Mantel-Haenszel analyses of litter-matched time-to-response data, with modifications for recovery of interlitter information.

Abstract

The Mantel-Haenszel procedure for comparing sets of time-to-response data is adaptable to data that can be stratified on other variables. A particular adaptation, which we have used, is one in which animals from the same litter have been assigned to different treatment groups, e.g., some to a control group and some to a drug treatment group. The time to response used was that of tumor appearance; death from other causes was considered a loss to observation. The initial litter-adjusted analysis seemed to have only limited advantages compared to analysis that ignored litters and could be interpreted as suggesting that litter matchihg was not advantageous. Contributing to the difficulty was the fact that in the litter-matched analysis no further information was forthcoming from the remaining similarly treated animals in a litter when there were no remaining contrastingly treated littermates. Several devices for recovering interlitter information from such remnants and for combining it with intralitter information are examined and applied.