Heat stress during the transition period is associated with impaired production, reproduction, and survival in dairy cows

Objectives of this retrospective cohort study were to determine the association of heat stress (HS) exposure during the periparturient period with production, health, reproduction, and survival during the first 90 d postpartum in dairy cows. A total of 5,722 Holstein cows (2,324 nulliparous and 3,397 parous) were categorized into environmental condition groups based on average temperature-humidity index (THI) exposure as thermoneutral (TN) or HS during the prepartum (PRE) and early postpartum (POST) periods into TN-TN (THI: PRE = 57.9 and POST = 63.7), TN-HS (THI: PRE = 68.2 and POST = 72.0), HS-TN (THI: PRE = 71.4 and POST = 65.6), and HS-HS (THI PRE = 72.2 and POST = 72.7). Nulliparous and parous cows were analyzed separately. In nulliparous cows, exposure to HS during the PRE, POST, or both PRE and POST periods was associated with a 1.7 kg/cow per day reduction in milk yield compared with TN-TN. Postpartum HS was associated with increases of 4.4 percentage points in incidence of retained placenta, 18.1 percentage points in incidence of metritis, and 2.0 percentage points in incidence of mastitis, but a reduction of 5.3 percentage points in pregnancy at first AI, and an increase of 4.5 percentage points in pregnancy loss compared with POST TN. Exposure to HS during PRE and POST periods was associated with increased removal from the herd. In parous cows, exposure to HS during PRE, POST, or both PRE and POST was associated with a 2.4-kg/cow per day reduction in milk yield when compared with TN-TN. There was an interaction between HS exposure PRE and POST and incidence of retained placenta, because POST HS was associated with an increase in incidence of 5.8 percentage points within PRE HS cows, whereas no difference was found within PRE TN cows. Postpartum HS was associated with an increase of 6.3 percentage points in incidence of metritis and tended to be associated with an increase of 1.9 percentage points in mastitis incidence compared with POST TN. There was an interaction between HS exposure PRE and POST and pregnancy per AI because POST HS was associated with reduced pregnancy incidence by 10.6 percentage points within PRE TN cows, whereas no difference was found within PRE HS cows. Removal from the herd increased in cows exposed to HS during the PRE or POST or PRE and POST. These data suggest that POST HS is associated with performance losses to a greater extent than pre-partum HS and that nulliparous and parous cows are prompt to losses associated with exposure to HS during the transition period. The results corroborate findings from manipulative experiments that showed improved milk production and reproduction in cows under HS provided with heat abatement but adds insights to the associations between HS and health.


INTRODUCTION
Genetic selection combined with improved management of dairy cows have resulted in an increase of more than 400% in individual cow milk yield over the last century (Vandehaar and St-Pierre, 2006).As production increases, a parallel increase in DM intake occurs to support the nutritional needs for milk synthesis, which influences the metabolic rate and results in additional heat production (Polsky and von Keyserlingk, 2017).More productive cows must dissipate additional heat, which makes them less capable of maintaining body temperature and more susceptible to heat stress (HS; Hahn, 1999).In US dairies, annual economic losses associated with HS are estimated to be close to $900 million (St-Pierre et al., 2003).Hence, attention to the detrimental effects of HS on cow performance and farm profitability has been extensively evaluated, including periods when cows are more challenged such as the transition period.
Throughout the prepartum period, cows undergo a process of tissue remodeling, in which mammary gland epithelia undergo apoptosis and cell proliferation to reestablish the secretory cells for a new lactation (Capuco et al., 1997).This process of regeneration is regulated by programmed cell death of old epithelial cells, followed by proliferation of a new generation of epithelial cells and regeneration of the mammary gland (Capuco et al., 1997), which is correlated with milk yield in the subsequent lactation (Steeneveld et al., 2013).Heat stress impairs proper involution of the mammary gland due to the reduced autophagic activity (Wohlgemuth et al., 2016) and reduced cell proliferation (Tao et al., 2011), consequently compromising productive performance in the subsequent lactation (do Amaral et al., 2009;Fabris et al., 2019).Similarly, postpartum cows exposed to HS have reduced number of mammary epithelial cells resulting in reduced milk yield (Collier et al., 2006).Yet, the interaction between HS exposure during the prepartum and postpartum periods on productive performance remains uncertain.
Peripartum diseases, such as metritis and mastitis, are highly prevalent during the early postpartum period with significant consequences to productive and reproductive performance and survival of dairy cows (Gröhn and Rajala-Schultz, 2000;Carvalho et al., 2019).Adequate function of innate immune cells is essential to protect the uterus and mammary gland tissue from pathogenic bacteria (Bronzo et al., 2020).Providing evaporative cooling to cows exposed to HS increased proliferation of lymphocytes (do Amaral et al., 2010) and improved phagocytosis and oxidative burst of neutrophils (do Amaral et al., 2011).Thus, it is clear that HS and consequent hyperthermia affect immune cells, which may predispose cows to diseases.In fact, Thompson and Dahl (2012) reported that cows exposed to HS during the prepartum period had greater incidence of retained placenta, mastitis, and respiratory problems.Heat stress also has marked negative effects on reproduction in dairy cows (Thatcher, 1974), likely because of the negative effects of hyperthermia on follicle growth, oocyte quality, ovulation, early embryo development, endometrial function, and the endocrine hormonal milieu (Wolfenson and Roth, 2019).
The literature has limited information on the timing during lactation when HS has its greatest effects and scarce information exists about the combined effects of HS in late gestation and early lactation on production, health, reproduction, and survival in dairy cows.Hence, the objectives were to determine the association between timing of HS exposure during the transi-tion period on production, health, reproduction, and survival during the first 90 d postpartum in Holstein cows.The combined exposure to HS during the preand postpartum periods was hypothesized to have a greater effect on cow performance than HS during only prepartum or only postpartum periods, and the latter 2 were expected to be detrimental to cows compared with no HS exposure.Furthermore, the effects of HS were expected to affect both nulliparous and parous cows.

