Sample size and power determination for a binary outcome and an ordinal exposure when logistic regression analysis is...

Area of Science:

• Biostatistics
• Medical Statistics
• Epidemiology

Background:

• Sample size determination is crucial for the reliability of statistical analyses.
• General methods for logistic regression sample size calculation can be data-intensive.
• Specialized, computationally simple methods are needed for specific research scenarios.

Purpose of the Study:

• To present practical methods for sample size determination in logistic regression with a binary outcome and a three-level quantitative exposure.
• To provide a framework for assessing the prognostic value of biomarkers like multidrug resistance gene (mdr1) expression.
• To simplify sample size calculations for studies involving ordinal or quantitative exposures.

Main Methods:

• Development of closed-form expressions for sample size calculations.
• Application to a three-level ordinal exposure with specified scaling.
• Assessment of sensitivity to exposure scaling and distribution.
• Evaluation of the linear trend assumption in log odds versus proportions.

Main Results:

• Computationally simple methods for sample size determination were derived.
• The methods are applicable to studies investigating trends in binary outcomes related to multi-level exposures.
• Sensitivity analyses were performed for key assumptions.

Conclusions:

• The presented methods offer a practical approach to sample size planning for specific logistic regression models.
• These techniques are valuable for research involving biomarker expression and treatment response, such as mdr1 in sarcoma.
• The approach simplifies the application of logistic regression in studies with quantitative or ordinal exposures.