Approximating the baseline hazard function by taylor series for interval-censored time-to-event data

Area of Science:

• Biostatistics
• Clinical Trials
• Oncology

Background:

• Oncology clinical trials often generate interval-censored time-to-event data.
• Current analysis methods may impute event times, introducing bias and potentially erroneous conclusions.
• Existing statistical methods like Cox regression are sensitive to imputation accuracy.

Purpose of the Study:

• To propose a novel statistical method to mitigate bias in analyzing interval-censored time-to-event data from oncology trials.
• To improve the accuracy of statistical inference in time-to-event analyses.
• To provide a robust alternative to common imputation practices.

Main Methods:

• A Taylor series approximation is proposed for the log baseline hazard function within Cox proportional hazards regression.
• The likelihood ratio test is utilized to determine the optimal order for the Taylor series approximation.
• Maximum likelihood techniques are employed for parameter estimation and statistical inference.

Main Results:

• The proposed Taylor series method effectively mitigates bias associated with imputed time-to-event data.
• Simulation studies demonstrate the method's validity and improved accuracy.
• Application to breast cancer trial data showcases its practical utility.

Conclusions:

• The novel Taylor series approximation offers a statistically sound approach for analyzing interval-censored time-to-event data in oncology.
• This method enhances the reliability of statistical inference, particularly for treatment effect evaluation.
• It provides a valuable alternative to potentially biased imputation techniques in clinical trial analysis.