Jaime Yan — Research & Writing

Git for SAS Programmers: Version Control in a GxP World

Tue, 06 Oct 2026 00:00:00 GMT

Why filename-versioned SAS programs are an audit-trail liability, the minimum Git vocabulary a clinical programmer needs, and how branch-per-output maps to QC.

Pipeline as Code: Automating the SDTM-to-ADaM-to-TLF Chain

Tue, 06 Oct 2026 00:00:00 GMT

How to turn the SDTM-to-ADaM-to-TLF chain into pipeline as code: explicit dependencies, pinned environments, hash-verified outputs, and stage-level failure isolation.

Becoming a Clinical Statistical Programmer in 2026: A Realistic Guide

Tue, 06 Oct 2026 00:00:00 GMT

What the job is, who gets hired, and how to train for it: a realistic 2026 guide to becoming a clinical statistical programmer, from SAS base to AI-assisted practice.

Human-Governed Validation of R Packages for Regulated Statistical Computing: LLM-Assisted Specifications and Auditable Test Evidence

Tue, 06 Oct 2026 00:00:00 GMT

Preprint (submitted to Pharmaceutical Statistics) — Open-source R packages are increasingly used in regulated statistical computing, but risk scores, coverage, and upstream tests do not by themselves demonstrate fitness for an intended use. This preprint presents a specification-driven validation prototype: versioned specifications define requirements and acceptance criteria; an isolated engine combines known-answer, metamorphic, property-based, fuzz, and mutation tests; and records enter a tamper-evident hash chain. An optional local LLM drafts structure only — expected values are captured characterization baselines or independently derived, and approval remains human. A public contract replay passed 15/15 LLM authority-boundary checks including a prompt-injection red-team set. A pristine synthetic R package passed 11/11 cases spanning four test techniques while four pre-injected defect variants each failed the targeted requirement; a jsonlite replay surfaced the documented digits=4 precision semantics; a synthetic time-to-event capsule reproduced prespecified Kaplan–Meier and Cox outputs and passed 50/50 row-order invariance checks. Public artifacts include source, specifications, machine-readable outputs, replay scripts, and SHA-256 manifests. The results demonstrate test-oracle construction, fail-closed model governance, and auditable correction — not package quality or regulatory compliance per se. Companion volume: the online book The GxP Statistical Computing Environment.

The 2026 CDISC AI Challenge: Winners, Open Code, and a Synthetic ADaM Showdown

Mon, 05 Oct 2026 00:00:00 GMT

The 2026 CDISC AI Innovation Challenge winners present today in Denver. None published code — except one adjacent R package. We read synadam's source and compare it to our own synthetic ADaM pipeline.

Pharma Daily — October 5, 2026

Mon, 05 Oct 2026 00:00:00 GMT

One pipeline, five first-hand sources: 3 deals, 0 FDA approvals and 12 upcoming readouts, collected and charted without an LLM.

Clinical R in Practice: The Open-Source Stack in Pharma

Thu, 01 Oct 2026 00:00:00 GMT

A 15-part series on R in pharma: pharmaverse ADaM, ARD tables, risk-based validation, targets pipelines, GxP Shiny, LLM agents — collected into a free bilingual book.

Building ADaM with admiral: The LEGO Method

Wed, 30 Sep 2026 00:00:00 GMT

A complete ADaM derivation built from composable bricks: the derive_* mental model, ADSL and BDS patterns, and why company templates sit on top of admiral instead of replacing it.

ARD: Analysis Results Data, and Why Tables Are Becoming Data

Wed, 30 Sep 2026 00:00:00 GMT

ARD turns tables from pixels into data — machine-reviewable, traceable, reusable. How the cards standard works, how gtsummary speaks it natively, and why reviewers will eventually ask for it.

Agents and MCP in the R Clinical Stack

Wed, 30 Sep 2026 00:00:00 GMT

From chat to autonomous workflow: how the Model Context Protocol standardizes what tools agents may touch, what an agent-native clinical pipeline looks like, and the guardrails that keep autonomy auditable.

From SDTM to Submission: The Clinical Data Flow, End to End

Wed, 30 Sep 2026 00:00:00 GMT

One map of every dataset, standard, and hand-off between the clinic and the regulator — SDTM, ADaM, define.xml, TLFs, and the submission package — and where R now sits at each stage.

The Next Five Years of Clinical Data Science

Wed, 30 Sep 2026 00:00:00 GMT

The series compressed into one argument: standards absorbing tooling, tooling absorbing validation, validation absorbing AI — where the field goes, and the book this series becomes.

rtables vs gt/gtsummary vs flextable: A Regulatory Showdown

Wed, 30 Sep 2026 00:00:00 GMT

One adverse-event table built three ways — the structural differences, the validation story of each framework, and a decision tree for which framework earns which role.

Shiny in GxP: From rhino to a Validation File

Wed, 30 Sep 2026 00:00:00 GMT

The complete path for a clinical Shiny app: rhino structure, shinytest2 suites, secure backends, async patterns, and the validation evidence file that lets QA sign.

