If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Sydney School of Veterinary Science, The University of Sydney, Camden, New South Wales 2570, AustraliaDepartment of Veterinary Population Medicine, University of Minnesota, St. Paul 55108
Département de pathologie et microbiologie, Faculté de médecine vétérinaire, Université de Montréal, Saint-Hyacinthe, Quebec, Canada J2S 2M2Mastitis Network, Saint-Hyacinthe, Quebec, Canada J2S 2M2
Département de pathologie et microbiologie, Faculté de médecine vétérinaire, Université de Montréal, Saint-Hyacinthe, Quebec, Canada J2S 2M2Mastitis Network, Saint-Hyacinthe, Quebec, Canada J2S 2M2
Antibiotic stewardship on dairy farms can be heightened through the implementation of selective dry-cow therapy (SDCT). However, some producers are concerned that this practice may be related to poor udder health outcomes in cows with high milk production at the time of dry-off. The objective of this study was to evaluate if the effect of culture-guided SDCT (Cult-SDCT) and algorithm-guided SDCT (Alg-SDCT) on dry-period intramammary infection (IMI) dynamics and postcalving udder health and performance [when compared with blanket dry-cow therapy (BDCT)] varied according to milk production level before dry-off. Data were compiled from clinical trials conducted in the United States and Canada that compared Cult-SDCT and Alg-SDCT to a positive control, i.e., BDCT. In those trials, cows were enrolled 1–2 d before dry-off, randomized to their dry-cow therapy strategy and followed until 120 d in milk of the subsequent lactation. The number of cows and quarters in the final data set were 1,485 and 5,097, respectively. Measured outcomes included quarter-level antibiotic use at dry-off, quarter-level IMI prevalence after calving, quarter-level dry-period IMI cure risk, quarter-level dry-period new IMI risk, cow-level clinical mastitis and removal from the herd during 1–120 d in milk, and somatic cell count and milk yield during 1–120 DIM. The primary objective of analysis was to investigate if the effect of Cult-SDCT and Alg-SDCT on these outcomes, when compared with BDCT, varied according to milk production level before dry-off. To do this, each cow was classified as having low, mid or high production, based on her milk yield tertile group at the most recent herd test before enrollment (low: <23.7 kg/d, mid: 23.7 to 30.4 kg/d, and high >30.4 kg/d). Multivariable generalized estimating equations were used to estimate risk differences and differences in means, and Cox regression was used to estimate hazard ratios. For Cult-SDCT, the proportion of quarters treated with dry-cow antibiotics within each milk production level were 40.7% (low), 41.7% (mid) and 47.2% (high). For Alg-SDCT, the proportions were 60.6% (low), 38.7% (mid), and 35.1% (high). Measures of udder health were not markedly different when comparing Cult-SDCT to BDCT and Alg-SDCT to BDCT. This was consistently observed in low, mid and high producing cows. In conclusion, the findings from this study indicate that Cult-SDCT and Alg-SDCT can be successfully implemented in cows of all milk production levels.
Risk factors for clinical or subclinical mastitis following infusion of internal teat sealant alone at the end of lactation in cows with low somatic cell counts.
Postcalving udder health and productivity in cows approaching dry-off with intramammary infections caused by non-aureus Staphylococcus, Aerococcus, Enterococcus, Lactococcus, and Streptococcus species.
). In North America, most farmers practice blanket DCT (BDCT), which involves the treatment of all quarters at dry-off. However, an increasing number of producers are switching from BDCT to selective DCT (SDCT), which is an antibiotic-sparing strategy that uses a screening test to allocate treatments to cows or quarters that are most likely to benefit. The global shift away from BDCT is due to an increasing awareness of antimicrobial stewardship among producers, economic incentives, government mandates, and industry responses to consumer demands (
Recent field trials have demonstrated that implementation of culture-guided SDCT (Cult-SDCT) or algorithm-guided SDCT (Alg-SDCT) can reduce antibiotic use at dry-off by 21 to 58% without compromising cow health in the subsequent lactation (
Evaluation of selective dry cow treatment following on-farm culture: Risk of postcalving intramammary infection and clinical mastitis in the subsequent lactation.
Randomized controlled non-inferiority trial investigating the effect of 2 selective dry-cow therapy protocols on antibiotic use at dry-off and dry period intramammary infection dynamics.
). However, it has also been shown that in some situations (especially when internal teat sealants are not used) implementation of SDCT can increase IMI during the dry period and increase clinical and subclinical mastitis in the subsequent lactation (
). Furthermore, recent studies have found that SDCT implementation was associated with slightly lower bacteriological cures during the dry period and small increases in postcalving SCC (
Monitoring udder health on routinely collected census data: Evaluating the short- to mid-term consequences of implementing selective dry cow treatment.
