Sample size calculations for risk equivalence testing in pharmacoepidemiology.

Area of Science:

• Pharmacoepidemiology
• Biostatistics
• Drug Safety

Background:

• Equivalence testing is established in pharmacokinetics (bioequivalence) and clinical trials (therapeutic equivalence).
• Its application in pharmacoepidemiology involves comparing a drug's risk against a known risk or another drug.
• Defining equivalence regions based on risk difference or ratio is crucial but complicated by binary, low-frequency outcomes.

Purpose of the Study:

• To propose usable sample size formulae for risk equivalence testing in pharmacoepidemiology.
• To provide a statistically sound method for concluding drug equivalence in observational studies.
• To contrast this approach with the common practice of non-rejection of the null hypothesis of equality.

Main Methods:

• Development and application of sample size formulae for one-group (known risk) and two-group (drug-drug) designs.
• Consideration of equivalence regions defined by risk difference or risk ratio.
• Calculation of sample sizes for specific scenarios, e.g., risk of 0.01, 20% maximal risk difference, one-sided test (alpha=0.05, beta=0.2).

Main Results:

• Sample size formulae are presented for risk equivalence testing.
• For a risk of 0.01 and 20% maximal risk difference, sample sizes are 15,309 (one-group) and 30,617 (two-group) per group.
• These sample sizes are deemed acceptable for practical use when the risk exceeds 1/1,000.

Conclusions:

• Risk equivalence testing with predefined limits and appropriate sample size calculation is the correct method for concluding equivalence in pharmacoepidemiology.
• This approach is superior to relying on the non-rejection of the null hypothesis of equality.
• The proposed methods offer practical solutions for sample size determination in this field.