Targeted progesterone supplementation improves fertility in lactating dairy cows without a corpus luteum at the initiation of the timed artificial insemination protocol

The study was conducted on a dairy farm in central Florida, milking 4,366 cows with yearly a rolling herd average milk yield of 10,949 kg. Weekly cohorts of cows were enrolled during 13 consecutive weeks and all inseminations were performed from January 27 to April 22, 2011. Primiparous (n = 515) and multiparous (n = 883) cows were housed separately in freestall barns equipped with sprinklers and fans. Cows received the same TMR to meet or exceed the nutrient requirements for a lactating Holstein cow producing 45 kg/d of milk with 3.5% fat and 3.2% true protein when DM intake is 25 kg/d (). Diet consisted of ryegrass silage, corn silage, ground corn, citrus pulp, solvent-extracted soybean meal, expeller soybean meal, corn gluten feed, molasses, minerals, and vitamins. Cows were fed twice and milked thrice daily.

All 1,398 Holstein cows were subjected to the 5-d timed AI program (Figure 1); 628 were enrolled at the first postpartum AI and 770 cows were enrolled during a resynchronized insemination after diagnosed as nonpregnant. Cows receiving the first AI postpartum had their estrous cycle presynchronized with 2 i.m. injections of PGF_2α (5 mL of Lutalyse Sterile Solution, dinoprost tromethamine, equivalent to 5 mg of dinoprost/mL; Pfizer Animal Health, Madison, NJ) administered 14 d apart, at 50 ± 3 and 64 ± 3 DIM. The timed AI protocol was initiated 12 d after the second PGF_2α of the presynchronization protocol, with mean and median DIM of 76 and 75, respectively. Cows receiving resynchronized inseminations were enrolled in the 5-d timed AI program following a nonpregnancy diagnosis at 34 ± 3 d after the previous insemination, with mean and median DIM of 156 and 127 DIM, respectively. The timed AI program consisted of an i.m. injection of 100 μg of GnRH (2 mL of Cystorelin, gonadorelin diacetate tetrahydrate equivalent to 43 μg of gonadorelin/mL; Merial Ltd., Duluth, GA) followed by 2 i.m. injections of PGF_2α 5 and 6 d later. A second i.m. injection of GnRH was administered concurrently with AI at 72 h after the first PGF_2α. For study purposes, the day of timed AI was labeled as study d 0 and the day of the first GnRH of the 5-d timed AI protocol as study d −8.

The tailheads of cows were painted daily with chalk starting on d −8 and removal of chalk was used as an indication of estrus. Cows detected in estrus between study d −8 and −3 were excluded from the statistical analyses. Cows detected in estrus from d −3 to −1 were inseminated on the same day and remained in the study.

All cows in the study had their ovaries evaluated by ultrasonography using a portable ultrasound equipped with a 7.5-MHz transrectal linear transducer (Easi-Scan, BCF Technology, Rochester, MN) on study d −8, and the location of CL and follicles greater than 10 mm in diameter were recorded. Cows bearing a CL greater than 10 mm in diameter were considered to be in diestrus (DI; n = 946). Cows lacking a CL were blocked by service number (first AI vs. resynchronized AI) and parity (primiparous vs. multiparous). Within each block, cows were allocated randomly into one of 2 treatments; progesterone supplementation or control. Cows assigned to receive progesterone supplementation (2CIDR; n = 218) were treated with 2 CIDR inserts, each containing 1.38 g of progesterone (Eazi-Breed CIDR Cattle Insert, Pfizer Animal Health, Madison, NJ) from study d −8 to −3; whereas control cows (CON; n = 234) received no supplemental progesterone. The proportions of CON, 2CIDR, and DI cows enrolled in the study in the first AI were 62.0, 64.7, and 36.1%, respectively.A second ultrasound evaluation of the ovaries on study d −3 was performed in all CON (n = 234) and 2CIDR cows (n = 218), and in a subset of DI cows (n = 354). Ovulation in response to the GnRH injection on study d −8 was considered when the cow had a follicle ≥ 10 mm on d −8 and a new CL was observed on d −3. Cows with follicles <10 mm on study d −8, but with a new CL on study d −3 were considered to have a new CL, but ovulated before study d −8.A subset of 56 CON, 47 2CIDR, and 49 DI cows had their ovaries scanned (Aloka SSD-500 equipped with a 7.5-MHz linear transducer, Aloka Co., Tokyo, Japan) on d 0 and 2, corresponding to the day of timed AI and again 48 h later. The diameter of all follicles greater than 5 mm was measured. Ovulation to the GnRH on d 0 was defined as the disappearance of one or more follicles ≥ 10 mm in diameter within 48 h.