LLMs Writing Trial Code: Production Cases, Real Limits

Wed, 30 Sep 2026 00:00:00 GMT

AI pair programmers for ADaM, LLM-assisted tables, QC drafting — the production case ledger from 2024-2026, and the exact wall each case hit when the rubber met validation.

Metadata-Driven Development: metacore to xportr

Wed, 30 Sep 2026 00:00:00 GMT

Spec as code: how metacore, metatools, and xportr turn your define metadata into checks, labels, and transport files — killing inconsistency at the source instead of in QC.

From Natural Language to CDISC Datasets

Wed, 30 Sep 2026 00:00:00 GMT

NL-to-CDISC generation, LLM-built analysis apps, auto slide decks — how close is the fully automatic submission, and which bottleneck is regulatory rather than technical?

The Pharmaverse Ecosystem Map: Who Builds What, and Why

Wed, 30 Sep 2026 00:00:00 GMT

admiral, tern, teal, metacore, xportr — who builds which package, how they interlock, who governs them, and how a company or a solo programmer enters the ecosystem.

Quarto for Clinical Study Reports

Wed, 30 Sep 2026 00:00:00 GMT

Parameterized CSR automation: branded documents, multi-study templating, the SAS engine for legacy code, and where auto-generated TLFs flow into review-ready reports.

Risk-Based R Validation: Open Source Under Inspection

Wed, 30 Sep 2026 00:00:00 GMT

The R Validation Hub method, hands-on: risk dimensions, the public risk-assessment app, package qualification records, and what an inspector actually asks for when your stack is open source.

The SAS-to-R Ledger: What Migrations Actually Cost

Wed, 30 Sep 2026 00:00:00 GMT

What large pharma migrations really cost — the training trap, the dual-run tax, the retraining ledger — and the ROI logic that survives an audit committee.

targets Pipelines: Reproducible Clinical Analyses

Wed, 30 Sep 2026 00:00:00 GMT

Incremental compute, crew parallelism, frozen environments — a pipeline that re-runs in minutes, replays exactly a year later, and answers an inspector by existing.

Validation as a Release Gate: Building a GxP Governance Layer for Shiny at Kardigan

Mon, 28 Sep 2026 00:00:00 GMT

How I designed the governance and validation layer for ClinVista, an internal Shiny data-review platform — and why it survived the decision to replace the platform itself.

Free TFL Mock Shell Generator: A Practical Walkthrough

Sun, 20 Sep 2026 00:00:00 GMT

Build clinical TFL mock shells, review variable bindings, and export documents and SAS/R program scaffolds with a free browser tool and a practical walkthrough.

Pharma Daily — September 17, 2026

Thu, 17 Sep 2026 00:00:00 GMT

One pipeline, five first-hand sources: 32 deals, 25 FDA approvals and 12 upcoming readouts, collected and charted without an LLM.

Pharma Daily — September 9, 2026

Wed, 09 Sep 2026 00:00:00 GMT

Sep 8–9: 17 deal events, 0 FDA approvals, 12 queued readouts; Novartis's $12B Avidity bet hits a Phase 3 miss and Monte Rosa signs a $2.1B license.

Novartis-Avidity at $12B: One Filing, One Win, One Miss

Tue, 08 Sep 2026 00:00:00 GMT

Six months after closing the $72/share Avidity buy: del-zota sits under FDA priority review, del-brax won Phase 1/2, and del-desiran failed HARBOR — NVS fell 13.9%.

Pharma Daily — September 8, 2026

Tue, 08 Sep 2026 00:00:00 GMT

Sep 7–8: $14.6B in disclosed totals, zero FDA approvals, and Novartis's $12B Avidity buyout's lead asset fails Phase 3. Monte Rosa licenses VAV1 degraders to Novartis.

Ivonescimab Clears OS vs Keytruda; the Hazard Ratio Waits for WCLC

Tue, 08 Sep 2026 00:00:00 GMT

HARMONi-2's OS win over pembrolizumab is real but detail-light: SMMT +17.3% on 2.7x volume, Akeso +15.3% in two sessions — and the HR itself lands at WCLC this month.

ADTTE and Time-to-Event: Programming the ADaM Analysis Dataset

Mon, 07 Sep 2026 00:00:00 GMT

ADTTE step by step: event and censoring definitions from the SAP, the censoring date cascade, CNSR semantics, partial dates at the event, and QC listings per subject.

Pharma Daily — September 7, 2026

Mon, 07 Sep 2026 00:00:00 GMT

Sep 6–7: zero deals, zero FDA approvals, and 12 Phase 2/3 readouts due within 90 days. The calendar, not the tape, is the story.

AgomAb's $200M IPO: An 85% Rally With No Data Behind It

Sat, 05 Sep 2026 00:00:00 GMT

AgomAb priced $200M at the midpoint in February, broke 8.4% on debut, fell to a −39.9% close by June, then rallied 85% on no data — now −1.4% vs offer as of Sep 4.

Aktis's $318M IPO: Lilly Took a Third, and the Tape Followed

Sat, 05 Sep 2026 00:00:00 GMT

Aktis priced 2026's first biotech IPO at the top of range with Eli Lilly buying $100M in the deal; the stock is +50% by September. The anchor playbook the class repeated.

