Sutrix

Biostatistics cores at hospitals, research networks and academic centers receive trial and chart review data as inconsistently shaped Excel files from investigators, requiring manual visual checks, custom SAS or R scripts, and cleaning that alone consumes a quarter of a study's time. Research centers with many investigators but few biostatisticians queue requests for one to two weeks before a first meeting, with further rounds of email before a report returns, making time the top complaint from investigators. A single multi-arm outcomes study with patient-reported outcomes can require three people and several weeks to produce a plan, analysis code, an independent check and a write-up. Before any protocol or grant is written, someone also has to manually search public registries and shared trial datasets across dozens of sites, a process that can take a week.

Sutrix runs each study through a fixed, ordered workflow: intake and audit, codebook creation, data cleaning, analysis plan, execution, independent validation, and report. Every upload is hashed and screened for 18 identifier classes, cleaning operations are listed with counts and require approval before applying, and the analysis plan is extracted from the protocol and signed with a fixed seed before anything runs. A second, independent derivation regenerates every result from the approved plan without reading the analysis code to reconcile against the first, and every number in the narrative is checked against results. Statisticians retain control at five approval gates (cleaning proposal, dataset, plan, validation, final report), with 41 cited statistical-practice rules marking what blocks, what must be declared, and what is advisory. The system connects to REDCap, Qualtrics, CSV, Excel and SAS transport files, plus public datasets like NHANES, and seals a reproducibility bundle (code, environment lock, hashes, reproduce script) to each report.