Outcome-Dependent Sampling Design and Inference for Cox's Proportional Hazards Model

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Sampling

Background:

• Failure time data analysis often faces challenges with biased sampling schemes.
• Existing sampling designs may not be cost-effective or statistically optimal.
• Efficient inference procedures are crucial for reliable results from biased data.

Purpose of the Study:

• To propose a cost-effective outcome-dependent sampling (ODS) design for failure time data.
• To develop an efficient inference procedure tailored for data collected using the ODS design.
• To evaluate the performance and power of the proposed ODS design against existing methods.

Main Methods:

• Derivation of estimators using a weighted partial likelihood estimating equation to address biased sampling.
• Development of criteria for optimal implementation of the ODS design in practical scenarios.
• Evaluation of small sample performance through simulation studies.

Main Results:

• Proposed estimators for regression parameters are consistent and asymptotically normally distributed.
• The ODS design demonstrates statistically superior power compared to alternative designs at the same sample size.
• The method was successfully illustrated using real-world data from the Uranium Miners Study.

Conclusions:

• The proposed outcome-dependent sampling design offers a statistically powerful and cost-effective solution for failure time data.
• The developed inference procedure provides efficient and reliable analysis for data collected under this biased sampling scheme.
• This approach enhances the analysis of complex survival data, as shown in the Uranium Miners Study example.