Alamar's $220M IPO: The Tools Company the Bifurcation Didn't Touch

Sat, 05 Sep 2026 00:00:00 GMT

Alamar upsized 20% to $219.9M with the shoe, popped +29.4% on debut, and trades +81.2% vs offer — a proteomics tools story priced on revenue, not trial risk.

Apnimed's $220.8M IPO: An NDA-Stage Bet the Tape Keeps Rewarding

Sat, 05 Sep 2026 00:00:00 GMT

Apnimed upsized 20% to $192M at the top of range, popped +56.3% on debut, and trades +71.6% vs offer — with an FDA decision on its sleep-apnea pill due Feb 28, 2027.

Attovia's $289M IPO: Phase 1 Itch Data the Tape Paid For

Sat, 05 Sep 2026 00:00:00 GMT

Attovia priced an upsized $289M IPO at the $17 top of range on August 4, popped +28.8%, and never closed below offer — Phase 1b itch data, runway into 2030.

Avalyn's $345M IPO: The Reformulation Play That Keeps Climbing

Sat, 05 Sep 2026 00:00:00 GMT

Avalyn upsized 41% to $300M at the top of range, popped +63.8%, and doubled by September on enrollment execution — with inhaled pirfenidone data due H2 2027.

BlossomHill's $150M IPO: Early-Stage, and the Tape Paid Anyway

Sat, 05 Sep 2026 00:00:00 GMT

BlossomHill upsized to $150M at $16.00 on August 6, 2026, traded flat on debut, then ground to +37.6% by the September 4 close on Phase 1 EGFR data and a dated path.

Braveheart's $382.5M IPO: Hengrui NewCo #2 Holds Its Pop

Sat, 05 Sep 2026 00:00:00 GMT

Braveheart priced $382.5M above range for one Hengrui-in-licensed HCM drug, popped +65.6%, and holds +48.6% a month in — where Kailera round-tripped. The NewCo anatomy.

Eikon's $381M IPO: Upsized at the Top, Down 40% Since

Sat, 05 Sep 2026 00:00:00 GMT

Eikon upsized to $381M at the top of range and broke −16.7% on debut, with no negative catalyst since. The class's clearest platform discount — and its H2 2026 binary.

Generate's $400M IPO: The Worst Debut of 2026 Climbs Back

Sat, 05 Sep 2026 00:00:00 GMT

Generate priced $400M at range midpoint, fell −20.9% on debut — 2026's worst first day — and fought back above water by September on Phase 3 execution.

Hemab's $347M IPO: +88% Debut, Then the Data Kept Coming

Sat, 05 Sep 2026 00:00:00 GMT

Hemab upsized 42% to $347M, popped +88.9%, then climbed on ISTH data and an FDA-endorsed Phase 3 path in Glanzmann thrombasthenia. A +147% tape, anatomized.

Kailera's $625M IPO: The Record That Gave Back the Pop

Sat, 05 Sep 2026 00:00:00 GMT

Kailera priced a then-record $625M at the top of range, popped +62.5%, and round-tripped to +1.4% by September. The Hengrui in-license anatomy and what reprices it.

Kardigan's $460M IPO: Three In-Licensed Heart Drugs, a Modest Premium

Sat, 05 Sep 2026 00:00:00 GMT

Kardigan priced 25M shares at the $16 top of range, popped +37.5% on debut, and holds +30.3% vs offer as of Sep 4 — with all three cardiac readouts guided to H1 2027.

Latigo's $397M IPO: A Phase 3-Ready Pain Bet the Tape Bought Early

Sat, 05 Sep 2026 00:00:00 GMT

Latigo priced $345.6M at the top of range on Aug 6, 2026 — $397.4M after a full greenshoe — behind a Phase 3-ready oral NaV1.8 pain drug; +28.0% vs offer as of Sep 4.

Odyssey's Second-Attempt IPO: Broke on Debut, Up 59.8% by September

Sat, 05 Sep 2026 00:00:00 GMT

Odyssey priced an upsized $279M IPO at the top of range on May 7, broke 8.8% on debut, then re-rated +59.8% by the Sep 4 close on two earnings prints.

Parabilis's Record IPO: The Platform That Trades Like a Product

Sat, 05 Sep 2026 00:00:00 GMT

Parabilis priced a record $670M biotech IPO above range after two upsizes; twelve weeks later the tape pays for one desmoid-tumor asset, not the Helicon platform.

Scribe's $148M IPO: The Platform Story the Tape Actually Paid For

Sat, 05 Sep 2026 00:00:00 GMT

Scribe priced an upsized IPO at the $15 top on Jul 23, closed at $148M with a full shoe plus a $7.5M Sanofi placement, and trades +91.1% vs offer.

Seaport's $260M IPO: PureTech's Glyph Bet Grinds Back Above Water

Sat, 05 Sep 2026 00:00:00 GMT

Seaport upsized 20% to $254.9M at the top of range, closed up 10.2% on debut, and trades +38.1% vs offer as of Sep 4 — on prodrug chemistry, not platform adjectives.

