The incidences of fetal losses in intensive dairy herds range from 7.6 to 20.5% ( Starbuck et al. , 2004 ; Pontes et al. , 2015 ), a 2.7-fold difference between herds. Causes for these apparent differences are not clear, although the wide variation in the incidence of this reproductive disorder could be explained by the inconsistent definition of gestation loss among studies, the different gestational at-risk periods considered, and the wide computation methods used among studies.

Non-infectious causes of abortion play an important role in premature pregnancy conclusion. However, the infectious causes of abortion have been a primary focus of attention and the control these infectious dise­ases is based on several points like vaccination, but non-infectious fetal losses still is likely in dairy herds ( López‐Gatius & García‐Ispierto, 2010 ). Studies that fo­cus on fetal losses have examined factors related to genetic abnormalities ( Flisikowski et al. , 2010 ) and secondary metabolites in diets offered to cows ( Santos et al. , 2003 ). Other circumstances that can terminate pregnancy include heat stress ( García-Ispierto et al. , 2006 ; Mellado et al. , 2016 ), high milk production ( Grimard et al. , 2006 ), cow parity ( Yániz et al. , 2010 ) and sub-optimal blood progesterone concentrations ( López-Gatius et al. , 2004 ). Additional risk factors for early fetal losses include intensive management ( López-Gatius et al. , 2009 ), parity ( Labèrnia et al. , 1996 ; López-Gatius et al. , 2004 ), twin pregnancies ( López-Gatius et al. , 2002 ; Silva-Del-Río et al. , 2009 ) and occurrence of mastitis ( Chebel et al. , 2004 ; Pinedo et al. , 2009 ).

When managing dairy cattle, it is important to understand the factors that might be related to fetal losses because knowledge of these influences can be used to focus herd health and management practices to reduce fetal losses. Many studies on gestation losses have been carried out during the early phase of fetal development. Therefore, further elucidation of factors that cause fetal losses throughout gestation in regions with prolonged warm weather may enable the development of more effective reproductive programs in intensive dairy herds. The objectives of the present study were to evaluate various factors associated with fetal losses in high-yielding lactating Holstein cows experiencing heat stress during an extended period of the year.

Risk factors for the occurrence of fetal losses are indicated in Table 1. Fetal losses were 23.8% (3,426/14,384). Cows that had dry periods longer than 60 days (range 61-401 days) were less likely to lose their fetuses in the subsequent pregnancy than cows having a dry period of 60 days or less (range 0-60 days). Cows with low BCS (range 2.25-3.0) at calving presented higher fetal losses than cows with adequate BCS (range 3.25-4.0). Likewise, BCS >2.75 15 d post­calving was associated with decreased risk for fetal losses. Cows with peak milk yield >38 kg (range 39-79 kg) presented higher odds of fetal losses than cows with low peak milk yield (range 24-38 kg). In addition, cows whose 305-d milk yield was higher than 9,000 kg (range 9,000-14,623) were more likely to have a fetal loss than cows with lower (range 6,857-9,000 kg) milk yield per lactation. Cows with retained placenta were 1.2 times more likely to have fetal loss than cows without this reproductive disorder. Cows suf­fering premature births had 1.2 greater ( p <0.05) risk of suffering fetal losses than cows with normal calving and cows carrying twins had more than twice the odds of losing their fetuses than cows carrying single fetuses.

Fetal loss was not associated with the occurrence of a previous abortion and lactation persistency ( p >0.05). Also, the occurrence of mastitis and uterine diseases did not represent a risk for fetal losses. The interactions between 305-milk yield × peak yield, 305-milk yield × body condition score at parturition, days dry × peak yield, days dry × body condition score at parturition and BCS at parturition and retained placenta were significant ( p <0.05) for the occurrence of fetal losses. The histogram describing the percentage of fetal losses throughout gestation exhibited a signifi­cant (two-sided p <|Z| 0.001) bimodal distribution with two well-defined peaks at around 45 and 220 days of gestation (median=131, inferior quartile=73, superior quartile=188; Figure 1).

In this study, cows with long dry periods were less prone to fetal loss rate compared with cows with dry period ≤60 days. This response is not clear because dry periods greater than 60 days have led to higher serum non-esterified fatty acids, β-hydroxybutyrate, and lower serum insulin and glucose concentrations postpartum, which indicates a poorer energy balance status compared to cows with shorter dry periods ( Weber et al. , 2015 ). Oppositely, cows with short dry periods (35-40 days) have greater dry matter intakes, better reproductive performance ( Shoshani et al. , 2014 ) and lost less weight post-partum than cows with >60-day dry periods, suggesting a better metabolic state ( Jolicoeur et al. , 2014 ). These results, however, are unsure because the im­proved reproductive function observed with short dry periods is due to the lower milk yield associated with short dry periods ( van Knegsel et al. , 2014 ). In fact, when some reproductive variables are adjusted for milk yield, short dry periods actually results in inferior fertility in the subsequent lactation ( Kuhn et al. , 2006 ).

Prolonged dry period frequently leads to greater BCS ( Weber et al. , 2015 ), but this was not the case in the present study where the difference in BCS of cows with short or long dry period was only 0.04 units. Hence, fetal losses were not associated with body energy reserves in cows with different dry period length.