Study Design and Sample Size
The study design was a retrospective cohort study.The exposure factor was HS during the transition period.The sample size was of at least 1,000 cows per group to provide sufficient power (α = 0.05; β = 0.20) to detect significant associations between environmental condition (EC) and milk yield (difference between groups of at least 1 kg/d, SD = 7), morbidity (difference of 5 percentage units between 2 of the 4 EC groups when morbidity ranges from 38 to 43%), pregnancy per AI (P/AI; difference of 5 percentage units between 2 of the 4 EC groups when P/AI ranges from 35 to 40%), and survival (difference of 3 percentage units between 2 of the 4 EC groups when survival in the herd ranges between 80 and 85%).

Cows and Herds
Data from 2 commercial herds located in central California were collected for this retrospective cohort study.Cows were classified based on temperaturehumidity index (THI) exposure during the last 28 d of gestation (PRE) and the first 28 d postpartum (POST) and considered as exposed to HS when that the average THI in PRE and POST periods were greater than 68 (Fabris et al., 2019).To address the exposure to HS during either PRE or POST or to both periods combined, only cows that calved in April, June, July, and September of each year (2012, 2013, and 2014) remained in the study.A total of 5,722 Holstein cows (2,324 nulliparous and 3,397 parous cows) were used for the study and categorized into EC groups based on THI exposure during PRE and POST as HS or thermoneutral (TN), resulting in 4 different groups: TN-TN (n = 1,040; mean THI PRE = 57.9± 0.4 and POST = 63.7 ± 0.3), TN-HS (n = 1,214; mean THI PRE = 68.2± 0.4 and POST = 72.0 ± 0.3), HS-TN (n = 1,917; mean THI PRE = 71.4± 0.4 and POST = 65.6 ± 0.3), and HS-HS (n = 1,551; mean THI PRE = 72.2± 0.4 and POST = 72.7  et al., 2017), and cows with a gestation length shorter than 256 or longer than 296 d were not included in the study.For consistency, prepartum nulliparous cows that became primiparous postpartum were designated as nulliparous cows throughout the manuscript, and prepartum parous cows that became postpartum multiparous were designated as parous cows.
Herd 1 milked 1,630 cows 4 times a day during the POST and then 2 times a day.The rolling herd average for 2013 was 12,500 kg of 3.5% FCM.Yields of milk were measured in 2 consecutive milkings once a month by the local DHIA laboratory (Hanford, CA).Cows were housed in dry lots with shades in the central area of the pen and over the feed bunk.Dry composted manure was used as bedding material under the shade structures, and it was added twice weekly and raked once daily.Soaker lines with nozzles were placed above the stanchions in the feed lane, and were activated for 1 min each 6 min when the ambient temperature reached 22°C.Cows were dried once weekly at 232 ± 3 d of gestation or earlier if production was too low.Dry cows were moved to a dry-lot pen.Nulliparous and parous cows were moved into separate prepartum dry-lot pens once weekly at 255 ± 3 d of gestation where they remained until calving.Postpartum primiparous and multiparous cows were grouped separately throughout the entire dry and lactating period, respectively.Cows received treatments with bST (Posilac, 500 mg sometribove zinc suspension for injection, Elanco Animal Health) every 14 d starting at 70 d postpartum until 14 d before the expected dry off date.
Herd 2 milked 5,230 cows 3 times a day during the entire lactation.The rolling herd average for 2013 was 13,635 kg of 3.5% FCM.Yields of milk were measured daily using automated electronic milk meters (Perfection 3000, Boumatic, Madison, WI).Lactating cows were housed in freestall barns with sand as the bedding material.Dry cows were also housed in freestall barns, but then moved to dry-lot pens in the prepartum groups until parturition.Freestall barns were equipped with fans in the central area above the beds.Soaker lines with nozzles were present in all freestall bans and dry-lot pens.They were placed above the stanchions, and were activated for 1 min each 6 min when the ambient temperature reached 22°C.
Cows were dried off once weekly at approximately 232 ± 3 d of gestation or earlier if production was too low.Dry cows were housed in freestalls, but cows and heifers in the last 3 wk of gestation were moved to prepartum dry lots and they calved either in the dry lots or in group maternity pens.Dry cows were fed once daily a TMR formulated to limit weight gain when the average intake was 13 kg/d of DM.Nulliparous and parous cows were moved into a prepartum pen at approximately 255 d of gestation and fed similar diets, except that parous cows were fed an acidogenic diet to minimize risk of hypocalcemia at the onset of lactation.Prepartum nulliparous and parous cows and postpartum primiparous and multiparous cows were grouped separately throughout the entire dry and lactating period, respectively.Cows received treatments with bST (Posilac; Elanco Animal Health) every 14 d starting at 70 d postpartum until 14 d before the expected dry off date.
In both herds, prepartum cows were fed once daily.Diets for prepartum nulliparous and parous cows differed in protein and mineral content.Nulliparous prepartum cows were fed diets with additional protein content and without addition of any acidogenic supplement, which was included in the diets of the prepartum parous cows.In herd 1, postpartum cows were fed twice daily, whereas in herd 2, postpartum cows were fed once daily.Lactating cow diets in both herds were formulated to meet or exceed the nutrient requirements of a 680-kg cow consuming 27 kg of DM and producing 45 kg of milk with 3.70% fat and 3.30% true protein (NRC, 2001).