SpyGlass's $172.5M IPO: Small Check, Steady Tape

Sat, 05 Sep 2026 00:00:00 GMT

SpyGlass priced $172.5M at midpoint with no upsize, popped +65%, and holds +81% by September on a 505(b)(2) glaucoma implant with two Phase 3s enrolling.

Veradermics' $295M IPO: A +479% Run Built on Two Data Days

Sat, 05 Sep 2026 00:00:00 GMT

Veradermics priced $295M at $17, popped +122%, then two pivotal readouts and a $442M follow-on at $100 took it to +479% of offer. The class's best tape.

Vogenx's $93M IPO: The Class's Smallest Deal More Than Doubled

Sat, 05 Sep 2026 00:00:00 GMT

Vogenx raised $93.4M all-in at $13 — the smallest 2026 biotech IPO — then ran +144.1% by the September 4 close with zero post-IPO disclosures. Float, not data.

Pharma Daily — September 5, 2026

Sat, 05 Sep 2026 00:00:00 GMT

Sep 4–5: Menarini/Gan & Lee $771M GLP-1 deal, NeuShen $80M financing, Novartis Lp(a) miss — China-originated assets took 100% of disclosed dollars.

The New Biotech IPO Class: Bigger, Later-Stage, and Judged Harder

Fri, 04 Sep 2026 00:00:00 GMT

21 US biotech IPOs raised ~$6.5B in 2026 through Aug 11 — median check $295M, 18 of 21 above offer as of Aug 28. What the class tells practitioners.

Pharma Daily — September 4, 2026

Fri, 04 Sep 2026 00:00:00 GMT

Sep 3–4: GSK-Hutchmed $110M upfront license, Medicus $1B ADC deal, ivonescimab Phase 3 OS win — China-originated assets dominated both deal flow and the tape.

Top 20 Pharma by Market Cap, Q3 2026: A $5.1 Trillion Table

Fri, 04 Sep 2026 00:00:00 GMT

As of Sep 4, 2026, the 20 largest pharma companies total $5.12T; Lilly alone is 20% of the table, Novo sits at #9, and WuXi AppTec leads YTD movers at +90%.

ADaM OCCDS Explained: Programming ADAE with Treatment Flags

Thu, 03 Sep 2026 00:00:00 GMT

ADAE from the OCCDS side: one row per event, the AE-to-ADSL merge, treatment-emergent flags driven by TRT01SDT, serious flags, and the QC defects reviewers catch.

ADaM BDS Structure: Building ADLB and ADVS Step by Step

Tue, 01 Sep 2026 00:00:00 GMT

The BDS skeleton behind ADaM analysis datasets: PARAM/PARAMCD/AVAL, baseline flags, change from baseline, and how ADVS and ADLB are built visit by visit.

SDTM AE Domain Mapping: A Worked Example from Raw EDC Data

Tue, 01 Sep 2026 00:00:00 GMT

A worked SDTM AE domain mapping example: MedDRA coding, serious flags, partial ISO dates, AESEQ derivation, SUPPAE, and CORE validation triage from a mock raw extract.

How to Write an SDTM Mapping Specification (Walkthrough)

Tue, 01 Sep 2026 00:00:00 GMT

How to write an SDTM mapping specification: column anatomy, a row-by-row VS domain walk, hygiene rules, and how specs become define-XML and machine-usable code.

SDTM Domain Basics: The Model Every Clinical Programmer Must Know

Tue, 01 Sep 2026 00:00:00 GMT

SDTM domains explained: domain classes, the topic-timing-qualifier variable pattern, USUBJID, controlled terminology, SUPPQUAL, and a day-one reading order.

ADSL Derivation Walkthrough: TRT01SDT, Flags, and the Merge Traps

Mon, 31 Aug 2026 00:00:00 GMT

How ADSL is built: deriving treatment dates and population flags from DM/EX/DS/SV, the one-row-per-subject rule, and the QC checks that catch real discrepancies.

The Clinical SP Bootcamp: A Learning Path from Raw Data to TLF

Mon, 31 Aug 2026 00:00:00 GMT

A systematic path into clinical statistical programming: CDISC fundamentals, modern SCE workflow, and where AI agents fit — every part with runnable takeaways.

Clinical SAS Interview Questions: What Actually Gets Asked

Mon, 31 Aug 2026 00:00:00 GMT

How clinical SAS interviews actually work: four rounds from SAS mechanics to GxP habits, real-format questions, and the signals strong answers contain.

Statistical Computing Environments: From SAS VM to Cloud SCE

Mon, 31 Aug 2026 00:00:00 GMT

What a statistical computing environment is for in clinical trials, why desktop SAS and shared drives failed audits, and how cloud SCE workflow changes day one.

From Mock Shell to RTF: How Clinical TLF Outputs Actually Ship

Mon, 31 Aug 2026 00:00:00 GMT

How clinical TLF outputs ship: reading mock shells, PROC REPORT tables and listings, RTF conventions, and the four-pass QC order that catches defects cheapest first.

AI Coding Assistants for Clinical Programmers: What Works in GxP

Sun, 30 Aug 2026 00:00:00 GMT

Where ChatGPT, Claude, and Copilot genuinely save time in SAS and R clinical programming, where they fail, and how to use them inside auditable GxP workflows.

