Consideration of culling bias in assessing the relationship between fetal survival and maternal age in dairy cows

This paper presents a method for assessing the potential bias associated with culling that could confound a relationship between maternal age at conception and fetal survival in dairy cows. A nominal variable (PREGSTAT) was created to distinguish between two groups of culled cows: those culled while open and those culled while pregnant. We tested for differences in fetal survival during the last completely followed pregnancies of cows in the two groups. Survival analysis using the Cox model and data from nine dairies in the San Joaquin Valley of California showed a higher risk of fetal loss for cows subsequently culled while open (relative risk (RR) = 2.7, P < 0.0001). Similar estimates of RR of maternal age on fetal survival were obtained from two models-one with one without the PREGSTAT variable-indicating that the relationship between maternal age and fetal survival was not confounded by culling. There was an interaction between maternal age and the PREGSTAT variable (P < 0.0001). The approach described here may be applicable in studies of other disease conditions that are subject to culling.