). Consequently, despite significant evidence to support increased implementation of SDCT, there is still uncertainty among some producers regarding the health risks. For example, we have recently encountered several farmers with concerns that SDCT may lead to worse udder health in cows with high milk yield at the time of dry-off. To our knowledge, this hypothesis has not been investigated in the aforementioned clinical trials of SDCT. However, observational studies have identified that increasing milk yield at dry-off can be associated with higher odds of new IMI during the dry period (
). A recent observational study of cows (n = 1,514) that did not receive DCT as part of an Alg-SDCT program in 36 New Zealand herds found that higher producing cows (>15 kg/d at the last herd test) had substantially higher odds of developing clinical mastitis during 1–60 DIM of the subsequent lactation (odds ratio = 4.8) when compared with lower producing cows (<10 kg/d;
Risk factors for clinical or subclinical mastitis following infusion of internal teat sealant alone at the end of lactation in cows with low somatic cell counts.
). Furthermore, the odds of subclinical mastitis increased by 7% for every 1 kg/d increase in milk yield at the last test (odds ratio = 1.07). In that study, the authors concluded that producers should consider cow milk yield when allocating treatments within an Alg-SDCT program, but also acknowledged that their study did not provide sufficient evidence to support the hypothesis that antibiotic DCT would mitigate yield-associated risks to udder health, as no treated comparison group was enrolled in their study. Given this context, we conducted a study using data from previously published, positively controlled SDCT trials (
Randomized controlled non-inferiority trial investigating the effect of 2 selective dry-cow therapy protocols on antibiotic use at dry-off and dry period intramammary infection dynamics.
Randomized controlled trial investigating the effect of 2 selective dry-cow therapy protocols on udder health and performance in the subsequent lactation.
) to investigate the hypothesis that SDCT may pose greater risks to udder health in high producing cows than in lower producing cows. The objective of this study was to evaluate if the effect of culture- and Alg-SDCT on dry-period IMI dynamics and postcalving udder health and performance (when compared with BDCT) varied according to milk production level before dry-off.
MATERIALS AND METHODS
Herd and Cow Enrollment in the Original Studies
Data from 2 positively controlled SDCT field trials conducted in the United States (
Randomized controlled non-inferiority trial investigating the effect of 2 selective dry-cow therapy protocols on antibiotic use at dry-off and dry period intramammary infection dynamics.
Randomized controlled trial investigating the effect of 2 selective dry-cow therapy protocols on udder health and performance in the subsequent lactation.
) were pooled into a single data set. Animal ethics were granted for both studies. The analyses reported in this study were not planned in the original research projects. No sample size calculations were conducted before conducting the current analysis.
Randomized controlled non-inferiority trial investigating the effect of 2 selective dry-cow therapy protocols on antibiotic use at dry-off and dry period intramammary infection dynamics.
Randomized controlled trial investigating the effect of 2 selective dry-cow therapy protocols on udder health and performance in the subsequent lactation.
), 1,275 cows from 7 dairy herds from California, Iowa, Minnesota, New York, and Wisconsin (bulk tank SCC ranging from 90,000 to 230,000 cells/mL; herd size ranging from 850 to 5,700 cows) were block randomized to BDCT, Cult-SDCT, or Alg-SDCT during summer 2018. Herd selection criteria included an existing relationship with the University of Minnesota, Cornell University, or DairyExperts Inc., maintaining an average bulk milk SCC <250,000 cells/mL during the past 12 mo, drying off at least 15 cows per week, being on a monthly DHIA testing schedule, and having high quality clinical mastitis and culling records. At dry-off, BDCT cows received an intramammary antibiotic (500 mg of ceftiofur hydrochloride, Spectramast DC, Zoetis) in all 4 quarters. Antibiotic treatments were selectively allocated to quarters of Cult-SDCT cows by treating only quarters from which aseptically collected milk samples tested positive on the Minnesota Easy 4Cast plate (University of Minnesota) after 30 to 40 h of incubation. For Alg-SDCT cows, antibiotic treatments were selectively allocated at the cow level, with all quarters receiving antibiotic treatment if the cow had either a DHIA test SCC >200,000 cells/mL during the current lactation or 2 or more clinical mastitis cases during the current lactation. All quarters of all cows were treated with an internal teat sealant (ITS). Quarter-level antibiotic use was 55% lower in both SDCT groups (culture and algorithm-guided), when compared with BDCT, and no differences were observed in dry-period IMI dynamics and postcalving udder health.