CDISC CORE Explained: Open-Source Validation for SDTM/ADaM Programmers

Sun, 30 Aug 2026 00:00:00 GMT

CDISC CORE is a free, open-source rule engine for SDTM and ADaM validation. How its YAML rules work, how to run it, and where rule-based checking stops.

LLM Agents in Clinical Trials: What's Real Beyond the Demos

Sun, 30 Aug 2026 00:00:00 GMT

How to tell demo-ware from deployable agentic AI in clinical trials: determinism, audit trails, 21 CFR Part 11, constrained architectures, and vendor questions.

LLMs in Clinical Statistical Programming: Proven vs Hype (2026)

Sun, 30 Aug 2026 00:00:00 GMT

A tiered survey of LLM evidence in clinical trial statistical programming: benchmarked results, promising single-team studies, vendor hype, and open gaps for 2026–2027.

One Number That Explains Where Rule-Based Validation Stops

Sun, 30 Aug 2026 00:00:00 GMT

69.7% of oncology CORE rules port cleanly to a graph constraint language. The 30% that don't are exactly where the dangerous errors live.

Temperature 0 Doesn't Buy You Reproducibility

Sun, 30 Aug 2026 00:00:00 GMT

Setting temperature to zero feels like determinism. It isn't — and in GxP work the difference will find you during an audit, not during development.

SAS to R Migration in Clinical Programming: A Practical Field Guide

Sun, 30 Aug 2026 00:00:00 GMT

Why pharma is moving from SAS to R, what breaks in practice (procedures, XPT files, QC), and how to migrate a validated macro library without a rewrite.

The Contradictions Your Validator Can't See

Thu, 27 Aug 2026 00:00:00 GMT

Rule-based SDTM validators structurally miss cross-domain contradictions. SHACL-SPARQL graph constraints plus a deterministic agent layer catch all 20 archetypes.

Double Programming Without the Duplication: AI-Generated QC Code

Tue, 25 Aug 2026 00:00:00 GMT

Independent QC re-programming costs 30–50% of clinical programming effort. An AI framework matched 97.1–100% of variables while keeping independence intact.

Benchmarking RAG for Clinical TLF Templates: 1,999 Experiments

Sat, 22 Aug 2026 00:00:00 GMT

Five LLM generation methods benchmarked across 1,999 bootstrap experiments on ICH E3-conformant TLF templates: hybrid RAG with reranking beats direct prompting.

Why LLM Agents Fail at Regulated Programming — the Process-DAG Fix

Thu, 20 Aug 2026 00:00:00 GMT

Free-form agent loops break down in GxP clinical programming. Structuring the work as a typed process DAG makes LLM agents reliable, replayable, and auditable.

Five Years of Automation Evidence: Real Gains, Mostly Low Grades

Wed, 19 Aug 2026 00:00:00 GMT

A GRADE-rated review of 2020–2025 automation evidence in statistical programming: real gains, mostly Low to Very Low quality evidence.

Modernizing a 558-Macro SAS Library Without Touching Validated Code

Sun, 16 Aug 2026 00:00:00 GMT

A bridge map, typed IR, and orchestrator wrap a legacy SAS TFL library unchanged — AI-ready JSON on day one, 80%+ cell-level parity, optional 92% code cut.

CAVE-Onc: Graph-Constrained Agentic Validation for Cross-Domain Contradictions in CDISC Oncology Submissions

Fri, 14 Aug 2026 00:00:00 GMT

PLOS One — Regulatory submissions in oncology must remain consistent across interdependent CDISC SDTM domains, yet cross-domain contradictions routinely survive conventional validation. CAVE-Onc introduces a graph-constrained agentic validation framework that models SDTM submission datasets as an RDF knowledge graph, combining declarative SHACL-SPARQL graph constraints with a deterministic LangGraph-based agent layer to detect cross-domain contradictions that domain-scoped rule-based validators cannot express. In a pre-registered evaluation, CAVE-Onc detected all 20 clinician-reviewed injected contradiction archetypes (vs 8/20 for the CDISC CORE engine and 6/20 for the Pinnacle 21 FDA engine, both 0/10 on cross-domain RECIST contradictions) — a construction validation of expressiveness — and stayed specific (0.06–0.09 flags/subject) on two real Project Data Sphere oncology trials mapped to SDTM, detecting 10/11 and 16/18 of applicable injected archetypes. CAVE-Onc writes all validation traces to a Merkle-chained, tamper-evident audit store, providing a foundation for 21 CFR Part 11 compliance.

Thin MCP, Thick Skills: Five Layers for Clinical Programming Agents

Thu, 13 Aug 2026 00:00:00 GMT

ClinAgent splits clinical programming capability into five layers, keeping MCP tools stateless and packaging domain expertise as testable skills.

Synthetic ADaM That Survives a Join: Knowledge Graphs, LLMs, Faker

Mon, 10 Aug 2026 00:00:00 GMT

Schema-only synthetic ADaM generation plateaus at 0.45 overall quality; enriching schemas from protocol/SAP/CRF knowledge graphs plus templates reaches 0.70.