Blood was sampled from a subset of 71 cows (CON = 24; 2CIDR = 24; DI = 23) by puncture of the coccygeal vein or artery into evacuated tubes containing K₂ EDTA (Vacutainer; Becton Dickinson, Franklin Lakes, NJ) on study d −8, −7, −5, −3, −2, 0, 2, 5, 8, 11, and 14. Samples were placed immediately in ice. Blood tubes were centrifuged at 3,000 × g for 15 min at 4°C for plasma separation within 6 h of collection; plasma samples were frozen at −20°C until assayed. Concentrations of progesterone were analyzed in samples by RIA using a commercial kit (Coat-a-Count, Siemens Healthcare Diagnostics, Los Angeles, CA). The sensitivity of the assay was 0.15 ng/mL, calculated as 2 SD below the mean counts per minute at maximum binding. Samples with known concentrations of progesterone (1.5 and 2.5 ng/mL) were incorporated into the assay for quality control. All samples were assayed in a single assay and the intra-assay CV was 6.4%. Concentrations of estradiol were analyzed in plasma samples collected on study d −3 and 0 by RIA using a commercial kit (Estradiol Double Antibody, Siemens Healthcare Diagnostics) validated previously for use in bovine samples (). The sensitivity of the assay was 0.55 pg/mL calculated as 2 SD below the mean counts per minute at maximum binding. Samples were analyzed in duplicate and repeated when the CV between duplicates was >15%. Plasma from a cow exhibiting standing activity, charcoal-stripped plasma from a male calf, and samples with known concentrations of estradiol (2.5, 5.0, and 10.0 pg/mL) were incorporated into the assay for quality control. Samples were analyzed in 4 assays; the intraassay CV averaged 12.5% and the interassay CV was 14.8%.

Pregnancy was diagnosed by transrectal ultrasonography on d 34 after AI. The presence of an amniotic vesicle containing an embryo with a heartbeat was used as the determinant of pregnancy. Pregnant cows on d 34 were reexamined for pregnancy by transrectal palpation 4 wk later, on d 62 of gestation. Pregnancy per AI was calculated by dividing the number of cows diagnosed pregnant at d 34 or 62 after AI by the number of cows receiving AI. Pregnancy loss was calculated as the number of cows that lost a pregnancy between d 34 and 62 after AI divided by the number of cows diagnosed pregnant on d 34 after AI. Cows that were detected in estrus before study d 32 were reinseminated and considered nonpregnant. Short AI interval was defined for cows reinseminated from study d 5 to 17.

Body condition of all cows was scored () on study d −8. For statistical analysis, BCS was categorized as low (≤ 2.50) or high (≥ 2.75). Yields of milk were recorded for individual cows once monthly using on-farm milk meters (Tru-Test Ltd., Manukau, New Zealand). Measurements for the first 3 mo postpartum were used to derive an individual 305-d production estimate (). Within parity group, cows were categorized according to milk production as being above or below the average for statistical analyses.