Data Collection
Weather data were obtained from the National Oceanic and Atmospheric Administration for the years of 2012, 2013, and 2014, from the station located at the Visalia Municipal Airport, Visalia, California, approximately 10 km from both farms.The data collected included the daily average dry bulb temperature and the relative humidity, which was used to calculate the THI as previously described (Ravagnolo and Misztal, 2000): where T is temperature (°C) and RH is relative humidity.
Health, production, reproduction, and survival data in the first 90 d postpartum were retrieved from the dairy management software Dairy Comp 305 (Valley Agricultural Software).All disease diagnoses were performed by herd personnel that were previously trained by veterinarians from the University of California Davis Veterinary Medicine Teaching and Research Center.Herd personnel evaluated cows daily after the first milking of the day for any signs of clinical disease, and treated cows accordingly.Cows were diagnosed as having retained placenta if the fetal membranes were not expelled within 24 h of parturition.Metritis was diagnosed based on transrectal palpation of the uterus for detection of fetid reddish to brownish uterine discharge that may or may not be accompanied by fever, and cows were treated with antimicrobial therapy (Excenel RTU sterile suspension, 50 mg/mL of ceftiofur as hydrochloride salt, Zoetis).Mastitis was diagnosed based on presence of abnormal milk with presence of flakes accompanied or not by visible inflammation of the mammary gland.Lameness was diagnosed based on visual appraisal of lameness score during standing and walking followed by routine therapeutic hoof trimming.Cows with reduced intake and interest in feed were checked for displaced abomasum, which was diagnosed by auscultation and percussion of the left paralumbar fossa region with a classical resonant sound followed by confirmation during corrective surgery.Hyperketonemia was not diagnosed in those cows.It is important to highlight that those cows calved between 2012 and 2014, and at the time it was not as common as it is today to have standard operation procedures for the diagnosis of hyperketonemia.Pneumonia was diagnosed based on increased respiratory frequency, presence of abnormal lung sounds at auscultation, and presence of fever.Morbidity was evaluated during the first 90 d postpartum and was defined as cows having at least one of the previously described diseases.Removal from the herd either by culling or death also was recorded.

Milk Yield
In herd 1, production of milk was measured once a month by the local DHIA laboratory (Hanford, CA).Production was measured for the first 3 mo postpartum and it was assumed that milk yield observed on that day represented the mean production of the respective month postpartum.The cumulative milk yield by 90 d postpartum for each cow was calculated using the monthly DHI milk test multiplied by 30.If a cow left the herd before 90 DIM, then the production in the month the cow left the herd was multiplied by the number of days she stayed in the herd on that month.In herd 2, daily milk yield was used to calculate the mean monthly milk yield for analysis of daily milk production.The cumulative milk yield by 90 d postpartum was computed by summing the daily milk production in the first 90 DIM or until the cow left the herd if before 90 DIM.

Reproductive Management
Both herds used the same reproductive management program.Synchronization of estrous cycle was performed by administration of 25 mg of PGF 2α (Lutalyse Sterile Solution, 5 mg/mL dinoprost as tromethamine salt, Zoetis) 14 d apart at 37 and 51 ± 3 d postpartum.Detection of signs of estrus based on removal of chalk from the tail started at 51 ± 3 d postpartum, and those in estrus were inseminated on the same day.Cows not inseminated within 11 d after the second dose of PGF 2α treatment were enrolled in the Ovsynch-56 timed AI protocol starting on d 62 ± 3 postpartum, such that every cow was programmed to receive their first AI by 75 ± 3 d postpartum.Cows were observed for estrus daily as previously described and those that returned to estrus were inseminated in the same morning and considered to be nonpregnant to the previous breeding.Pregnancy was diagnosed by transrectal ultrasonography on d 32 ± 3 after AI.The presence of an amniotic vesicle containing an embryo with a heartbeat was used as the criteria to determine pregnancy.Pregnant cows on d 32 after AI were reexamined for pregnancy 4 wk later, on d 60 of gestation.Pregnancy per AI was calculated by dividing the number of cows diagnosed pregnant at d 32 or 60 after AI by the number of cows receiving AI.Reproductive data were collected only for the first AI.Pregnancy loss was calculated as the