Fine-Tuning Llama 3.1 8B for admiral R Code: From 0.36 to 0.82

Fri, 07 Aug 2026 00:00:00 GMT

Base Llama 3.1 8B scores 0.36 on admiral code generation. LoRA fine-tuning plus knowledge-graph validation gets it to 0.82 — without sending data to an API.

A Non-Destructive Methodological Framework for Modernizing Legacy Clinical Reporting Systems for AI-Driven Pharmacoinformatics: A SAS Case Study

Wed, 13 May 2026 00:00:00 GMT

arXiv — Legacy SAS macro libraries resist AI integration because they emit opaque RTF with no machine-readable layer, yet source-level change triggers re-validation. This preprint presents a non-destructive framework — a bridge map, typed Intermediate Representation, and orchestrator that wrap the legacy library unchanged — applied to a 558-component industrial SAS library. Coexistence mode delivers AI-ready JSON output on day one; optional consolidation achieved a 92% SAS code reduction. Parity validation reached at least 80% cell-level parity on 11 of 14 reports from an internal Phase III study and 100% parity (4,764 cells) on the public CDISCPilot01 benchmark, and LLM experiments demonstrated IR-based table summarization, anomaly detection, and configuration generation.

ClinAgent: A Five-Layer Architecture for Autonomous Clinical Trial Statistical Programming

Fri, 16 Jan 2026 00:00:00 GMT

medRxiv — This preprint presents the five-layer ClinAgent architecture — A2UI rendering, skill router, thick skills, thin MCP tools, and compliance infrastructure — under a "Thin MCP, Thick Skills" design that augments any MCP-compatible AI coding agent (Augment Code, Claude Code, Cline, Cursor) with clinical programming capabilities rather than acting as an autonomous agent itself. Skills are independently testable, and the infrastructure layer provides audit logging and data masking for GxP compliance. Nine skills (SK-001–SK-009) were validated on a production Phase 2 cardiovascular study (11 ADaM domains, 93,239 synthetic observations): log analysis achieved 100% precision (1 error, 7 warnings over 10 logs), all 56 ADSL variables were matched, and specification generation reached 72.1% derivation accuracy overall and above 96% on simple domains.

ClinAgent: AI-Assisted Methodology for Clinical Trial Data Processing and Statistical Programming

Thu, 01 Jan 2026 00:00:00 GMT

Biology Methods and Protocols — Clinical trial statistical programming remains labor-intensive: ADaM dataset derivation, TLF generation, and QC programming consume substantial effort per submission. ClinAgent is a skill-and-tool layer that augments any MCP-compatible AI coding agent with clinical-programming capabilities: nine skills (SK-001 Study Setup through SK-009 eSub Packaging) package prompts, deterministic rule engines, and decision trees, while thin MCP tools provide stateless I/O for SAS datasets, Excel specifications, and log files. All nine skills passed functional validation on artifacts from a single production Phase 2 study — deterministic components matched all 56 ADSL variables, and prompt-based specification generation reached 72.1% derivation accuracy (>96% on simple domains, <55% on complex domains, with wide confidence intervals making these point estimates upper bounds) — while end-to-end productivity gains were not measured and remain future work. Human review gates and deterministic audit trails preserve the traceability and reproducibility required for regulatory submissions.

Eliminating QC Programming Duplication Through Claude AI-Assisted Independent Code Generation: A Practical Framework for Regulatory-Compliant Validation

Thu, 01 Jan 2026 00:00:00 GMT

PharmaSUG 2026 — Independent QC programming — recreating production programs from specifications alone — accounts for an estimated 30–50% of total clinical programming effort. This paper presents a Claude AI-assisted workflow that generates independent QC code in Python directly from ADaM specifications, combining a QC Trace Tree, a Decision Router, Agent Skills, and an automated code review engine to reduce duplication between production and QC streams while preserving operational independence through isolated AI instances, with the human QC programmer as the genuinely independent review layer. On the CDISCPilot01 benchmark, the framework achieved 97.1%–100% variable-level match across five ADaM domains and passed all 13 assertions. Companion code is publicly available.

Schema-Preserving Generation of Clinical TLF Templates and Executable R Code via Iterative LLM-Guided Debugging

Thu, 01 Jan 2026 00:00:00 GMT

PharmaSUG 2026 — Manual authoring of CSR TLF templates is resource-intensive, and naive LLM prompting suffers schema drift and regulatory inconsistencies. This paper compares five LLM generation methods for producing ICH E3/CDISC-conformant TLF templates across 1,999 instance-matched bootstrap experiments on three LLM providers, using JSON Patch (RFC 6902) to preserve schema fidelity, and separately evaluates iterative LLM-guided debugging for translating the templates into executable R code. A hybrid RAG approach with reranking significantly outperformed direct prompting (mean quality score 85.7 vs 81.7, p < 0.05, consistent across providers and therapeutic areas), and iterative LLM-guided debugging raised R-code execution success from a low zero-shot rate to 70% within 3–5 rounds, with higher-fidelity templates needing fewer iterations.

