Associations between dry period length and time to culling and pregnancy in the subsequent lactation

The association between dry period length (DPL) and time to culling and pregnancy in the subsequent lactation may be important for the economically optimal length of the dry period. Therefore, this study aimed to (1) quantify the association between DPL and hazard of culling and pregnancy in the subsequent lactation; (2) develop continuous functions of DPL for the hazard ratios of culling and pregnancy; and (3) investigate the effect of a cause-specific hazards model and a subdistribution model to analyze competing events. The data used in this observational cohort study were from dairy herd improvement milk test lactation records from 40 states in the United States. After edits, there remained 1,108,515 records from 6,730 herds with the last days dry in 2014 or 2015. The records from 2 adjacent lactations (current, subsequent) were concatenated with the DPL of interest, 21 to 100 d, in between both lactations. We defined 8 DPL categories of 10 d each. Kaplan-Meier survival curves were used to show associations between DPL and time to culling or pregnancy for 3 lactation groups: lactation 1 and 2, lactation 2 and 3, and lactation 3 and greater. To control for confounding factors in Cox proportional models, we included 6 current lactation covariates and 3 time-dependent variables in the survival models. Hazard ratios of culling were estimated for 4 days in milk (DIM) categories from 1 to 450 DIM. Hazard ratios of pregnancy were estimated for 3 DIM categories from 61 to 300 DIM. Competing risk analysis of 8 disposal codes (i.e., farmer reported reasons) for culling and the culling event for pregnancy were conducted by a cause-specific hazards model and a subdistribution model. Hazard ratios were also estimated as quadratic polynomials of DPL. Compared with the reference DPL category of 51 to 60 d, hazard ratios of culling and pregnancy of the other 7 DPL categories ranged between 0.70 and 1.49, and 0.93 and 1.15, respectively. Short DPL were associated with lower risk of culling in the early lactation but not over the entire lactation. Short DPL were associated with greater hazard of pregnancy. Trends in hazard ratios over the ranges of the 8 DPL categories were not always consistent. Competing risk analysis with both models provided little differences in hazard ratios of culling and pregnancy. In conclusion, variations in DPL were associated with meaningful differences in the hazard ratios for culling and pregnancy and minor differences in the relative frequency of disposal codes. Subdistribution hazards models produced hazard ratios similar to cause-specific hazard models. The quadratic polynomials may be useful for decision support on customization of DPL for individual cows.