Clinical Statistical Workflow

From research questions to reviewable delivery.

Clinical statistics connects evidence, decisions, data standards, specifications, programs, validation, and review. AI supports how these parts are organized and carried forward, while professional roles shape judgment, interpretation, and approval.

01 · Business Workflow

Clinical statistical delivery

Each stage carries professional decisions into a defined artifact, review point, and downstream dependency.

  1. 01

    Protocol

    Decision
    Interpret objectives, endpoints, populations, visits, estimands, and study constraints.
    Artifact
    Traceable requirements and questions requiring confirmation.
    AI and system support
    Locate relevant statements, connect sources, and surface gaps or inconsistencies.
    Professional judgment
    Confirm the clinical question and its statistical interpretation.
  2. 02

    Statistical strategy / SAP

    Decision
    Define estimands, analysis sets, methods, intercurrent events, missing data, multiplicity, and sensitivity analyses.
    Artifact
    Statistical decisions, SAP content, and analysis specifications.
    AI and system support
    Retrieve applicable knowledge, organize possible approaches, and check Protocol/SAP consistency.
    Professional judgment
    Select methods, resolve assumptions, and approve the statistical strategy.
  3. 03

    SDTM

    Decision
    Define source mappings, domains, controlled terminology, timing, and traceability.
    Artifact
    SDTM specifications, datasets, and validation findings.
    AI and system support
    Propose mappings, apply repeatable rules, and organize exceptions for review.
    Professional judgment
    Confirm mapping exceptions and interpretations of data standards.
  4. 04

    ADaM

    Decision
    Define analysis populations, parameters, derivations, baselines, and analysis time points.
    Artifact
    ADaM specifications, datasets, and traceability to SDTM and SAP decisions.
    AI and system support
    Translate approved decisions into structured derivations and check consistency.
    Professional judgment
    Confirm analysis meaning, derivation rules, and fit for statistical use.
  5. 05

    TFL

    Decision
    Define displays, statistical calculations, shells, programs, and output structure.
    Artifact
    TFL shells, programs, tables, listings, figures, and results.
    AI and system support
    Turn specifications into clear, repeatable tasks and check the expected structure.
    Professional judgment
    Review statistical interpretation, program logic, and output presentation.
  6. 06

    QC / Review

    Decision
    Reconcile data, programs, results, versions, and independent findings.
    Artifact
    QC evidence, differences, review records, and resolved decisions.
    AI and system support
    Run repeatable validation, locate differences, and assemble review context.
    Professional judgment
    Perform independent QC and determine whether findings are resolved.
  7. 07

    Submission

    Decision
    Confirm completeness, consistency, version identity, and delivery readiness.
    Artifact
    Reproducible, traceable, and reviewable delivery materials.
    AI and system support
    Assemble validation and traceability evidence across the delivery package.
    Professional judgment
    Approve the formal delivery and its supporting evidence.

02 · AI Workflow

AI-supported work across the delivery path

AI assists retrieval, organization, transformation, execution, and checks, while professional roles complete statistical judgment, interpretation, and approval.

  1. 01Knowledge context
  2. 02Statistical decision
  3. 03Structured specification
  4. 04Assisted execution
  5. 05Validation
  6. 06Human review
  7. 07Traceable delivery

03 · Product 01

Clinical Knowledge Wiki

RAG & Vector Knowledge Base

A traceable knowledge system for organizing guidelines, standards, study documents, and professional experience into content that can be retrieved, checked, and reused.

  1. 01Sources
  2. 02Evidence
  3. 03Knowledge structure
  4. 04RAG & vector retrieval
  5. 05Human review
  6. 06Versioned knowledge

AI and system support

  • Locate evidence in long documents and preserve relevant context.
  • Connect related concepts, terminology, and semantic matches.
  • Organize key summaries, differences, conflicts, and knowledge needs.

Professional judgment

  • Decide whether a source and interpretation apply to the current context.
  • Resolve conflicts, review knowledge content, and approve formal versions.

Current capabilities

  • Organized sources, evidence, knowledge entries, versions, and relations.
  • Document preparation and traceable evidence locations.
  • Knowledge entries and a professional review workspace.

Next workflow

  • Wiki-based knowledge organization with keyword, semantic, and vector retrieval.
  • Complete professional review and versioned knowledge publication.
  • Reviewed, traceable knowledge context for Clinical Workflow.
View project →

04 · Product 02

Clinical Workflow

Statistical Delivery Workflow

A clinical statistical delivery workflow connecting study requirements, statistical decisions, data standards, analysis specifications, program outputs, and quality review.

  1. 01Protocol
  2. 02Statistical strategy / SAP
  3. 03SDTM
  4. 04ADaM
  5. 05TFL
  6. 06QC / Review
  7. 07Submission

AI and system support

  • Organize requirements and retrieve reviewed knowledge context.
  • Translate confirmed decisions into structured specifications and repeatable tasks.
  • Support consistency checks, validation, and review preparation.

Professional judgment

  • Interpret the study and decide estimands, populations, methods, exceptions, and derivations.
  • Review program logic and results, perform independent QC, and approve delivery.

Current capabilities

  • Fixed clinical stages and explicit business dependencies.
  • Action selection, review records, and knowledge-version references.
  • Study materials, representative deliverable flows, and change records.

Next workflow

  • A continuous record connecting decisions, specifications, execution, and artifacts.
  • Stage-based execution connecting validation, professional review, and formal delivery.
  • Knowledge-version references, identified knowledge needs, and review feedback.
View project →

05 · Architecture

Two products connected by traceable context

The knowledge product and the delivery workflow retain separate responsibilities while exchanging versioned context and review feedback.

01Clinical Knowledge Wiki

→ Reviewed, versioned knowledge context with sources and applicability

← Knowledge needs, differences, and review feedback requiring professional assessment

02Clinical Workflow

06 · Governance

Professional judgment, system support, and approval form a responsibility chain

The system makes repeated work more consistent, while professional roles carry the corresponding responsibilities for statistical interpretation, quality review, and formal approval.

Professional roles
Clinical interpretation, statistical methods, assumptions, exceptions, independent review, risk acceptance, and final approval.
Knowledge product
Sources, evidence, knowledge structure, retrieval context, review state, and versioned knowledge.
Clinical workflow
Business stages, decisions, specifications, artifacts, validation state, review, and delivery traceability.
AI support
Retrieval, organization, approach preparation, transformation, assisted execution, consistency checks, and review preparation.

07 · Projects

Existing product foundations

The current project cases document the evidence, workflow capabilities, and next direction for each product.

Clinical Knowledge WikiClinical Workflow

Workflow Insights

Related insights

Focused accounts connected to the workflow stages shown above.

  1. Reviewed

    From protocol to SAP: making statistical decisions executable

    The protocol defines the scientific question; the SAP must turn it into analysis choices that can be reviewed before programming begins.

  2. Reviewed

    From SDTM and ADaM to TFL: what automation changes

    Automation is most useful when specifications, programs, validation, review, and release form one controlled delivery chain.

  3. Reviewed

    From document retrieval to reusable knowledge

    Source evidence, professional interpretation, review records, and knowledge versions connect document retrieval to the clinical workflow.