), 569 cows from 9 dairy herds in Quebec (herd size ranging from 26 to 165 cows) were randomized to 4 different DCT strategies between July 2015 and May 2016. Herd selection criteria included an average bulk tank SCC <250,000 cells/mL and being on a monthly DHIA testing schedule. The 4 DCT strategies evaluated were BDCT with ITS, BDCT without ITS, Cult-SDCT with blanket ITS, and Cult-SDCT with ITS only used in quarters that did not receive DCT. Of the 4 treatment groups, the BDCT with ITS (n = 148) and Cult-SDCT with blanket ITS group (n = 126) matched the BDCT and Cult-SDCT treatment groups within the US study, respectively, and were consequently eligible to be included in the current data set. Cows in the eligible BDCT group received an intramammary antibiotic (200,000 IU of Penicillin G Procaine and 400 mg of Novobiocin, Novodry Plus, Zoetis Canada, Kirkland) in all 4 quarters at dry-off. The same antibiotic treatment was selectively allocated to quarters of cows randomized to the Cult-SDCT group included in the current study, by treating only quarters from which aseptically collected milk samples tested positive on the Petrifilm Aerobic count plate (3M Petrifilm; 3M) after 24 h of incubation. No Alg-SDCT cows were enrolled in the Canadian study. Quarter-level antibiotic use was 57.9% lower in the Cult-SDCT groups (Cult-SDCT with blanket ITS and Cult-SDCT with ITS only used in quarters that did not receive DCT), when compared with BDCT, and no differences were observed in dry-period IMI dynamics and postcalving udder health.
Treatment Groups
Three DCT approaches were compared in this study: BDCT, Cult-SDCT, and Alg-SDCT. Cows in the BDCT (positive control, n = 555 cows) group received an antibiotic treatment and ITS in all quarters. Cows in the Cult-SDCT group received intramammary antibiotic treatments in quarters that tested positive using the Minnesota Easy 4Cast plate (US study, n = 410 cows) or 3M Petrifilm (n = 126 cows). Cows in the Alg-SDCT group (n = 394) were treated with antibiotics in all quarters if they failed to meet either of the follow criteria: (1) test-day SCC <200,000 cells/mL for all tests in the current lactation, and (2) less than 2 cases of clinical mastitis during the current lactation.
Sampling at Enrollment and Follow-Up During the Dry Period and Postcalving
Aseptic quarter-milk samples were collected 1 d (Canada) or 2 d (the United States) before dry-off using a standardized sampling technique outlined in
and cultured to establish IMI status at dry-off (laboratory methods outlined later). Postcalving quarter-milk samples were collected on 1 (the United States) or 2 (Canada) occasions in the first 18 DIM. Given that the US sample was collected between 1 and 13 DIM, the culture results for Canadian cows were retrospectively consolidated into a single data-point by excluding the later sampling if the first sampling yielded a valid result (i.e., noncontaminated). If no valid results from 1 to 13 DIM were available, then the cow was excluded from the current study. Consequently, all quarters in the final data set had an IMI status from a single milk sample collected at 1–13 DIM. Cows were monitored for visual signs of illness during the dry period as part of the farm's regular husbandry practices. After calving, the udder and foremilk of cows were examined at milking time by milking staff for evidence of clinical mastitis (abnormal milk, inflammatory signs in the udder, and systemic inflammatory signs). Somatic cell count and milk yield data were measured at monthly intervals as part of the herd's regular DHIA schedule and, along with health event data, were extracted from the electronic herd records.
