Time until first significant difference in in vivo tumor growth experiments

Area of Science:

• Oncology
• Biostatistics

Background:

• Standard analysis of in vivo tumor growth experiments often identifies the time when treatment group volume distributions become significantly different.
• This common practice is statistically deficient, as its Type I error rate exceeds the nominal 5% unless multiple comparison corrections are applied.
• Investigators may incorrectly interpret the time of first significance as a fixed statistical parameter rather than a result influenced by experimental design.

Purpose of the Study:

• To highlight the statistical deficiencies in common methods for analyzing in vivo tumor growth data.
• To propose more robust statistical approaches for comparing treatment groups in tumor growth studies.

Main Methods:

• Critique of the conventional method of determining the time of first significant difference in tumor volume distributions.
• Discussion of the influence of experimental design (sample size, measurement spacing) and model parameters on the time of first significance.
• Advocacy for alternative statistical modeling techniques.

Main Results:

• The conventional method inflates the Type I error rate, leading to false positives.
• The time of first significance is not a fixed parameter but depends on true model parameters and experimental design.
• Statistical power and the time of first significance are influenced by factors like sample size and measurement frequency.

Conclusions:

• Investigators should avoid relying solely on the time of first statistical significance for in vivo tumor growth analysis.
• Modeling tumor growth curves or estimating volume doubling times are recommended as more reliable statistical approaches.
• Adopting these advanced methods can lead to more accurate and reproducible research findings in oncology.