After the last patient visit and the final data query, a clinical trial enters one of its least visible but most consequential stretches: the gap between database lock and regulatory submission. Biostatisticians hand-program, validate, and double-program the tables, listings, and figures (TLFs) that regulators actually read, a process that routinely consumes weeks per cycle. Industry analyses place the cost of each lost day at roughly half a million dollars in unrealized sales.
The stakes climb higher in fatal, rapidly progressing diseases such as amyotrophic lateral sclerosis (ALS), where median survival after diagnosis runs two to five years. Sponsors typically respond by adding programmers, a remedy that scales poorly when the volume of required analyses expands late in a filing.
Veristat, a global clinical research organization (CRO) and consultancy specializing in complex studies, announced that its InStat biometrics platform cut the time to deliver validated biostatistical outputs for Clene Nanomedicine's pivotal ALS drug application by 80 percent. The interval from locked study data to submission-ready TLFs fell from up to six weeks to one week. Clene is a late-stage biotechnology company.
ALS accelerated approval: NfL analyses on the critical path
Clene's October 2026 accelerated approval NDA, a New Drug Application for CNM-Au8 in ALS filed under the US Food and Drug Administration's (FDA) Subpart H pathway, rests heavily on neurofilament light (NfL) biomarker analyses. NfL, a blood-based marker of neuronal injury, anchors the submission because the FDA indicated in a Type C meeting that the biomarker could potentially serve as a reasonably likely surrogate endpoint in ALS. The NfL analysis package, not a conventional clinical endpoint readout, therefore formed the statistical center of gravity of the application.
When the volume of required analyses expanded beyond the original plan, InStat allowed Clene to absorb the added work and still submit in 2026. The engagement spanned a dozen deliverables, each governed by its own statistical analysis plan; Veristat reported comparable results on two studies for a second sponsor.
"Veristat moved at unprecedented speed, eliminating weeks of hand-programming that is also the source of most errors," said Rob Etherington, CEO of Clene Nanomedicine. "Every minute from database lock to submission matters. InStat enabled us to move from locked data to reviewable NfL biomarker tables in days, allowing for concurrent review and verification of the outputs instead of waiting until the end of a long cycle. That kind of speed, with the rigor we require, is what lets us keep pace with the urgency of ALS."
How does InStat automate TLFs without LLM-generated code?
InStat does not ask a large language model (LLM) to write analysis programs. Artificial intelligence (AI) agents instead translate biostatisticians' descriptions into machine-readable specifications. The design aims to deliver automation speed without the regulatory exposure of LLM-generated analysis code:
- Zero-code specification: AI agents convert biostatistician descriptions into formal analysis specifications
- Validated engines: a pre-built library of statistical engines, not generated code, performs every computation
- Traceability: each output links back to its specification for regulatory audit
- Expert review: Veristat biostatisticians review every deliverable before release
- System-agnostic design: sponsors keep their preferred electronic data capture system and formatting standards
- One-week turnaround: reported delivery of submission-ready TLFs within days of database lock
"Everyone hoped AI would bring speed, but quality may be the bigger story," said Kyle McBride, Vice President of AI & Innovation at Veristat. "In every side-by-side comparison with traditional hand programming, any discrepancies traced back to the manual side, never to InStat. An error in an NDA puts an approval at risk. While much of the industry is still working on either speed, quality, or compliance, we've solved all three. Faster used to mean riskier. Not anymore, and that means new therapies can reach patients sooner."
A one-week TLF cycle changes submission economics
Compressing the TLF cycle from six weeks to one carries a quantifiable economic value, according to independent, peer-reviewed estimates rather than vendor claims. In a 2024 analysis in Therapeutic Innovation & Regulatory Science, lead author Zachary P. Smith, coauthor Joseph A. DiMasi, and senior author Kenneth A. Getz of the Tufts Center for the Study of Drug Development recalculated delay-day costs at approximately $500,000 per day in unrealized sales, replacing a decades-old $4 million estimate. The same analysis put direct trial conduct costs at $40,000 per day on average and above $55,000 per day for Phase 3 studies.
By those estimates, five weeks recovered on the critical path represents on the order of $17.5 million in earlier-realized sales value. The Tufts authors cautioned that the per-day figure declines by roughly $80,000 each year. The table below summarizes how the automated and manual workflows compare.
| Attribute | Traditional manual TLF programming | InStat automated workflow |
|---|---|---|
| Time from database lock to validated TLFs | Up to six weeks per cycle | Approximately one week |
| Code source | Hand-programmed and double-programmed by statisticians | Pre-validated statistical engine library |
| Specification process | Manual statistical programming specifications | AI agents convert biostatistician descriptions into specifications |
| Review model | Sequential review at end of programming cycle | Concurrent review and verification of outputs |
| Reported error pattern in side-by-side comparisons | Discrepancies traced to manual programming | No discrepancies attributed to platform, per Veristat |
Company performance claims, including the 80 percent reduction and the side-by-side error comparisons, reflect Veristat's reported results for this engagement and have not been independently published.
Automated TLFs kept Clene's ALS accelerated approval on schedule
Veristat's InStat platform cut validated TLF delivery for Clene's CNM-Au8 ALS submission from up to six weeks to one, an 80 percent reduction that let an expanded NfL analysis package reach the FDA in 2026. The approach pairs AI-written specifications with pre-validated statistical engines rather than LLM-generated code, keeping every output traceable and biostatistician-reviewed. With the FDA's review decision ahead, a Phase 3 confirmatory study planned for 2027, and InStat already deployed on a second sponsor's studies, the collaboration offers an early model for how automated biostatistics can shorten the path from database lock to regulatory submission.
This article is based on a press release issued by Veristat and was produced under Drug Discovery News' AI Editorial Guidelines.