Statistical Analysis
Nulliparous and parous cows were analyzed separately using the same models to facilitate interpretation of the data and because there were no changes in the interpretation when analyzed together.The continuous dependent variables analyzed included daily milk yield and cumulative milk yield in the first 3 mo of lactation.The binary dependent variables analyzed included risk of individual diseases (retained placenta, metritis, mastitis, and lameness) and of morbidity, pregnancy at fist AI, pregnancy loss, and survival by 90 d postpartum.
The responses analyzed for time to event included days to morbidity and days to leaving the herd.
Daily or cumulative milk yield were analyzed by ANOVA using the MIXED procedure of SAS (version 9.4; SAS/STAT, SAS Institute Inc.).Distribution of residuals and homogeneity of variance were evaluated after fitting the statistical models for continuous data.
For the average or cumulative milk yield, the linear mixed-effects model was: where β 0 is the intercept, β 1 represents the 3 estimated coefficients of the fixed effect of EC (TN-TN vs. TN-HS vs. HS-TN vs. HS-HS), β 2 represents the 2 estimated coefficients of the fixed effect of gestation length (short vs. normal vs. long), and β 3 represents the 5 coefficients for the random effect of herd-year.Herd-year was used because data were collected from 3 consecutive years in herd 1 (herd-year A, B, and C) and 2 consecutive years in herd 2 (herd-year D and E).
For the daily milk yield, the mixed-effects model with repeated measures was: where β 0 is the intercept, β 1 represents the 3 estimated coefficients of the fixed effect of EC (TN-TN vs. TN-HS vs. HS-TN vs. HS-HS), β 2 represents the 2 estimated coefficients of the fixed effect of gestation length (short vs. normal vs. long), and β 3 represents the 2 estimated coefficients for month postpartum (1 vs. 2 vs. 3), β 4 represents the 6 estimated coefficients for the interactions between EC and month postpartum, and β 5 represents the 5 coefficients for the random effect of herd-year, and β 6 represents the coefficients for the random effect of cow nested within EC.Month was the factor in the REPEATED statement, and the covariance structure with the smallest Akaike's information criterion was chosen, and therefore we used the first-order autoregressive structure for equally spaced measurements.The Kenward-Roger method was used to calculate the approximate denominator degrees of freedom for the F tests in the statistical models.
Risk of individual diseases or morbidity, P/AI, and pregnancy loss were analyzed by logistic regression using the GLIMMIX procedure of SAS fitting a binary distribution and a logit link function.For statistical analysis of incidence of retained placenta and metritis only, the average THI exposure during the first 5 d postpartum was used to categorize cows as being exposed to POST HS or POST TN.The rational for that was because retained placenta was diagnosed at either 1 or 2 d postpartum, whereas metritis was diagnosed with a mean (±SD) of 4.5 ± 2.4 d postpartum.Because mastitis and lameness were diagnosed with a mean of 31.0 ± 25.4 and 31.9 ± 23.8 d postpartum, the average THI exposure during POST was used.
The model for analysis of binary data was where the log of odds of the event was modeled with a logistic regression that included the intercept β 0 , β 1 representing the 3 estimated coefficients of the fixed effect of EC, β 2 representing the 2 estimated coefficients of the fixed effect of gestation length, and β 3 representing the 5 coefficients for the random effect of herd-year.The Kenward-Roger method was used to calculate the approximate denominator degrees of freedom for the F tests in the statistical models.The ILINK function was used to convert the odds into predicted probabilities of an event for the different EC.
The interval from calving to diagnosis of the first disease event or to removal from the herd were analyzed with Cox's hazard regression model using the PHREG procedure of SAS.Cows that did not develop morbidity or survived by 90 d postpartum were censored.The adjusted hazard ratio and the 95% confidence inter-val were calculated.The statistical model used for the Cox's hazard regression analyses was: where logarithm of the ratio of the hazard of time over the basal hazard was a function of the set of predictors β 1 representing the 3 estimated coefficients of the fixed effect of EC, β 2 representing the 2 estimated coefficients of the fixed effect of gestation length, and β 3 representing the 5 coefficients for the effect of herd-year.Orthogonal contrasts were used to evaluate the effects of EC exposure to HS during the prepartum period (TN-TN + TN-HS vs. HS-TN + HS-HS), the effects of EC exposure to HS during the postpartum period (TN-TN + HS-TN vs. TN-HS + HS-HS), and the interaction between EC exposure to HS during the pre and postpartum period (TN-TN + HS-HS vs. TN-HS + HS-TN).Statistical significance was considered at P ≤ 0.05, and tendency when 0.05 < P ≤ 0.10.Data are presented as least squares means and standard error of the mean, unless otherwise stated.

RESULTS
Descriptive statistics depicting the productive performance, incidence of disease, reproduction at first AI, and survival of cows included in the study according to herd and parity are presented in Table 1.Distribution of nulliparous cows based on EC category was 17.8, 23.8, 28.4, and 29.9% for TN-TN, TN-HS, HS-TN, and HS-HS, respectively, whereas the distribution for parous cows was 18.4, 19.4,37.0, and 25.2% for TN-TN, TN-HS, HS-TN, and HS-HS, respectively.

Nulliparous Cows
By design, cows classified as TN during the PRE (TN-TN and TN-HS) were exposed to smaller (P < 0.001) prepartum THI compared with those classified as HS (HS-TN and HS-HS; PRE TN = 63.0 vs. PRE HS = 71.7 ± 0.4; Figure 1A and Table 2).Moreover, cows classified as TN during POST (TN-TN and HS- TN) were exposed to smaller (P < 0.001) postpartum THI compared with those classified as HS (TN-HS and HS-HS; POST TN = 64.4 vs. POST HS = 72.3± 0.3).

Parous Cows
Cows classified as TN during PRE (TN-TN and TN-HS) were exposed to smaller (P < 0.001) THI compared with those classified as HS (HS-TN and HS-HS; PRE TN = 63.1 vs. PRE HS = 71.8± 0.4; Figure 1B and Table 3).In addition, cows classified as TN (TN-TN and HS-TN) during POST were exposed to smaller (P < 0.001) THI compared with those classified as HS (TN-HS and HS-HS; POST TN = 64.7 vs. POST HS = 72.4± 0.3).day and a total of 178 kg in the first 90 d postpartum compared with TN-TN (Table 2).In addition, repeated measure analysis was conducted to evaluate changes over time during the first 90 d postpartum and results are presented in Figure 2A.The β coefficients and respective standard error for predictors in the models for milk yield and cumulative milk are presented in Supplemental Table S3 (https://data.mendeley.com/datasets/y5x5jjp8bn/3).

Productive
Parous Cows.As with nulliparous cows, interactions (P ≤ 0.003) also were observed between PRE and POST exposure to HS for daily milk yield and cumulative milk yield during the first 90 d postpartum.Exposure to HS during the PRE or POST or PRE and POST was associated with a reduction (P < 0.05) in milk yield of 2.4 kg/cow per day resulting in a cumulative decrease of 269 kg in the first 90 d postpartum compared with TN-TN (Table 3).In addition, repeated measure analysis was conducted to evaluate changes over time during the first 90 d postpartum and results are presented in Figure 2B.The β coefficients and respective standard errors for predictors in the models for milk yield and cumulative milk are presented in Supplemental Table S4 (https://data.mendeley.com/datasets/y5x5jjp8bn/3).

