How should data on murine spontaneous abortions be expressed and analysed?



Problem:  Spontaneous abortions (resorptions) in the CBAxDBA/2 are normally reported as number or resorptions/total number of implantations pooling >5 mice (R/T). The significance of differences between groups is 2 or Fisher's Exact test determined using non-parametric methods (e.g. based on a priori predictions). Recently, it has been suggested that medians with box-plots replace the accepted standard. This has led to reported rates of abortion of zero, where a median (that divides a population into the 50% of mice > the median and the 50% < the median) cannot logically exist. This method gives equal value to a mouse with 0/2 = 0% abortions and 0/10 = 0% abortions, and deprives readers of key data required to verify P values. The fact that there is an irreducible minimum loss due to chromosomally abnormal embryos (2–5%) is also ignored.

Methods:  Raw data on 177 individual CBAxDBA/2 matings was analysed by median, mean, and geometric mean, along with R/T from 6 independent experiments with 5–10 mice per group.

Result:  Individual abortion rates showed a non-Gaussian distribution, but the mean and median differed by <0.5%. R/T data from the 6 independent experiments were normally distributed. Only mean and geometric mean enable between-group P value calculation.

Conclusion:  Although it is possible to compare individual mice (and even individual implantation sites), as the relevant question is the significance of differenced between treatments in groups of mice and reproducibility, the established classical method should continue to be used.