Innovator›Clinical Development›Study & Statistics

Study blueprint → structured protocol → estimand/SAP → SDTM → ADaM → Python statistics → TLF/QC → SAS interoperability

Clinical Study & Statistical Programming OS

Build the executable evidence chain around the same programme and candidate used by Clinical Development, MIDD and Qwen Progression. The statistical specification is authoritative; Python is the native execution engine and SAS is an optional interoperability target.

M11-shaped structured protocolPython-native statisticsSDTM / ADaM / DefineSAS XPT + code export

Authority & validation boundary

BayesPharma makes the workflow inspectable; it does not turn a generated artifact into regulator acceptance.

Fail closed

Evidence semantics: approved source observations remain MEASURED; Clinical Data Spine, SDTM, ADaM, Define metadata, Python analysis and reconciliations are COMPUTED; an accountable independent statistical/design review can create REVIEWED evidence. No generated artifact advances a clinical phase or authorizes dosing.

Conformance boundary: “M11 structured” does not mean certified M11 conformance. BayesPharma structural SDTM/ADaM/Define checks do not replace supported external CDISC/FDA validation. Generated SAS programs are not executed by BayesPharma and require no SAS license to generate.

Workspace inventory

Refresh the three canonical workspaces without changing state.

Blueprints—

Estimands—

SAPs—

SDTM packages—

TLFs—

SAS drafts—

Latest workspace JSON

Loading…

1 · DESIGNM11 structured ≠ certified

Study blueprint & protocol

Create a programme-linked blueprint, then version a structured protocol. Use existing Clinical OS study IDs—no second study identity.

Blueprint request JSON

{
  "study_id":"",
  "title":"Phase II proof-of-concept study",
  "phase":"phase2",
  "therapeutic_area":"",
  "indication":"",
  "rationale":"Decision-linked structured protocol design.",
  "objectives":{"OBJ1":"Estimate the primary treatment effect."},
  "planned_population":"Population defined by the programme protocol.",
  "inclusion_criteria":[],"exclusion_criteria":[],
  "arms":[{"arm_id":"A","label":"Control","intervention":"Control","dose_regimen":"Per protocol"},{"arm_id":"B","label":"Active","intervention":"Investigational product","dose_regimen":"Per protocol"}],
  "visits":[{"visit_id":"V1","name":"Baseline","nominal_day":0,"activities":["baseline assessments"]}],
  "endpoints":[{"endpoint_id":"EP1","name":"Primary endpoint","endpoint_type":"primary","objective_id":"OBJ1","variable":"AVAL","timepoint":"Primary timepoint","measurement_method":"Per protocol","analysis_population":"FAS"}],
  "stopping_rules":[],"safety_monitoring":[],"critical_to_quality_factors":[],"data_handling_principles":[],"actor":"Clinical Scientist"
}

Protocol version request JSON

Ready.

2 · ESTIMAND + SAP

Lock the question before programming

Estimands are first-class objects. SAPs bind estimands, analysis populations, primary/sensitivity analyses and TLF shells to a protocol version.

Estimand request JSONSAP request JSONIndependent design review JSON

Ready.

3 · DATA STANDARDSExternal validator still required

Clinical Data Spine → SDTM → ADaM → Define

Transforms only immutable Clinical Data Spine snapshots. Mapping and derivations are explicit; transformed rows remain COMPUTED artifacts.

SDTM request JSONADaM request JSONDefine metadata request JSON

Ready.

4 · PYTHON STATISTICS

Prespecified native execution

No arbitrary-code endpoint. Whitelisted methods execute against a locked ADaM hash with Python/package versions captured in every run.

Analysis specification JSON

Ready.

5 · QC / REVIEW

Reconcile before consequential use

Compare numeric results to an independent implementation, then require a reviewer principal distinct from author/executor to sign the statistical interpretation.

Reconciliation request JSONIndependent statistical review JSON

Ready.

6 · SAS INTEROPERABILITYSAS execution = false

Generate SAS from the same spec

BayesPharma can produce a SAS program and XPT files without owning SAS. A customer can execute the program in its validated SAS environment and return results for reconciliation.

SAS program request JSON

Commercial boundary: no SAS/Viya subscription is required for BayesPharma's Python execution, XPT conversion or SAS source generation. SAS execution remains an optional customer-side integration later.

Ready.

Recommended operational flow

Use the existing Clinical Data handoff for source imports/QC, then return here for standards and analysis.