Cows lacking a CL on study d −8 were assigned to treatments in a randomized block design. Cows were blocked by service number (first postpartum AI or resynchronized AI) and parity (primiparous or multiparous) and randomly assigned to either CON or 2CIDR. By design, contemporary herdmates that had a CL on d −8 were labeled as DI and used as a positive control to evaluate the effectiveness of progesterone supplementation in restoring fertility of cows without CL relative to a cow with CL. We have previously shown that cows without a CL at the beginning of the synchronization protocol have a 13-percentage-point decrease in P/AI compared with cows in diestrus (). Similarly, anovular cows had a 12-percentage-point decrease in P/AI compared with estrous cyclic cows (). It was assumed that cows in 2CIDR would have similar P/AI to that of DI cows, and both would be 13 percentage points greater than cows on the CON group. Therefore, a sample size calculation was performed using the POWER procedure of SAS version 9.3 (SAS Institute Inc., Cary, NC) to allow for sufficient experimental units to detect an increase in P/AI of 13 percentage points following progesterone supplementation of cows without a CL (CON vs. 2CIDR; α = 0.05; β = 0.20). Pregnancy per AI for CON cows was assumed to be 30% (). Under such assumptions, a total of 215 cows per treatment were calculated to be required.Data were analyzed initially with only randomly assigned cows (CON and 2CIDR) and then with all 3 treatments (CON, 2CIDR, and DI). Because estimates of differences between CON and 2CIDR remained basically the same when both strategies were used, the approach with 3 treatments was selected to identify differences in fertility responses between CON and DI and between 2CIDR and DI.Binary responses were analyzed by logistic regression using the GLIMMIX procedure of SAS. Parity (primiparous vs. multiparous), BCS category (≤ 2.50 or ≥ 2.75), milk yield category (above or below the mean milk production within parity), service number (first postpartum AI vs. resynchronized AI), and the respective interactions with treatment were included as covariates. For P/AI and pregnancy loss, sire, technician, and their interactions with treatment were also included in the statistical models. Multivariable logistic models were built and a backward stepwise elimination method was applied, in which covariates were continuously removed when P > 0.10. Treatment was forced into the final model in all analyses. To evaluate if cows in estrus at AI had differential response to treatments, an additional model was built to analyze P/AI and pregnancy loss with the effects of estrus at AI and the interaction between treatment and estrus at AI. Adjusted risk ratios (ARR) and respective 95% CI were estimated by analyzing binary data with a modified Poisson regression model with the GENMOD procedure of SAS using a log link function and correction for data dispersion (; ).Pregnancy per AI was also analyzed for the subset of 806 cows in which ovarian responses on d −8 and −3 were evaluated by ultrasonography. The statistical models were the same as indicated above, but also included the presence of a new CL on d −3 and the interaction between treatment and presence of new CL.Continuous and discrete variables were analyzed by ANOVA using the GLIMMIX procedure of SAS fitting a normal distribution. Tests for normality of residuals and homogeneity of variances were conducted for each variable using the guided data analysis option of SAS. Data which did not fulfill the assumptions of ANOVA were transformed accordingly. Back-transformed data using the ilink function of SAS are presented for clarity. The effects of treatment, parity, and interaction between treatment and parity were included in the model as independent variables. Concentrations of progesterone and estradiol were analyzed as repeated measures and cow nested within treatment was considered a random effect. The concentration of progesterone on the day of AI was included as a covariate in the analysis of progesterone concentrations from study d 2 to 14. The first-order autoregressive was selected as covariance structure because it presented the smallest Schwarz's Bayesian information criterion value.Treatment differences with P ≤ 0.05 were considered significant and 0.05 < P ≤ 0.10 were considered as tendencies.

The ovulatory response to the first GnRH was not affected by treatment (Table 1). However, because some of the cows not bearing a CL were likely in metestrus on d −8, a greater (P = 0.001) proportion of CON and 2CIDR cows had a newly formed CL on d −3 compared with DI cows. Both the ovulatory response to the first GnRH on d −8 (58.2 vs. 46.7%; ARR = 1.24; 95% CI = 1.08 to 1.43) and the presence of a newly formed CL on d −3 (63.7 vs. 51.4%; ARR = 1.22; 95% CI = 1.08 to 1.39) were greater (P < 0.001) for cows receiv ing the first AI postpartum than for those receiving resynchronized AI. Only 2.3% (8 of 354) of DI cows did not present a visible CL on d −3. Therefore, in spite of a smaller occurrence of newly formed CL, the proportion of cows with a CL on d −3 was greater (P = 0.001) for DI compared with CON and 2CIDR cows (Table 1). Cows subjected to the first AI postpartum were more likely (P < 0.001) to bear a visible CL at PGF_2α than those receiving resynchronized AI (81.7 vs. 74.8%; ARR = 1.12; 95% CI = 1.04 to 1.20). The proportion of cows that had a synchronized ovulation within 48 h of the second injection of GnRH of the timed AI was greater (P = 0.05) for DI than for CON (ARR = 1.16; 95% CI = 1.03 to 1.32) and intermediate for 2CIDR cows. An interaction (P = 0.02) between treatment and presence of a newly formed CL on d −3 was observed for the diameter of the ovulatory follicle on d 0 (Table 1). Diameter of the ovulatory follicle did not differ among treatments in cows without a newly formed CL; however, for cows that had a newly formed CL on d −3, the diameter of the ovulatory follicle was larger (P < 0.05) for CON than 2CIDR (Table 1).

