Randomization for bioequivalence & crossover trials

Randomization.

Generate verifiable, reproducible randomization schedules for bioequivalence, bioavailability, and crossover trials — including replicate designs for highly variable drugs. Every schedule carries a seed, algorithm version, and SHA-256 verification code so an independent auditor can reproduce it. Runs entirely in your browser; your data never leaves this device.

Already have a schedule? Verify it →

tool 01
client-side only
verifiable

Study parameters

configure

Design

Method

Core randomization

Total volunteers

Drugs

Seed

Block size

Allocation ratio

Audit metadata

optional

Study / protocol code

Sponsor / CRO

Protocol version

Protocol date

Generated by

Recorded in the audit header and included in the verification code — changing the name changes the code.

Checked by

Numbering & blinding

optional

Reserve subjects

Stratification (optional)

Single-axis: name:n, comma-separated — e.g. Male:12, Female:12. Multi-axis (factorial): separate factors with | and give a per-cell count — e.g. Site: A, B, C | Sex: M, F | n: 8 makes 6 strata of 8. Each stratum is randomized independently and balanced on its own. Overrides total volunteers.

Numbering & reserves

Rand. no prefix

Start

Digits

Separate enrollment number

Methodology & validation

reference

Algorithm

Allocation uses a seeded, platform-independent mulberry32 pseudo-random number generator (RNG_VERSION 1.0) driving a Fisher–Yates shuffle. The same seed and parameters produce a bit-for-bit identical schedule on any device or browser — no Math.random(), no server, no network. The seed is recorded with every schedule.

Balance

Subjects are assigned in permuted blocks so each sequence (crossover) or arm (parallel) receives an equal share within every block. Williams designs are variance-balanced: each treatment appears once per period and each ordered pair of treatments is adjacent equally often, neutralising first-order carryover. Replicate designs (partial TRR/RTR/RRT, full TRTR/RTRT) follow their canonical templates for HVD / reference-scaled analyses.

Known-answer tests

A committed test suite pins the RNG output and full schedules to fixed expected values (e.g. seed 12345 on a 2×2×2 design yields a known verification code), so any unintended change to the algorithm is caught before release.

Verification & reproduction

Each schedule is serialised to a canonical string and hashed with SHA-256 (Web Crypto). The first 16 hex characters are shown as a verification code on screen, in the PDF, and in every export filename. An independent auditor can re-enter the seed and parameters, regenerate the list, and confirm the identical hash — making the schedule tamper-evident.

Version

Tool v1.0.0 · RNG mulberry32 v1.0. Algorithm and version are stamped on every schedule and export.