Overview
Stratified Permuted Block Design extends Permuted Block Randomization by maintaining separate randomization lists for each unique combination of stratification factors and site. This guarantees that treatment arms are balanced not only overall, but also within each clinically meaningful subgroup.
It is the most common algorithm for double-blind, multi-site Phase II/III confirmatory trials.
How it works
- All defined stratification factors (e.g., NYHA Class, LVEF Method) are combined with the site identifier to form a stratum key.
- A separate permuted block list is maintained for each stratum key.
- When a subject is randomized, the system reads their stratification variable values and site, looks up the correct stratum list, and assigns the next available allocation.
Example — 2 stratification factors, 2 sites:
| Stratum key | Separate block list |
|---|---|
| Site 01 / NYHA II / Simpson's Biplane | [TRT, PBO, TRT, PBO, …] |
| Site 01 / NYHA II / Teichholz | [PBO, TRT, PBO, TRT, …] |
| Site 01 / NYHA III / Simpson's Biplane | [TRT, TRT, PBO, PBO, …] |
| Site 02 / NYHA II / Simpson's Biplane | [PBO, TRT, TRT, PBO, …] |
| … | … |
Each stratum is balanced independently, so the final analysis can adjust for the stratification factors and site with full statistical validity.
When to use
Use Stratified Permuted Block Design when:
- The trial has multiple sites and site-level balance is required.
- One or more baseline characteristics (prognostic factors) need to be balanced across arms.
- The protocol specifies stratified randomization (common in Phase III trials).
- The regulatory submission requires proof of balance on key covariates.
Configuration
In the Randomization app, select Stratified Permuted Block Design as the algorithm, then configure the settings below.
Algorithm settings
| Setting | Description |
|---|---|
| Study Groups | One row per treatment arm: Label, Code, Weight, Description |
| Block size | Subjects per block. Use 4 or 8 for 1:1 allocation |
| Generate randomization list | Real-time (recommended) — a new block is generated when needed |
| Level of stratification | By Site — separate lists per site (most common) or By Study — single lists across all sites |
Stratification Factors
After saving the main settings, add one row per stratification variable in the Stratification Factors Information section. For each factor:
| Field | Description |
|---|---|
| Stratification From | Source type — select A CRF |
| CRF / Default Version | The CRF containing the variable (e.g., CRF 4 - NYHA Classification - V.1.0) |
| Visit / Event | The event when the data is collected (e.g., Screening) |
| Field | The Radio/Checkbox field name within the CRF |
| Options | Auto-populated from the field's answer choices |
Prerequisite: Only Radio and Checkbox CRF fields can be used as stratification variables. The field must exist in the CRF before you can select it here. See Form Builder for field type guidance.
Worked example — CARDIO-RELIEF 2025
| Setting | Value |
|---|---|
| Blinding Method | Double Blind |
| Algorithm | Stratified Permuted Block Design |
| Study Groups | CardioRelief CRF-101 (TRT, weight 1) + Placebo (PBO, weight 1) |
| Block size | 4 |
| List generation | Real-time |
| Level of stratification | By Site |
| Randomization event | Randomization / Baseline |
Stratification Factors:
| Stratification From | Visit | Field | Options |
|---|---|---|---|
| CRF 4 - NYHA Classification - V.1.0 | Screening | NYHA Class | Class I, Class II, Class III, Class IV |
| CRF 6 - Echocardiography (LVEF) - V.1.0 | Screening | LVEF Method | Simpson's Biplane, Teichholz, Visual Estimate, 3D Echo |
This configuration creates up to 4 × 4 × (number of sites) strata. For a 10-site trial, that is up to 160 separate block lists.
Number of strata
Total strata = (options in factor 1) × (options in factor 2) × … × (number of sites, if By Site)
| Factors | Options | Sites | Total strata |
|---|---|---|---|
| 1 factor, 3 options | 3 | 5 | 15 |
| 2 factors, 4 options each | 4 × 4 = 16 | 10 | 160 |
| 3 factors, 3 options each | 3 × 3 × 3 = 27 | 8 | 216 |
Important: Many strata remain under-enrolled in most trials. Each stratum with an odd number of subjects at study close will be slightly imbalanced. Limit stratification factors to those that are strong prognostic factors per the statistical analysis plan.
Regulatory acceptance
Stratified randomization is endorsed by ICH E9 §3.4 and is the standard approach in most Phase III confirmatory trials. The stratification factors and block sizes used must be pre-specified in the protocol and randomization SOP before trial start.