Incidence of Diseases
Nulliparous Cows.Retained placenta was diagnosed in 8% (187/2,325; Table 1) of nulliparous cows.Exposure to HS in PRE and POST was not associated with the incidence of RP (Table 2).Metritis affected 47.2% (1,098/2,325; Table 1) of nulliparous cows.An interaction (P = 0.007) between PRE and POST exposure to HS was observed for the incidence of metritis.Exposure to HS during the PRE or POST or PRE and POST was associated with increased (P < 0.05) incidence of metritis of 12 percentage points compared with TN-TN (Table 2).Mastitis was diagnosed in 3.9% (92/2,325; Table 1) of the cows.Exposure to HS in POST was associated (P = 0.03) with increased incidence of mastitis (Post TN = 3.3 vs. POST HS = 5.2 ± 1.3; Within a row, values with different superscripts in the same row differed (P < 0.05). 1 TN = thermoneutral; HS = heat stress.Environmental condition (EC) was categorized based on temperature-humidity index (THI) exposure during the prepartum (PRE) and postpartum (POST) periods as TN-TN (mean THI PRE = 57.9± 0.4 and POST = 63.7 ± 0.3), TN-HS (mean THI PRE = 68.2± 0.4 and POST = 72.0 ± 0.3), HS-TN (mean THI PRE = 71.4± 0.4 and POST = 65.6 ± 0.3), or HS-HS (mean THI PRE = 72.2± 0.4 and POST = 72.7 ± 0.3). 2 Pre = effect of prepartum exposure to heat stress (TN-TN + TN-HS vs. HS-TN + HS-HS); Post = effect of postpartum exposure to heat stress (TN-TN + HS-TN vs. TN-HS + HS-HS); Pre × Post = interaction between the effect of prepartum and postpartum exposure to heat stress (TN-TN + HS-HS vs. TN-HS + HS-TN).
3 Evaluated in the first 90 d postpartum.For statistical analysis of incidence of retained placenta and metritis, the average THI exposure during the first 5 d postpartum was used to categorize cows as being exposure to Post HS or Post TN. 4 Morbidity defined as cows having at least one of the following diseases: retained placenta, metritis, mastitis, lameness, displaced abomasum, or pneumonia.sociation was found between HS exposure during the transition period and incidence of lameness.Morbidity affected 54.3% (1,263/2,325; Table 1) of nulliparous cows, and an interaction (P < 0.001) between PRE and POST exposure to HS was observed for morbidity.Exposure to HS during the PRE or POST or PRE and POST was associated with an 18 percentage-point increase (P < 0.05) in morbidity compared with TN-TN (Table 2).In addition, the hazard of morbidity in the first 90 d postpartum increased (P < 0.001) 72, 74, and 89% for TN-HS, HS-TN, and HS-HS, respectively, compared with TN-TN (Table 4; Figure 3A).Data from the logistical regression model for each individual disease with odds ratio and respective 95% confidence interval is presented in Supplemental Table S1 (https://data.mendeley.com/datasets/y5x5jjp8bn/3).The β coefficients and respective SE for predictors in the models for incidence of disease are presented in Supplemental Table S3 (https://data.mendeley.com/datasets/y5x5jj-p8bn/3).
Parous Cows.Retained placenta affected 9.1% (308/3,397; Table 1) of parous cows.An interaction (P = 0.02) between PRE and POST exposure to HS was observed for the incidence of retained placenta (Table 3).Among cows that experienced PRE HS, those that also experienced POST HS (HS-HS) had greater (P < 0.05) incidence of retained placenta compared with HS-TN.Metritis was diagnosed in 18.4% (624/3,397; Table 1) of parous cows.A tendency (P = 0.08) was observed for an association between prepartum HS exposure with decreased incidence of metritis (PRE TN = 27.4 vs. PRE HS = 23.6 ± 2.6; Table 3).Exposure to HS in POST was associated (P = 0.009) with increased incidence of metritis (POST TN = 19.7 vs. POST HS = 29.4 ± 3.0; Table 3).Mastitis affected 8.3% (282/3,397; Table 1) of parous cows.Exposure to HS in POST tended (P = 0.08) to be associated with increased incidence of mastitis (POST TN = 8.0 vs. POST HS = 9.9 ± 1.2; Table 3).Lameness was diagnosed in 10.4% (355/3,397; Table 1) of parous cows, but no association was found between HS exposure during the transition period and incidence of lameness.Morbidity affected 33.8% (1,149/3,397; Table 1) of parous cows.Exposure to HS in POST tended (P = 0.07) to be associated with  increased morbidity (Post TN = 41.2 vs. Post HS = 44.7 ± 2.3; Table 3).In addition, the hazard of morbidity in the first 90 d postpartum increased (P = 0.01) 26% in HS-HS compared with TN-TN (Table 4; Figure 3B).Data from the logistical regression model for each individual disease with odds ratio and respective 95% confidence interval is presented in Supplemental Table S1.The β coefficients and respective SE for predictors in the models for incidence of disease are presented in Supplemental Table S4 (https://data.mendeley.com/datasets/y5x5jjp8bn/3).

