Accelerated Failure Time Survival Model to Analyze Morris Water Maze Latency Data

Clark R. Andersen, Jordan Wolf, Kristofer Jennings, Donald S. Prough, Bridget E. Hawkins

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

Traumatic brain injury (TBI) induces cognitive deficits clinically and in animal models. Learning and memory testing is critical when evaluating potential therapeutic strategies and treatments to manage the effects of TBI. We evaluated three data analysis methods for the Morris water maze (MWM), a learning and memory assessment widely used in the neurotrauma field, to determine which statistical tool is optimal for MWM data. Hidden platform spatial MWM data aggregated from three separate experiments from the same laboratory were analyzed using 1) a logistic regression model, 2) an analysis of variance (ANOVA) model, and 3) an accelerated failure time (AFT) time-to-event model. The logistic regression model showed no significant evidence of differences between treatments among any swims over all days of the study, p > 0.11. Although the ANOVA model found significant evidence of differences between sham and TBI groups on three out of four swims on the third day, results are potentially biased due to the failure of this model to account for censoring. The time-to-event AFT model showed significant differences between sham and TBI over all swims on the third day, p < 0.045, taking censoring into account. We suggest AFT models should be the preferred analytical methodology for latency to platform associated with MWM studies.

Original languageEnglish (US)
Pages (from-to)435-445
Number of pages11
JournalJournal of Neurotrauma
Volume38
Issue number4
DOIs
StatePublished - Feb 15 2021

Keywords

  • Morris water maze
  • latency
  • learning and memory
  • survival analysis
  • traumatic brain injury

ASJC Scopus subject areas

  • Clinical Neurology

Fingerprint Dive into the research topics of 'Accelerated Failure Time Survival Model to Analyze Morris Water Maze Latency Data'. Together they form a unique fingerprint.

Cite this