement the part that proves the value first.
Read more04
Improve from there
We add the checks and visibility needed to keep it useful.
Read moreThe first call is a practical review of your use case and the right next step.
Talk to Us