Reproductive Performance
Nulliparous Cows.Exposure to HS in PRE was not associated with pregnancy at first AI in nulliparous cows.Exposure to HS in POST tended (P = 0.07) to be associated with reduced P/AI on d 32 after the first AI (Post TN = 42.1 vs. Post HS = 38.1 ± 2.6; Table 2) and was associated (P = 0.01) with reduced P/AI on d 60 after first AI (POST TN = 38.7 vs. POST HS = 33.4± 2.5; Table 2) because nulliparous cows exposed to POST HS had increased (P = 0.04) pregnancy loss (POST TN = 7.8 vs. POST HS = 12.3 ± 2.5; Table 2).Data from the logistical regression model for each reproductive outcome with odds ratio and respective 95% confidence interval is presented in Supplemental Table S2 (https://data.mendeley.com/datasets/The effect of prepartum EC (P < 0.001), effect of postpartum EC (P < 0.001), interaction between the effect of pre-and postpartum EC (P = 0.009), effect of month (P < 0.001), and interaction between the effect of month and EC (P < 0.001).Panel B: effect of prepartum EC (P < 0.001), effect of postpartum EC (P < 0.001), interaction between the effect of pre and postpartum EC (P < 0.001), effect of month (P < 0.001), and interaction between the effect of month and EC (P < 0.001).×: Within month, TN-TN vs. TN-HS differ (P < 0.05); ♦: within month, TN-TN vs. HS-TN differ (P < 0.05); *: within month, TN-TN vs. HS-HS differ (P < 0.05).Data presented as LSM and error bars are SEM.y5x5jjp8bn/3).The β coefficients and respective SE for predictors in the models for reproductive outcomes are presented in Supplemental Table S3 (https://data.mendeley.com/datasets/y5x5jjp8bn/3).Parous cows.Interactions (P ≤ 0.01) between PRE and POST exposure to HS were observed for P/AI on d 32 and 60 after the first AI (Table 3).Exposure to HS in POST was associated with depressed P/AI, but the effect was greater in cows exposed to TN in PRE.There was no association between HS exposure and loss of pregnancy in parous cows.Data from the logistical regression model for each reproductive outcome with odds ratio and respective 95% confidence interval is presented in Supplemental Table S2.The β coefficients and respective standard error for predictors in the models for reproductive outcomes are presented in Supplemental Table S4.
Parous Cows.By 90 d postpartum, 6.9% (235/3,397; Table 1) of parous cows were sold.An interaction (P = 0.03) between PRE and POST exposure to HS was observed for culling because within cows exposed to PRE TN, POST HS tended (P = 0.07) to be associated with an increase in culling by 2.9 percentage points, whereas there was no association for cows exposed to HS during the PRE (Table 3).Mortality affected 3.0% (102/3,397; Table 1) of parous cows and exposure to HS in POST was associated (P = 0.004) with increased mortality (POST TN = 2.2 vs. POST HS = 4.0 ± 0.7; Table 3).Overall, 9.9% (337/3,397; Table 2) of parous left the herd by 90 d postpartum.A tendency for an interaction (P = 0.09) between PRE and POST exposure to HS was observed for removal from the herd.Exposure to HS during the PRE or POST or PRE and POST were associated with greater (P < 0.05) removal from the herd by 5 percentage points compared with TN-TN (Table 3).In addition, the hazards of removal from the herd were 60, 48, and 59% greater (P < 0.05) for TN-HS, HS-TN, and HS-HS, respectively, compared with TN-TN (Table 5; Figure 3D).The β coefficients and respective standard error for predictors in the models for removal from the herd outcomes are presented in Supplemental Table S3.

