How to Calculate Confidence Interval Non-Inferiority

Non-inferiority testing is a statistical method used to determine whether a new treatment or product is not significantly worse than an existing standard. This guide explains how to calculate confidence intervals for non-inferiority testing, including the formulas, assumptions, and practical applications.

What is Non-Inferiority Testing?

Non-inferiority testing is used in clinical trials and product comparisons to ensure that a new intervention is not significantly worse than an established standard. Unlike superiority testing, which looks for improvements, non-inferiority testing focuses on demonstrating that the new treatment is at least as effective as the standard.

The key concept is the non-inferiority margin (δ), which represents the maximum acceptable difference between the new treatment and the standard. If the upper bound of the confidence interval for the difference between treatments is below δ, the new treatment is considered non-inferior.

Confidence Interval Formula

The confidence interval for the difference between two means (μ₁ - μ₂) is calculated using the formula:

(μ₁ - μ₂) ± t*(s√(1/n₁ + 1/n₂))

Where:

μ₁ = mean of the new treatment
μ₂ = mean of the standard treatment
t = critical t-value from t-distribution
s = pooled standard deviation
n₁ = sample size for new treatment
n₂ = sample size for standard treatment

The pooled standard deviation is calculated as:

s = √[( (n₁-1)s₁² + (n₂-1)s₂² ) / (n₁ + n₂ - 2)]

Where s₁ and s₂ are the standard deviations of the two groups.

How to Calculate Non-Inferiority CI

Determine the sample means (μ₁ and μ₂) and standard deviations (s₁ and s₂) for both treatments.
Calculate the pooled standard deviation using the formula above.
Find the critical t-value for your desired confidence level and degrees of freedom (n₁ + n₂ - 2).
Calculate the margin of error using the t-value, pooled standard deviation, and sample sizes.
Construct the confidence interval by adding and subtracting the margin of error from the difference in means (μ₁ - μ₂).
Compare the upper bound of the confidence interval to the non-inferiority margin (δ). If the upper bound is less than δ, the new treatment is non-inferior.

Note: Non-inferiority testing requires careful planning of the non-inferiority margin (δ) based on clinical relevance and practical considerations.

Example Calculation

Suppose we have two treatments:

New treatment: mean = 70, standard deviation = 10, sample size = 30
Standard treatment: mean = 75, standard deviation = 8, sample size = 30

Non-inferiority margin (δ) = 5

Calculations:

Pooled standard deviation: s = √[(29×10² + 29×8²)/58] ≈ 9.17
Degrees of freedom = 30 + 30 - 2 = 58
For 95% confidence, t-value ≈ 2.002
Margin of error = 2.002 × 9.17 × √(1/30 + 1/30) ≈ 4.32
Confidence interval: (70-75) ± 4.32 → (-5, -9.32)

The upper bound of the confidence interval (-9.32) is less than the non-inferiority margin (-5), so we conclude the new treatment is non-inferior.

Interpreting Results

A non-inferiority confidence interval provides several key pieces of information:

The estimated difference between treatments
The precision of that estimate (width of the interval)
Whether the upper bound of the interval is below the non-inferiority margin

If the upper bound of the confidence interval is below δ, we can conclude with the specified confidence level that the new treatment is not significantly worse than the standard. If the upper bound is above δ, we cannot conclude non-inferiority.

Common pitfalls include:

Choosing an inappropriate non-inferiority margin
Assuming non-inferiority implies equivalence
Ignoring the precision of the estimate (interval width)

FAQ

What is the difference between non-inferiority and equivalence testing?

Non-inferiority testing shows that a new treatment is not worse than a standard, while equivalence testing shows that the new treatment is neither better nor worse. Non-inferiority is often used in regulatory settings where showing the new treatment is not worse is sufficient.

How do I choose the non-inferiority margin?

The non-inferiority margin should be based on clinical relevance and practical considerations. It represents the maximum acceptable difference between treatments that would still make the new treatment acceptable for use.

What if my sample size is small?

With small sample sizes, the confidence interval will be wider, making it more difficult to demonstrate non-inferiority. You may need to increase the sample size or accept a wider non-inferiority margin.