|Abstract:||Pregnancy loss is a multifactorial condition that compromises reproductive performance in dairy operations. Despite the high oocyte fertilization rate in dairy cows, only 28 % of those maintain a pregnancy to term. Pregnancy loss is estimated to cost U$600.00 per case. Identification of cows losing the pregnancy as early as possible can be helpful in providing timely opportunities for rebreeding, thus potentially minimizing economic losses. Traditionally, early pregnancy diagnosis is performed via ultrasonography, starting at 30 days, which provides information regarding embryo viability, uterine health, and ovarian structures. In addition, this technique allows the diagnosis of twin pregnancy that is three times more likely to be lost than a singleton. Despite its benefits ultrasonography requires well-trained personnel and incurs additional costs involving equipment purchase and maintenance. The use of biomarkers has been studied throughout the years, based on a demand for an easier, less costly, and more accurate test. Pregnancy-associated glycoproteins (PAG) is the most common biomarker marker to assess pregnancy status in cows. Produced and secreted on the maternal circulation by binucleate giant cells. Measurement of PAG in the blood has high sensitivity when performed between 25 32 days of gestation, however, the specificity can be as low as 83%. One of the major components that affect test accuracy is pregnancy loss. It has been reported that cows experiencing early pregnancy loss, present lower plasma concentrations of PAG. Another indirect biomarker to detect pregnancy in cows is progesterone. Cows experiencing pregnancy loss showed lower concentrations of this hormone, in comparison to cows keeping the pregnancy. The development of a threshold for PAG and progesterone that can predict pregnancy loss may aid in management decisions to provide earlier rebreeding opportunities. It was hypothesized that the plasma concentration of PAG and progesterone is reduced and can predict pregnancy loss in cows experiencing a high-risk pregnancy. Additionally, it was hypothesized that the concentration of PAG and progesterone are increased and can predict twins. High-risk pregnancy (HR) were characterized using transrectal ultrasonography 37 days post-AI based on the following criteria: small embryo size (SE, embryo < 15 mm, n=10), slow heartbeat (SH, <60 beats per minute, n = 11), extra amniotic membrane (EM, additional amniotic membrane, n=3). A cohort of twins (TW, n = 41) diagnosed at day 37 post-AI was also enrolled. Twins were also subgroups in unilateral (UT, n=17) and bilateral (BT, n=24). Each HR and TW cow was paired with the same parity cow carrying a normal singleton at d 37 post-AI (CON, n = 65). Blood samples were collected to measure PAG and progesterone at 37, 44, and 51 post-AI. Statistical analysis was performed using ANOVA, logistic regression and receiver operation characteristics (ROC) with JMP. Pregnancy loss at day 51 post-AI was greater (P < 0.01) in HR than CON and TW (CON=1.5%; HR=87.5%; TW=12.2%). Concentration of PAG at day 37 post-AI did not differ (P = 0.75) among groups (CON = 5.3 ± 0.7; HR = 4.8 ± 1.2; TW = 4.0 ± 0.9 ng/ml). The subgroup SE showed a statistical difference regarding the concentration of PAG at day 51 post-AI (P < 0.05), EM showed a tendency (P < 0.10) whereas SH, UT and BT did not when compared to CON. Concentration of progesterone at day 37 post-AI was greater in TW than HR and CON, and lower (P < 0.01) in HR than CON cows (CON = 7.0 ± 0.3; HR = 5.9 ± 0.4; TW = 8.4 ± 0.3 ng/ml). Regression and ROC analysis for PAG at day 37 post-AI did not find a threshold to predict pregnancy loss (P = 0.24) or twins (P = 0.30). Regression and ROC analysis for progesterone at day 37 post-AI found that a threshold of 6.5 ng/ml predicted (P < 0.01) pregnancy loss with an area under the curve (AUC) of 0.65, and threshold of 7.2 ng/ml predicted (P < 0.01) twins with AUC of 0.70. In summary, pregnancy loss and twins were predicted with only moderate accuracy by progesterone concentration at day 37 post-AI and the variability in PAG concentrations at day 37 post-AI was insufficient to generate a threshold to predict pregnancy loss and twins in Holstein lactating cows.