DISCUSSION
This study demonstrates the association between HS exposure throughout the late gestation and the early lactation period with production, risk of diseases, reproduction, and survival in the first 90 d postpartum in Holstein dairy cows.The literature is vast on experiments evaluating the effects of providing evaporative cooling to cows exposed to HS on productive performance; however, limited data exist from controlled experiments evaluating the effect of providing heat abatement on health and survival in dairy cows.The present observational study was designed to attempt to isolate the effects of exposure to HS at critical times of the lactation cycle, late gestation and early lactation on postpartum performance.Although an observational study is limited as there are cofounders in addition to the variable of interest, our results showed that cows exposed to HS in either PRE or POST were associated with impaired production, reproduction, and more likely to develop early lactation diseases.In almost all instances, exposure to HS in either PRE or POST or in both periods was associated with depressed production, reproduction, and increased risk of diseases in nulliparous and parous cows.
It is well established that reducing HS by providing evaporative cooling during the entire dry period (Tao and Dahl, 2013) or during lactation (Huber et al., 1994) benefits lactation performance in dairy cows, although most of the data from controlled experiments involved parous cows.Parity is an important factor to be considered because nulliparous and parous cows do not always have the same response to dietary and management interventions during the transition period (Husnain and Santos, 2019;Santos et al., 2019).Nulliparous are still growing and nutritional needs need to  address accretion of tissue.Also, nulliparous undergo extensive mammogenesis during late pregnancy.Thus, it is possible that effects of exposure to HS might differ between the 2 parity groups.The present data clearly showed that both nulliparous and parous cows suffer when exposed to HS in either PRE or POST or in both periods.Nevertheless, it is important to point out that responses were not always the same between the parity groups.Exposure to HS was associated with increased risk of metritis in nulliparous cows regardless of if HS exposure occurred PRE or POST or PRE and POST, whereas in parous cows only POST HS was associated with increased metritis, and PRE HS was associated with decreased metritis.Nulliparous cows typically have greater incidence of metritis than parous cows (Table 1), likely related to dystocia and trauma during calving.Uterine diseases are linked with reduced DMI and altered feeding behavior prepartum (Hammon et al., 2006;Huzzey et al., 2007), thus suggesting that reduced intake might be a risk factor for uterine diseases.Nulliparous cows are known to have less DMI as percentage of BW than parous cows (Hayirli et al., 2002), whereas their needs for tissue accretion and growth are greater than those in parous cows (NRC, 2001).It is possible that the reduced DMI typically caused by HS prepartum affects the supply of nutrients to a larger extent in nulliparous than parous cows that predisposes them to uterine diseases, including metritis.Differences in response to HS between parity groups were not limited to risk of metritis.Exposure to POST HS in nulliparous cows was associated with a depression in P/ AI, whereas in parous cows the effect of POST HS was dependent on PRE HS and the greatest decrease in P/ AI associated with POST HS was observed in parous cows exposed to PRE TN.One of the most common observations in cows that suffer from HS is the reduction of milk yield.Although it has been suggested that early lactation cows suffer less from the detrimental effects of HS compared with mid and late lactation cows (Perera et al., 1986), the present findings reveal that exposure to HS at any time in the transition period was associated with depressed production in both nulliparous and parous cows.In addition, controlled experiments showed that providing evaporative cooling either during the entire dry period (Fabris et al., 2019;Tao and Dahl, 2013) or during lactation (Huber et al., 1994) markedly improves productive performance in dairy cows.Holstein cows are adapted to temperate climate and exposure to temperatures above 25°C induces increased heat loss to maintain body temperature (Berman, 2005).When exposed to high ambient temperature or high THI, dairy cows are unable to properly thermoregulate and they often respond with reduced DMI, which is thought to explain approximately half of the reduced milk yield (Baumgard and Rhoads, 2012).Capuco et al. (2003) showed that milk synthesis is based on the number and secretory activity of mammary epithelial cells.Cows exposed to HS have reduced number of mammary epithelial cells resulting in reduced milk yield (Collier et al., 2006).Thus, the combined reduction in nutrient supply and direct effects of hyperthermia on the mammary gland likely explains the reduced milk production in cows exposed to HS (Collier et al., 1982;West, 2003).Furthermore, acute HS alters energy metabolism and shifts tissue sensitivity to insulin (Wheelock et al., 2010;Baumgard and Rhoads, 2012), which alters the supply of nutrients to the mammary gland.In addition to the aforementioned effect, we showed that HS increases the risk of morbidity in early lactation, and there is a strong link between peripartum diseases and production in dairy cows (Carvalho et al., 2019).Altogether, exposure to HS during late gestation and early lactation depresses productive performance by affecting mammogenesis, reducing DMI, altering supply of nutrients to the mammary gland, and increasing the risk of peripartum diseases.
Adequate innate immune function is essential to protect cows against infections (Bronzo et al., 2020).Providing evaporative cooling to cows exposed to HS enhances in vitro innate immune cell function based on increased cytokine production (do Amaral et al., 2009), lymphocytic proliferation, and TNFα expression (do Amaral et al., 2010(do Amaral et al., , 2011)).Likewise, an in vitro experiment demonstrated that PMNL had less phagocytosis and oxidative burst when exposed to increased temperature (Lecchi et al., 2016).Not only cows are affected by changes in environmental conditions, but increased temperature is associated with more pathogens in the environment of the cow (Godden et al., 2003;Makovec and Ruegg, 2003), which likely contributes to an increased risk of infectious diseases.High environmental temperature is recognized as an important risk factor for metritis and mastitis (DuBois and Williams, 1980;Hogan et al., 1989).These environmental conditions likely trigger a combination of impaired host resistance and increased pathogen challenge that favors disease.Indeed, in the present study exposure to HS at any time in the transition period in nulliparous and POST HS in parous cows was associated with increased risk of morbidity.Postpartum HS was associated with retained placenta, and although this disorder happens in the first 24 h of calving, it is important to recognize that this is an observational study and, therefore, changes in environmental conditions such as from TN to HS were not abrupt.For instance, cows classified as TN-HS were exposed to a mean THI equal to 68.2 during PRE; however, during the last few days of gestation, those cows were exposed to greater THI than the mean THI for the entire prepartum period resulting in some exposure to HS conditions.Also, it is important to describe that threshold used for THI or upper limit temperature to discriminate HS in dairy cows are usually based on effects on milk yield (Berman, 2005), and little is known about thresholds that might affect health.Thus, the association between retained placenta in POST HS is probably linked to the exposure to greater THI during the last few days of gestation which could affect proper immune function and ability of cows to detach the placenta adequately.
Hyperthermia is known to affect follicle growth, oocyte quality, early embryo development, endometrial function, and the endocrine milieu in dairy cows (Thatcher, 1974;Wolfenson and Roth, 2019).Indeed, the present study clearly showed that exposure to HS, particularly postpartum depressed P/AI in both nulliparous and parous cows, and increased the risk of pregnancy loss in nulliparous cows.Cows exposed to HS have impaired endocrine pathways that alter the production and secretion of ovarian hormones such as estradiol and inhibin (Wolfenson and Roth, 2019), which affects follicular dominance.Secretion of LH (Wise et al., 1988) and progesterone (Wolfenson et al., 2002) are altered by HS, which can impair final maturation and ovulation of the pre-ovulatory follicle and affect establishment and maintenance of pregnancy.Development of the pre-ovulatory follicle begins months before ovulation and insults such as HS (Torres-Júnior et al., 2008) or diseases (Ribeiro et al., 2016) can have long-lasting effects on reproduction.Torres-Júnior et al. (2008) reported that oocytes can be damaged by HS as early as 105 d before ovulation and Ribeiro et al. (2016) showed that inflammatory diseases that affect cows in early lactation reduces morula and conceptus quality and induces damages to the conceptus that carry on during later stages of pregnancy.Thus, it is possible that direct insults caused by HS and the associated increased risk of diseases observed in the present study explain the reduced P/AI and increased risk of pregnancy loss.
Survival in the first 90 d postpartum was affected by PRE and POST HS and both nulliparous and parous cows exposed to PRE or POST or PRE and POST HS had increased hazard of leaving the herd.Cows exposed to PRE or POST or PRE and POST HS had reduced milk yield and increased risk of morbidity, both of which are known to reduce survival of cows in the herd (Bascom and Young, 1998).Peripartum diseases can cause death and exposure to POST HS increased morbidity and death, the latter contributing to the increased removal from the herd in the first 90 d postpartum.Early lactation diseases are well known risk factors for survivability in dairy herd (Fetrow et al., 2006).Because removal from the herd was evaluated in the first 90 d postpartum only, it is unlikely that reproductive status played a role on decisions to retain or cull a cow.Thus, it is suggested that increased death, reduced production, and associated increased morbidity resulted in the increased hazard of leaving the herd and reduced survival by 90 d postpartum in cows exposed to HS.
Although the data presented in this study provide insights into the effects of HS on performance of both nulliparous and parous cows, we need to address the limitations of the study.Data used were collected from 2 dairy farms located in proximity in the southwestern United States with similar health and reproductive management but housing cows differently.It is possible that HS effects on dairy cows vary with region and management imposed.For instance, both farms did not provide adequate evaporative cooling to prepartum cows other than protection against solar radiation with shade and soaker lines in the feedbunk.It is very likely that the effects of HS might be attenuated in farms in which forced ventilation is available to all cows, dry and lactating (Fabris et al., 2019).Also, this is a retrospective cohort study that establishes associations, but not causation.We also caution readers that even though all disease diagnosis were performed by trained farm personnel, our data set comes from different herds and years, therefore it is possible to have variation due to change in farm personnel over the years that may affect disease diagnose.Nevertheless, we tried to minimize this by including 2 herds and data from 3 consecutive years.Finally, by default, the cows from the different EC groups did not calve at the same time of the year; therefore, as all observational studies evaluating EC, the present study suffers from the possible confounding factors that might have occurred over time in the 2 herds.Nevertheless, the results herein corroborate data from controlled experiments thus providing some support to the present findings.