Low BCS at calving and 15 d postpartum predis­posed cows to have higher fetal losses. This is in line with observations of Starbuck et al. (2004) who reported that cows in average body condition (2.75-3.25) maintained 92.1% of pregnancies whereas those with lower condition (≤2.50) maintained 84.2% of pregnancies during weeks 5-9 of gestation.

These data demonstrate that energy status is an important factor regulating the survival of fetuses. Fat mobilization to meet energy requirements during early lactation is unavoidable because of insufficient feed intake in high-yielding dairy cows ( Wheelock et al. , 2010 ). This loss of weight during the first months of lactation, particularly in cows with low body energy reserves can exacerbate the negative correlation with progesterone secretion, which is harmful to fetal sur­vival ( Grimard et al. , 2006 ). Cows with lower BCS would be expected to eat more, which increases the metabolism of progesterone and results in lower plasma concentrations of this hormone ( Rabiee et al. , 2002 ). Additionally, low BCS is associated with low-quality embryos ( Carvalho et al. , 2014 ) and may be long-term carry-over effects of low body energy reserves on the impaired development of fetuses.

Cows with peak milk yield >38 kg and 305-d milk yield >9,000 kg were more likely to have a fetal loss. It has been shown that very high milk production at the beginning of lactation induce a physiological state of negative energy balance because milk yield, milk proteins, fat and lactose increase rapidly and exceed feed intake ( Bertoni et al. , 2009 ). During peak lactation, over 80% of the available glucose in the body is partitioned to the mammary gland for milk synthesis ( Hocquette & Bauchart, 1999 ). Additionally, the majority of fats mobilized from adipose tissue contribute to milk fat synthesis ( Grummer, 2008 ). This situation leads to a decline in cow fertility ( Walsh et al. , 2011 ) with fetal losses as one of the major causes of reproductive failure.

The higher fetal losses in cows experiencing retained placenta could be the result of failure of complete uterine recovery and subsequent subclinical endometritis. This condition has been associated with an increase of late embryo/early fetal loss ( Lima et al. , 2013 ). Additionally, uterine diseases also modulate the endocrine function of the uterus and ovary, which would hamper fetal development ( Dobson et al. , 2007 ).

The first peak of fetal losses observed in the present study (around 45 days) is in line with various previous observations where early fetal losses range between 12 and 20% ( Starbuck et al. , 2004 ; Pontes et al. , 2015 ). The second peak of fetal losses occurred in the third trimester of gestation, which is not in line with other reports where it was found that mid-to-late fetal losses were lower than those occurring in early pregnancy in lactating cows ( Jousan et al. , 2005 ). Most late-term abortions are caused by infectious diseases, but in the current study, a good vaccination effort directed to abortigenic diseases for dairy cows was implemented, although this does not rule out the possibility of infectious diseases for which cows were not immunized. Therefore, causes of abortions around 220 days of pregnancy remain unclear. This high increment of fetal losses in late-gestation could be due to intense ambient thermal stress ( Mellado et al. , 2016 ) via its actions to compromise placental weight and mass ( Dunlap et al. , 2015 ), which harm normal placental vascular development and placental inefficiency ( Marai et al. , 2007 ), and finally intrauterine growth restriction ( Regnault et al. , 2002 ), or to a diverse bacterial species associated with opportunistic infections of the placenta and fetus resulting in abortion ( Anderson, 2007 ). Another possible cause of fetal losses at the end of gestation could be the presence of Neospora caninum in the herd ( Klauck et al. , 2016 ). This protozoa causes abortion due to tissue destruction in the fetus, the maternal immune response, placental inefficiency due to inflammation, or a combination of these responses ( Dubey et al. , 2006 ). Despite a preparturition and then yearly vaccination against Brucella abortus , this disease is present in this herd ( Mellado et al. , 2014 ), therefore, this disease could be another possible cause of late fetal losses in the present study.

In the present study, the occurrence of premature parturition was associated with greater odds of fetal losses. A plausible explanation for this response is the presence of cows infected with Neospora caninum , a widespread disease in northern Mexico ( Salinas et al. , 2005 ) that causes calves to be born prematurely ( McAllister et al. , 2000 ). In the present study, fetal losses were aggravated with multiple fetuses as compared to a single fetus, despite the fact that bovine females had the uterine capacity to support two fetuses. These findings agree with other studies where the survival of single pregnancies is better than that of twin pregnancies ( Silva-Del-Río et al. , 2009 ; Mellado et al. , 2016 ).

As twinning is a consequence of higher potential for milk yield ( Macmillan et al. , 2018 ), it seems that the higher fetal losses in cows bearing twin pregnancies are mediated by metabolic stress associated with higher milk yield ( Weber et al. , 2013 ). In high-yielding dairy cows, the increased body tissue mobilization and an increased metabolic rate emit metabolic sensory signals that negatively affect reproduction ( Chagas et al. , 2007 ).

These findings indicate that higher peak milk yield and 305-d milk production correlates with an increased risk of fetal losses. These results also show that, under the present conditions, it is necessary to maintain BCS at 3.0 units or greater at calving in order to reduce fetal losses and that the occurrence of retained placenta, twin pregnancies and short days dry are important indicators to identify cows at risk for fetal losses. Finally, these observations show that fetal losses under the warm weather of the present study are equally likely at the beginning or late stages of gestation in well-managed high-yielding Holstein cows.