GxP-Agent: Process-DAG Topology for Reliable Clinical Trial Programming with LLM Agents

Thu, 01 Jan 2026 00:00:00 GMT

arXiv — LLM agents fail in regulated programming work when their execution is unstructured. GxP-Agent organizes clinical trial programming as a process DAG — typed nodes for data ingestion, derivation, validation, metadata application, and export — so that agent behavior is constrained, replayable, and auditable. The topology yields measurably higher task reliability than free-form agent loops and produces the evidence trails GxP review requires.

Evidence Behind the Automation of Clinical Trial Statistical Programming: A Scoping Review of Technology Adoption, Validation Frameworks, and AI/ML Integration (2020–2025)

Mon, 29 Dec 2025 00:00:00 GMT

medRxiv — This scoping review (reported per PRISMA-ScR) maps the 2020–2025 evidence base on automating clinical trial statistical programming — from macro-based tooling to LLM-driven approaches — characterizing study types, claimed efficiency gains, evaluation rigor, and open gaps. From 1,247 records, 262 studies were included; reported gains include 15–25% development-time reduction for pharmaverse TLF tools, 30–50% effort reduction for risk-based validation with CI/CD, and 75–85% SDTM conversion-time reduction for REDCap2SDTM, alongside 88–93% F1 for domain-specific LLMs on clinical NLP versus 60–85% code-generation accuracy for general models. Evidence quality is predominantly Low to Very Low: only 12 of 527 validation papers (2.3%) report quantitative outcomes, and no RCTs comparing validation approaches exist, defining the critical research priorities for the field.

An End-to-End Approach to Fine-Tune Small LLMs for Generating Admiral R Code in Statistical Programming

Wed, 01 Jan 2025 00:00:00 GMT

PhUSE US Connect 2025 — Base small LLMs perform poorly on Admiral code generation (OPS 0.36), and even strong general-purpose LLMs offer no built-in correctness verification for ADaM compliance. This paper presents an end-to-end pipeline for fine-tuning small, locally deployable LLMs on curated Admiral examples — fine-tuning LLaMA 3.1 8B with LoRA and validating output through a knowledge-graph system built from Admiral documentation. The pipeline covers data preparation, training, and evaluation, achieving reliable ADaM code generation without sending proprietary data to external APIs.

Automating SAS and R Code Interpretation and Debugging: A Practical Pipeline for Statistical Programmers

Wed, 01 Jan 2025 00:00:00 GMT

PhUSE US Connect 2025 — Statistical programmers spend substantial effort writing and debugging SAS and R code. This paper describes an LLM-based pipeline that translates natural-language queries into executable SAS, R, or Python code, runs it in a Jupyter kernel, captures runtime errors from the kernel, and feeds them back to the LLM for iterative repair until the code succeeds or a retry limit is reached, exposed through a Gradio-based user interface. Evaluated on a factorial corpus of 2,700 runs across Python, R, and SAS with injected syntax, runtime, and logical/statistical errors, the system achieved success rates above 96% across languages, resolving about 95% of syntax and 90% of runtime errors versus about 88% and 87% for logical and statistical errors.

Enhancing Clinical Trial Data Queries with LLMs and Neo4j: A Flexible Framework for ADaM Dataset Management

Wed, 01 Jan 2025 00:00:00 GMT

PhUSE US Connect 2025 — This paper models CDISC ADaM datasets in a Neo4j graph database via a reusable graph schema derived from ADaM specification files, and uses LLMs with LangChain to translate natural-language questions into both Cypher and SQL queries, making data checking more flexible, particularly for ad hoc tasks. Benchmarked head-to-head on 90 protocol-derived questions over synthetic ADaM datasets (ADSL, ADAE, ADCM, ADLB, ADVS) held in both PostgreSQL and Neo4j, LLM-generated Cypher achieved higher accuracy than SQL (73.33% vs 66.67% overall), with the advantage holding across difficulty tiers.

A Novel Pipeline for Generating Realistic Synthetic CDISC ADaM Datasets Using Large Language Models and Knowledge Graphs

Wed, 01 Jan 2025 00:00:00 GMT

PhUSE US Connect 2025 — Creating synthetic ADaM data that faithfully represents real trial characteristics is challenging. This paper combines knowledge graphs built from clinical trial documentation (protocols, SAPs, CRFs) with Faker-based generation, using LLMs to enrich and reorganize the JSON schemas that drive data generation. A three-way comparison of direct JSON, LLM-enhanced JSON, and template-based generation shows the template-based variant achieving the highest overall quality score (0.70 vs 0.45 for direct JSON-schema generation), with improved structural integrity and cross-dataset relationships.

GenAI Assisted Code Conversion: From SAS to R Standard ADaM Templates

Wed, 01 Jan 2025 00:00:00 GMT

PharmaSUG 2025 — As clinical trial Analysis & Reporting shifts toward open-source languages, our organization needed to convert its SAS-based ADaM standards library to R. This paper presents a GenAI-assisted conversion methodology using a structured workflow of source pre-processing, logical code segmentation, and iterative prompt engineering that accumulates conversion instructions from previously converted templates, followed by human validation comparing converted R outputs against the validated SAS-generated ADaM datasets. Using ChatGPT-4o and Claude 3.5 Sonnet, the effort converted about 78% of the SAS ADaM templates, with about 66% of the converted code used without modification, though macro-heavy templates still required significant human intervention.