CONCLUSIONS
Exposure to HS either in late gestation or in early lactation was associated with reduced milk production in the first 3 mo of lactation.Nevertheless, HS in POST was associated with performance losses to a greater extent than HS in PRE, mainly because of more adverse health and reproduction responses.Although most literature on HS has focused on parous cows, the present results clearly showed that the effects of adverse ambient temperature are detrimental to nulliparous cows as well.Exposure to HS in POST increased morbidity and reduced P/AI, both of which are known to affect survivability of cows in dairy herds.In fact, nulliparous and parous cows exposed to HS had increased hazard of leaving the herd in early lactation.In both studied herds, PRE and POST cows were provided some heat abatement with shade to reduce radiation and water spraying to increase conductive and convective cooling.Based on the current findings it is possible that heat abatement pre-and postpartum would improve performance but suggests that priority should be given to cows in early lactation, nevertheless controlled experiments evaluating the effect of heat abatement pre-and postpartum on health and reproductive performance warrant further research.

1
TN = thermoneutral; HS = heat stress.Environmental condition (EC) was categorized based on temperature-humidity index (THI) exposure during the prepartum (PRE) and postpartum (POST) periods as TN-TN (mean THI PRE = 57.9± 0.4 and POST = 63.7 ± 0.3), TN-HS (mean THI PRE = 68.2± 0.4 and POST = 72.0 ± 0.3), HS-TN (mean THI PRE = 71.4± 0.4 and POST = 65.6 ± 0.3), or HS-HS (mean THI PRE = 72.2± 0.4 and POST = 72.7 ± 0.3).2Pre = effect of prepartum exposure to heat stress (TN-TN + TN-HS vs. HS-TN + HS-HS); Post = effect of postpartum exposure to heat stress (TN-TN + HS-TN vs. TN-HS + HS-HS); Pre × Post = interaction between the effect of prepartum and postpartum exposure to heat stress (TN-TN + HS-HS vs. TN-HS + HS-TN).3Evaluated in the first 90 d postpartum.For statistical analysis of incidence of retained placenta and metritis, the average THI exposure during the first 5 d postpartum was used to categorize cows as being exposure to Post HS or Post TN.
Menta et al.: HEAT STRESS AND PERFORMANCE ± 0.3).Cows that calved in August were not included in the study to avoid large variation in the number of cows representing each environmental category group.For instance, if we included August in the data set, we would have an addition of approximately 1,500 cows into HS-HS group, and we thought that could inflate the statistical analysis.Gestation length was categorized as short if ≤269 d, normal if 270 to 282 d, or long if ≥283 d as previously described (Vieira-Neto Menta et al.: HEAT STRESS AND PERFORMANCE number of cows that lost pregnancy between gestation d 32 and 60 divided by the number of pregnant cows on d 32.

Table 1 .
Menta et al.: HEAT STRESS AND PERFORMANCE Descriptive statistics according to parity group and herd

Table 2 .
Association between environmental condition (EC) exposure during the transition period and production, incidence of diseases, reproduction, and survival during the first 90 d postpartum in nulliparous Holstein cows (LSM)

Table 3 .
Menta et al.:HEAT STRESS AND PERFORMANCE Association between environmental condition (EC) exposure during the transition period and production, incidence of diseases, reproduction, and survival during the first 90 d postpartum in parous Holstein cows(LSM and SEM) Within a row, values with different superscripts in the same row differed (P < 0.05).
Menta et al.: HEAT STRESS AND PERFORMANCE

Table 4 .
Cox's regression model for rate of morbidity in the first 90 d postpartum in nulliparous and parous Holstein cows

Table 5 .
Menta et al.: HEAT STRESS AND PERFORMANCE Cox's regression model for rate of removal from the herd in the first 90 d postpartum in nulliparous and parous Holstein cows Menta et al.: HEAT STRESS AND PERFORMANCE