Prediction of pregnancy following transfer of in‐vitro produced embryos using recipient health records, accelerometer sensors and meteorological data in Australian dairy cows

Embryo transfer (ET) in dairy cattle has long been used for the dissemination of elite genetics. However, achieving consistently high pregnancy rates remains a major challenge, particularly in high‐producing lactating dairy cows receiving in‐vitro produced (IVP) embryos. We hypothesised that recipient selection and ET program timing could be optimised using individual items of recipient and meteorological data that are routinely available on dairy farms. To investigate this hypothesis, we conducted an observational study that explored associations between recipient characteristics, health events, sensor‐derived activity alerts, meteorological variables and their ability to predict successful pregnancy following ET (P/ET) in an on‐farm IVP embryo program. Cow health data and ear tag health sensor alerts from a commercial dairy herd in southern Australia were used to compile a retrospective dataset of 5770 embryo transfers (ETs) in 4681 recipients. Meteorological records from two nearby weather stations were used to calculate temperature humidity index (THI). Univariable generalised linear models were used to test associations between predictors and P/ET. Outputs from models included risk differences (RD), P‐values, 95% confidence intervals and area under the curve (AUC). Population RD were calculated using tabular methods. Recipient parity, days in milk (DIM), calving interval, metritis events and ‘very sick’ ear‐tag sensor alerts were associated with P/ET, as were temperature and THI variables. Area under the curve values were close to 0.50 (range 0.50–0.55) and population RD values from many exposure variables were close to 0, indicating that optimising the ET program using individual variables as done in this study, under the conditions in this herd may have limited impact on the overall P/ET. Therefore, further research is needed to evaluate other methodologies to predict P/ET. Furthermore, future studies should incorporate non‐recipient factors, such as embryo quality. It is unclear if the associations observed in this study are due to causal relationships; therefore, we recommend that observational and intervention studies be conducted to verify this.