Sample size formula for proportional hazards modelling of competing risks

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Research Methodology

Background:

• Competing risks are common in medical studies, where multiple failure types can occur.
• Traditional cause-specific hazard models may not fully capture the impact of prognostic factors on specific failure types.
• Subdistribution hazard models offer an alternative approach for analyzing competing risks.

Purpose of the Study:

• To develop and validate approximate sample size formulas for proportional hazards modeling of subdistribution hazards.
• To account for both independent and correlated covariates in competing risk analyses.
• To provide practical tools for researchers designing studies with competing risks.

Main Methods:

• Derivation of approximate sample size formulas for subdistribution hazard models.
• Conducting numerical simulations to assess the accuracy and performance of the proposed formulas.
• Applying the formulas to real-world data from a randomized clinical trial and a prospective prognostic study.

Main Results:

• The proposed approximate sample size formulas provide valid estimates for planning competing risk studies.
• Numerical simulations confirm the reliability of the formulas under various scenarios.
• The formulas are applicable to studies with both independent and correlated covariates.

Conclusions:

• The developed sample size formulas are valuable tools for researchers in biostatistics and clinical epidemiology.
• Accurate sample size calculation is crucial for the efficient design and interpretation of studies involving competing risks.
• This work facilitates more robust analyses of therapeutic and prognostic factors in the presence of multiple failure types.