BRAVE: The Backbone of Sample Size Calculation

In clinical research, one of the most common questions is: How many participants do I need? The answer is rarely a fixed number. Instead, the study size must be determined according to the primary research objective. For comparative studies, the core statistical framework that guides this process can be summarized by the mnemonic BRAVE.

What is BRAVE?

BRAVE represents the five key statistical components used to estimate sample size for studies that compare groups or test hypotheses:

B — Beta / Power

Beta is the probability of a Type II error, meaning the study fails to detect a real difference when one truly exists. Statistical power is defined as:

Power = 1 - β

In most clinical studies, power is set at 80% or 90%.

R — Ratio

Ratio refers to the allocation ratio between study groups, such as 1:1 or 1:2. If one group is more difficult to recruit, an unequal ratio may be used. However, unequal allocation usually requires a larger total sample size to preserve the same statistical power.

A — Alpha

Alpha is the probability of a Type I error, meaning the study concludes that a difference exists when it actually does not. In most clinical research, alpha is conventionally set at 0.05 for a two-sided test.

V — Variability

Variability reflects the expected spread of the outcome measure, usually expressed as the standard deviation (SD) or variance for continuous outcomes. Greater variability makes it harder to detect a true difference and therefore requires a larger sample size.

E — Effect Size

Effect size is the smallest clinically meaningful difference that the researcher wants the study to detect. The smaller the expected effect, the larger the sample needed to demonstrate it.

BRAVE Across Different Analysis Strategies

Although BRAVE is the main framework for hypothesis-testing studies, its role changes depending on whether the objective is descriptive, comparative, or predictive.

1. Descriptive Strategy: Focus on Precision

For descriptive objectives, such as estimating prevalence, incidence, or a mean value, the goal is not to test a hypothesis but to estimate a parameter with acceptable precision.

What is used

Researchers perform a precision analysis, not a traditional power analysis.

Key components

The calculation is usually based on:

Confidence level
Expected variability
Margin of error

What is missing

There is no beta/power and no group ratio, because there is no comparison between groups.

In this setting, the “E” concept does not mean effect size between groups; instead, it refers more closely to the desired precision of the estimate.

2. Comparative Strategy: Explore vs. Explain

This is the main setting in which the full BRAVE framework is applied. However, the meaning of effect size depends on the type of comparative question being asked.

Explanatory (Causal) Research

The aim is to determine whether an exposure or intervention causes an outcome. In this context, the effect size should be based on:

Previous causal literature
Pilot data
The minimal clinically important difference (MCID)

Here, BRAVE is used in its classic form: to ensure the study has enough power to detect a clinically meaningful effect.

Exploratory Research

The aim is to identify which factors may be associated with an outcome. In this context, researchers are often less focused on powering the study for one single exposure and more concerned with achieving a sample size large enough to support:

Stable estimates
Multivariable analysis
Broad exploration of candidate variables

Thus, BRAVE may still inform the study, but practical and modeling considerations often become equally important.

3. Predictive Strategy: Beyond BRAVE

Predictive research, such as the development of a clinical prediction model, follows a different logic. The goal is not to test whether one variable differs between groups, but to build a model that performs well in new patients.

BRAVE is not the main framework

Modern guidance does not recommend relying on traditional power analysis, p-values, or BRAVE alone for prediction studies.

What is used instead

Sample size is guided by performance analysis, with attention to:

Number of candidate predictors
Number of outcome events
Model complexity
Expected model performance
Calibration and shrinkage
Risk of overfitting

Older rules such as 10 events per variable (EPV) are widely used as rough heuristics, but more modern approaches recommend using model-based criteria, such as those proposed by Riley and colleagues, because they better address prediction error and model optimism.

In predictive studies, the central question is not, “Can I detect a statistically significant effect?” but rather, “Can I build a model that will perform reliably in future patients?”

Summary Table

Objective	Strategy	Main Focus	BRAVE Usage
Descriptive	Precision Analysis	Confidence intervals and margin of error	Partial use of alpha and variability
Comparative	Power Analysis	Statistical significance and clinically meaningful difference	Full BRAVE framework
Predictive	Performance Analysis	Overfitting, calibration, discrimination, generalizability	Not based on BRAVE

Key Message

A defensible sample size begins with a clear understanding of the study’s primary objective.

If the aim is to describe, focus on precision.
If the aim is to compare, use BRAVE.
If the aim is to predict, focus on performance.

In short:

Descriptive studies seek precision. Comparative studies must be BRAVE. Predictive studies require performance-based planning.