Chop-lump tests for vaccine trials.

Area of Science:

• Biostatistics
• Clinical Trials
• Epidemiology

Background:

• Randomized vaccine trials often face challenges in detecting efficacy when infection rates are low.
• Existing methods like the burden of illness (BOI) test can lack power with rare events.
• Subgroup analyses, while potentially powerful, rely on untestable assumptions.

Purpose of the Study:

• To introduce novel statistical tests, "chop-lump" Wilcoxon (CLW) and t-tests (CLT), for comparing vaccine and placebo groups in randomized trials.
• To enhance statistical power in vaccine trials, particularly when infections are infrequent.
• To provide a more robust alternative to existing methods for analyzing vaccine efficacy.

Main Methods:

• Proposed "chop-lump" (CL) methods that remove an equal number of zero outcomes from both vaccine and placebo groups before analysis.
• Utilized a permutation approach to establish the null distribution for the CLW and CLT tests.
• Compared the power of CLW and CLT against traditional Wilcoxon and t-tests, as well as BOI tests, through simulations.

Main Results:

• The CLW test demonstrated greater power than the standard Wilcoxon test under local alternatives when infection rates are equal.
• The "gap" between zero and non-zero outcomes significantly influences the power of the CLT.
• Simulations indicated CL tests can be more powerful than BOI tests in specific scenarios.

Conclusions:

• Chop-lump tests offer a valuable enhancement for analyzing vaccine trial data, especially with low infection rates.
• These new methods provide increased statistical power and can lead to more sensitive detection of vaccine effects.
• The study illustrates the application of CL tests using data from HIV and malaria vaccine trials.