Effect of non-random missing data mechanisms in clinical trials

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Data Analysis

Background:

• Missing data is a common challenge in clinical trials.
• Non-ignorable missing data mechanisms can introduce bias.
• Simple analyses often ignore missing data, potentially leading to flawed conclusions.

Purpose of the Study:

• To characterize non-ignorable missing data mechanisms using a two-parameter model.
• To investigate the impact of non-randomly missing data on analyses that ignore missing values.
• To evaluate the effects on bias and statistical power for different response variable types.

Main Methods:

• A two-parameter model was employed to define missing data mechanisms.
• The study analyzed the consequences of ignoring non-randomly missing data.
• Simulations or theoretical derivations were used to assess bias and power for binary and normal outcomes.

Main Results:

• The bias and power effects of non-randomly missing data escalate with increased non-randomness.
• Bias can manifest as positive or negative, and power may deviate from 'missing at random' scenarios.
• Ignoring missing data can severely compromise analysis integrity, especially with high non-randomness.

Conclusions:

• Analyses ignoring non-randomly missing data can yield seriously flawed results, even with minimal missing proportions.
• The impact is more severe for normally distributed response variables with unequal variances.
• Careful consideration of missing data mechanisms is crucial for valid clinical trial analysis.