The concentration of progesterone in plasma on study d −8 was greater (P < 0.001) for cows in DI than for those not bearing a CL (Figure 2). As expected, CON and 2CIDR cows had similar (P = 0.88) concentrations of progesterone on the d −8. The incorporation of 2 CIDR inserts to the timed AI program increased (P < 0.05) circulating concentrations of progesterone, which were consistently greater for 2CIDR than for CON cows throughout the timed AI program. A treatment by day interaction was observed (P < 0.001) for circulating progesterone from study d −7 to −3. This interaction occurred because concentrations of progesterone in CON and DI cows increased from d −7 to d −3, whereas 2CIDR cows had a slight decline in progesterone concentrations during the same period (Figure 2). Concentrations of progesterone were greater (P = 0.02) for DI (0.31 ng/mL) compared with 2CIDR and CON cows (0.19 ng/mL) on d 0. Treatment did not affect progesterone concentrations during the first 14 d after AI (Figure 3).

Although the overall proportion of cows detected in estrus at AI was similar among treatments (Table 2), an interaction (P < 0.001) between treatment and the presence of a newly formed CL on d −3 influenced detection of estrus. For cows without a new CL, detection of estrus was less for CON than 2CIDR and DI (21.8 vs. 41.0 vs. 47.0%). On the other hand, for cows with a newly formed CL, detection of estrus was greater for CON, followed by 2CIDR, and then DI (40.4 vs. 33.3 vs. 25.0%).

Circulating concentrations of estradiol did not differ among treatments on d −3, when the first PGF_2α of the 5-d timed AI protocol was administered (Table 1). However, on d 0, when AI was performed, CON and 2CIDR cows had lower (P < 0.001) concentrations of estradiol than DI cows.

Treatment affected (P < 0.01) P/AI on d 34 and 62 after insemination (Table 2). Pregnancy per AI for 2CIDR and DI did not differ; however, they were both greater (P < 0.01) than the CON. In fact, supplementing progesterone in cows without a CL on d −8 increased the risk of pregnancy on d 62 by 58% (ARR = 1.58; 95% CI = 1.30 to 1.93). No interactions between treatment and parity, milk yield, or BCS were observed for P/AI. Cows receiving the first AI postpartum, which was preceded by presynchronization with PGF_2α, had greater P/AI on d 32 and 64 than cows receiving resynchronized inseminations (Table 3). The benefits of supplementing progesterone to cows without a CL on d −8 were similar whether they received the first or resynchronized AI, and no differences were observed between 2CIDR and DI cows according to service number. When detection of estrus at AI was included in the statistical models, an interaction (P < 0.01) between treatment and estrus at AI was observed to affect P/AI on d 34 and 62 (Table 4). Cows in estrus had greater P/AI in all treatments; however, the benefits to pregnancy from supplemental progesterone were observed only among cows that were not detected in estrus at AI.

For the subset of 806 cows with ovarian responses on d −8 and −3, presence of a new CL influenced (P < 0.01) P/AI on d 34 and 62 after insemination. Furthermore, an interaction (P < 0.01) between treatment and presence of a new CL on d −3 was observed because the increment in P/AI for cows supplemented with progesterone was greater among those that did not have a new CL than in those with a new CL on d −3. On d 62, in cows with a new CL detected on d −3, the P/AI did not differ between 2CIDR and DI (47.4 vs. 51.1%), but were both greater (P < 0.01) than CON (37.2%). Similarly, for cows without a new CL on d −3, the P/AI did not differ between 2CIDR and DI (37.8 vs. 45.3%), but were both greater (P < 0.01) than CON (11.5%).

The proportion of cows that were reinseminated from d 5 to 17 after AI was greater (P = 0.02) for CON than DI and 2CIDR cows (Table 2). For the subset of 806 cows with ovarian responses on d −8 and −3, a tendency for an interaction (P = 0.08) between treatment and presence of new CL on d −3 was observed for the proportion of cows with short AI interval. For cows with a new CL, treatment had no effect (P = 0.97) on short AI interval (CON = 4.6 vs. 2CIDR = 4.8 vs. DI = 5.3%). Conversely, for cows without a new CL on d −3, those in 2CIDR or DI groups had a lower (P < 0.01) incidence (2.0 and 3.3%) of short AI interval than CON (19.4%).

Pregnancy loss between d 34 and 62 of gestation did not differ with treatments (Table 2) in both the first and the resynchronized AI (Table 3). Cows that showed estrus at AI had smaller (P < 0.01) pregnancy loss than those not detected in estrus (7.4 vs. 1.2%; Table 4).