Methods for analysis of censored tumor growth delay data

Area of Science:

• Oncology
• Biostatistics
• Experimental Therapeutics

Background:

• Tumor growth delay assays are crucial for evaluating experimental therapies.
• Censored data, arising from animal death or limited follow-up, complicates analysis.
• Accurate statistical methods are needed to handle incomplete tumor growth data.

Purpose of the Study:

• To evaluate the accuracy of the Kaplan-Meier method for estimating median tumor growth delay with censored data.
• To compare the statistical power of different tests for analyzing censored tumor growth data.
• To provide guidance on analyzing incomplete data in preclinical cancer research.

Main Methods:

• Computer simulations were used to assess estimation bias and statistical test power.
• Log-normal distribution and independence of censoring were assumed.
• Kaplan-Meier, log-rank, Gehan-Wilcoxon, mu, and t-tests were evaluated.

Main Results:

• The Kaplan-Meier method provides unbiased estimates of median growth delay, even with significant censoring.
• Omitting censored data leads to biased estimates.
• Tests designed for censored data (log-rank, Gehan-Wilcoxon) show markedly higher power than those for complete data (mu, t-tests) at censoring rates of 40%.

Conclusions:

• The Kaplan-Meier method is a robust tool for analyzing tumor growth delay data with censoring.
• Statistical tests accounting for censoring are essential for reliable comparisons of therapeutic efficacy.
• These methods enable reliable quantitative and qualitative analysis of growth delay data with up to 30% censoring.