Ultrasonographic determination of day of parturition based on fetal biparietal diameter in pregnant bitches: Comparison of simple linear regression and mixed linear regression in breed-specific models and maternal weight grouped models.

The aim of this retrospective study was to compare simple linear regression and mixed linear regression on data grouped by breed or maternal weight group. The comparison was done to find the most accurate model for predicting day of parturition in pregnant bitches in clinical practice. The retrospective data consisted of fetal biparietal diameter determined by ultrasonography and day of parturition for all included bitches. The study population was divided into five maternal weight groups (miniature (≤5 kg), small (>5 to 10 kg), medium (>10 to 25 kg), large (>25 to 40 kg), and giant (>40 kg)) with three breeds in each group with 26 miniature-breed bitches, 13 small-breed bitches, 19 medium-breed bitches, 22 large-breed bitches, and 20 giant-breed bitches. The data was used to develop models to determine the number of days before parturition based on fetal biparietal diameter. A statistically significant effect was seen for grouping by maternal weight group (p<0.0001) and by breed (p=0.0057). Breed-specific models were derived and compared to each other within the same maternal weight group. Statistically significant differences between some miniature-breed and small-breed bitches were found using mixed linear regression analysis. The accuracies of all models were given as number of births within ±1 and ±2 days of estimated day of parturition and compared to an acceptable limit of 80% at ±2 days. All breeds and maternal weight groups except Dogue de Bordeaux and giant-breed bitches met the limit. Poor accuracies were seen when applying data from each breed to the maternal weight grouped model. Simple linear regression analyses were compared to mixed linear regression analyses. The simple linear regression analyses obtained the best accuracies for most of the breeds which is most likely to be caused by overestimation. Comparison of Chihuahua and German Shepherd to other studies showed similar accuracies between the highest reported and the two linear models. We recommend the use of breed-specific models based on mixed linear regression analysis in clinical practice. Further research is needed to analyze the differences between the two linear models and to confirm the tendency of more accurate predictions of day of parturition for medium-breed, large-breed, and giant-breed bitches when using breed-specific models.