On the use of logrank scores in the analysis of litter‐matched data on time to tumour appearance

Previously, Mantel and Ciminera¹ introduced an application of the Mantel‐Haenszel procedure to litter‐matched time‐to‐tumour data on treated and control animals with a device for recovering ‘interlitter information’ from non‐informative litters. An example is given here to show that the Mantel‐Ciminera procedure can confound litter effect and treatment effect; their procedure is shown to produce a large chisquare value on contrived data containing no information on the treatment effect on time to tumour. An alternative procedure is described which completely adjusts for litter effect. This procedure is adaptable to any scoring system. As an illustration, it is carried out with logrank scores to reanalyse the data of Mantel and Ciminera.