C*Power — Historical Control Borrowing Calculator

The Problem

Every animal study demands a concurrent control group — subjects that receive no treatment, serving as the living baseline against which all effects are measured. This requirement is both scientifically essential and legally mandated.

Yet your laboratory has already conducted many similar experiments. You possess data from dozens — perhaps hundreds — of control animals observed under nearly identical conditions. Can that accumulated evidence reduce the number of new control animals your next study requires?

The answer is: yes, sometimes, and by a precisely quantifiable amount — but only when the right methodology is applied. Guesswork will not suffice. This tool provides the rigorous framework to determine that amount.

BfR Recommendation 013/2026 (20 March 2026) establishes that when historical controls have demonstrated stability over time, the number of concurrent control animals may be reduced. However, concurrent controls can never be eliminated entirely. This tool operationalizes that recommendation through a concrete statistical framework.

The 3R Principle — Reduce. We are bound by an ethical imperative to use the minimum number of animals necessary. But "minimum" must be determined by science — not by convenience, intuition, or wishful thinking.

How It Works

1

Enter your data

Summary statistics (n, mean, SD) from each previous matching control group in your lab — one row per study, control animals only.

2

Assess consistency

The tool estimates how much your controls vary between studies — the between-study heterogeneity τ. This single parameter governs how much you may borrow.

3

Get your answer

A concrete number: how many concurrent controls your next study requires, and how many animals are spared.

Key Concepts

Ancient principles, modern application — the foundational ideas behind historical borrowing.

τ

Between-Study Heterogeneity (τ)

Even in the same lab with the same protocol, results shift between experiments. Tau measures how much. Higher τ means less you can borrow from the past — a natural governor on overconfidence.

n_eff

Effective Sample Size (ESS)

Your 80 historical control animals are not worth 80 concurrent ones. After accounting for between-study variability, perhaps only 25 "count." ESS quantifies this discount — the true currency of borrowed evidence.

M

The Borrowing Engine

A principled statistical method that automatically borrows less when historical data are inconsistent. No subjective tuning required; the data speak for themselves. (Technically: the Meta-Analytic Predictive prior, or MAP prior.)

R

Robustification

An additional safeguard: mixing in an uninformative component to protect against the unexpected — the scenario where past and future diverge more than the model anticipates.

When This Tool Cannot Help

Historical borrowing requires that past and future experiments are exchangeable — drawn from the same underlying reality. The tool will not produce valid results if:

Different lab, strain, or age range — historical data must come from the same laboratory, species/strain, and comparable age group
Data older than 5 years — per OECD TG 451 (the OECD guideline for carcinogenicity studies, widely adopted as best practice for historical control data quality), conditions drift over time and distant data lose relevance
Fewer than 3 historical studies available for automatic τ estimation (5+ recommended). With only 1–2 studies, the tool switches to manual τ mode — you may proceed, but must assume a τ value and document that assumption
Control types don't match — e.g., using untreated controls to inform a vehicle control group (different baselines due to handling or injection stress)
Known temporal drift in the endpoint — seasonal effects, colony genetic drift, or other systematic time trends that violate exchangeability
Elimination of concurrent controls entirely — this is never valid, regardless of how consistent your historical data appear