Comparing SQL and Graph Database Query Methods for Answering Clinical Trial Questions with LLM-Powered Pipelines

Wed, 01 Jan 2025 00:00:00 GMT

PharmaSUG 2025 — Should clinical data exploration use SQL or graph databases? This paper benchmarks both query paradigms behind LLM-powered natural-language pipelines on 150 representative clinical trial questions over synthetic ADaM data, comparing accuracy, latency, and the classes of questions each paradigm handles well. The proposed RagQL-Nav framework — query decomposition, intelligent routing, and dual-query validation — reached 91% accuracy on complex queries, roughly 12 points above single-system approaches, offering practical guidance on when graph approaches earn their complexity.

AI-Enhanced Chatbot for Streamlined Clinical Trials Analysis and Document Management

Mon, 01 Jan 2024 00:00:00 GMT

PhUSE US Connect 2024 — Statistical programming teams lose hours searching specifications, protocols, and analysis documentation. This paper describes a retrieval-augmented chatbot that indexes clinical trial documents and analysis artifacts and answers natural-language questions against the retrieved content, with a local-deployment option that keeps confidential documents and data in-house. A fine-tuned LLM also maps mock table shells to SAS macro calls for end-to-end TLF generation, with a SAS kernel executing the generated code.

LLM-Enhanced Training Agent for Statistical Programming

Mon, 01 Jan 2024 00:00:00 GMT

PharmaSUG 2024 — This paper presents an LLM-enhanced training agent for statistical programming, implemented as a custom GPT-4 GPT built via prompt engineering with uploaded reference documents (SDTMIG, ADaMIG, ICH E9, and internal SAS guidance) and a slash-command interface. The agent guides learners through personalized curricula, lessons, and real-world simulation projects, tests their knowledge with quizzes and coding exercises, and gives immediate, personalized feedback, with positive feedback from a pilot testing phase.

Methodology for Automating TOC Extraction from Word Documents to Excel

Mon, 01 Jan 2024 00:00:00 GMT

PharmaSUG 2024 — Programmers tracking clinical study deliverables often transcribe the table of contents of SAP mock-shell Word documents into Excel trackers by hand. This paper presents an Excel-based VBA tool — chosen because company policy bars external tools — that extracts the TOC from Word mock-shell documents into an Excel tracker at the click of a button, locating the TOC field programmatically and stripping page numbers from entries. The resulting trace system supports monitoring each study deliverable’s progression and validation status.

JSON Data Generation: Linking Statistical Analysis with Large Language Models

Mon, 01 Jan 2024 00:00:00 GMT

PHUSE/FDA Computational Science Symposium (CSS) 2024 — This poster presents a framework that extracts schema from structured clinical datasets (e.g., CDISC ADaM) into JSON, uses it to design a graph-based database (Neo4j) with graph embeddings stored in a vector database, and enables LLMs to answer natural-language questions by retrieving relevant variables and generating SQL or Cypher queries — automating clinical data analysis workflows.

A Macro Utility for CDISC Datasets Cross Checking

Sun, 01 Jan 2023 00:00:00 GMT

PharmaSUG 2023 — CDISC validation tools such as Pinnacle 21 check standards compliance but not dataset content, and A&R programmers need study- and therapy-area-specific checks throughout the analysis lifecycle. This paper presents a configurable SAS macro utility that systematizes cross checking across SDTM and ADaM datasets; its modular sub-macro framework lets users plug in custom checking rules, extensible beyond CDISC datasets. The macro generates a formatted Excel discrepancy report with a hyperlinked summary tab and one sheet per issue category, supporting faster data review cycles.

Tips to Read In and Output Excel Spreadsheets in SAS

Sun, 01 Jan 2023 00:00:00 GMT

PharmaSUG 2023 — Excel is widely used in clinical trials to save and transfer data, and analysis & reporting programmers routinely receive spreadsheets and produce customized Excel reports for other departments. This paper collects practical techniques for reading different Excel formats (xls, xlsx, xlsm, xlsb, csv) into SAS via LIBNAME engines and PROC IMPORT — including how to choose between them by environment and format — and for writing well-formatted outputs with PROC EXPORT and ODS EXCEL. It offers practical decision tips for choosing import/export methods and output formats.

A Framework for Interactive Ad-hoc Request Handling: Empowering Clinical Insights through Interactive Plots

Sun, 01 Jan 2023 00:00:00 GMT

PHUSE/FDA Computational Science Symposium (CSS) 2023 — This poster presents a framework for producing interactive plots to handle ad-hoc analysis requests in the clinical domain. It standardizes the end-to-end process — Anaconda-based environment setup, package version control via renv/pipenv lock files, Jupyter Lab as the IDE, interactive visualization with Plotly (Python) and ggplot2 plus Plotly/htmlwidgets (R), and portable self-contained HTML output — with results validated by crosschecking between SAS, R, and Python. Example outputs include swimmer, waterfall, and Kaplan-Meier plots with drop-downs and range sliders.