Microbiological Culture of Milk Samples
Laboratory-based aerobic milk culture was conducted to identify the cause of IMI at dry-off and postcalving, so that IMI dynamics could be compared among treatment groups. These results were not used to guide antibiotic treatments at dry-off. Milk culture methods at the US and Canadian laboratories followed a similar procedure. For samples in the Canadian study, samples were thawed at room temperature, and inoculated onto Columbia and MacConkey agar using disposable plastic loops (0.01 mL). Plates were incubated at 35 ± 2°C for up to 48 h. In the US study, samples were thawed at room temperature, and inoculated onto Trypticase Soy Agar with 5% sheep blood using disposable loops (0.01 mL). Plates were incubated at 37 ± 2°C for 42–48 h. In both studies, samples were classified as contaminated if 3 or more distinct morphotypes were identified. Isolates were identified in both studies using a MALDI-TOF mass spectrometer (Microflex; Bruker Daltonics Inc.). Peaks produced by each isolate were analyzed by the MALDI-TOF Biotyper reference library. In the US study, the confidence level for each diagnosis reported by the software was used in the following fashion: >2.0, species-level diagnosis recorded; 1.8–2, genus-level diagnosis recorded; <1.8, MALDI-TOF diagnosis not recorded and traditional identification methods used. In the Canadian study, the confidence level was used in a similar fashion except that the lower-limit for genus-level diagnosis was 1.7, instead of 1.8. Consequently, the Canadian data set had a small number of isolates (n = 88) that were reported at the genus level that were between 1.7 and 1.8. Traditional identification methods included colony morphology, catalase reaction, gram stain, and cytology. For all milk culture results, non-aureus Staphylococcus spp. isolates with less than 2 colonies (<200 cfu/mL) and Bacillus spp. isolates with less than 5 colonies (<500 cfu/mL) were disregarded, and the quarter was considered uninfected (
Effect of SDCT on Quarter-Level Dry-Period IMI Dynamics
The analysis log for this study can be found at https://rpubs.com/samrowe/sdctxmilkyield. Intramammary infection status at dry-off and postcalving, and dry-period IMI dynamics, were calculated after the data set was assembled so that a consistent set of case definitions could be used across all quarters. Intramammary infection status at enrollment and postcalving was used to determine a dry-period IMI cure and new IMI. Consequently, quarters missing an IMI status at one or both periods were not assigned a value for these outcome variables. Only quarters with an IMI at the enrollment sample were considered at risk for a dry-period IMI cure. Dry-period IMI cure cases were defined as a quarter with a species-level IMI present at enrollment that was not isolated in the postcalving sample. Dry-period new IMI cases were defined as a quarter with a species-level IMI at calving that was not originally present in the enrollment sample. If species-level diagnosis was not available for an isolate, then it was matched at the genus level. For example, a quarter with a dry-off IMI caused by Staphylococcus chromogenes and postcalving IMI caused by Staphylococcus spp. would be classified as cure = 0. Multivariable generalized estimating equations (GEE) in the ‘geepack' package (
) were used to determine the effect of the DCT approach (i.e., Cult-SDCT vs. BDCT and Alg-SDCT vs. BDCT) on IMI prevalence at 1–13 DIM, and dry-period IMI cure risk and new IMI risk (family = binomial, link = identity). Consequently, the effect measure derived from these models was a population-averaged risk difference (RD). Models included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States) as they were hypothesized to potentially cause confounding. Standard error estimates were adjusted for the clustering of quarters within cows and cows within herds by specifying clustering at the highest level (i.e., herd) and using an independence working covariance structure (
). Generalized linear mixed models and linear mixed models were also conducted as a sensitivity analysis which yielded similar results (see analysis log).
Effect of SDCT on Cow-Level Postcalving Clinical Mastitis and Culling or Death
Survival analysis was conducted using the ‘survival' package in R (
) to evaluate the associations between DCT approach and cow-level clinical mastitis and removal from the herd during the first 120 d of the subsequent lactation. Clinical mastitis cases were defined as the first case of clinical mastitis (determined using herd records) from calving until 120 DIM. Cows were right-censored at 120 DIM or at time of removal from the herd. Removal from the herd events were defined as any record of death or culling in farm records. Kaplan-Meier survival curves were generated. Hazard ratios were estimated using Cox proportional hazards regression, using a sandwich variance estimator to account for the clustering of cows within herds. Multivariable models included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States), as they were hypothesized to potentially cause confounding. The proportional hazards assumption for each covariate was assessed by visualizing Schoenfeld residuals against time and using the Schoenfeld test. Covariates with hazards shown to vary over time (β ≠ 0, P < 0.05) were stratified to allow for multiple baseline hazards.
Effect of SDCT on Postcalving SCC and Milk Yield
Test-day SCC and milk yield data were analyzed in a longitudinal data set, with multiple observations (herd tests) per cow during 1–120 DIM. Somatic cell counts were loge transformed. Generalized estimating equations were used to determine the association between DCT approach and monthly test-day milk yield and logeSCC (family = Gaussian, link = identity) during 1–120 DIM. Standard error estimates were adjusted for the clustering of tests within cows using an independence working covariance structure. Models included the following covariates: parity, DIM at herd test, SCC before dry-off, milk yield at last herd test, and farm identification. Consequently, the β coefficient for each model provides an estimate for the adjusted, population-averaged difference in SCC or milk yield for cows dried off using different dry-off approaches. Assumptions of normality of residuals and homoscedasticity were evaluated.
Strategies Used to Evaluate if Milk Yield at Dry-Off Modified the Effect of SDCT on Health and Production Outcomes
The major objective of this study was to investigate if the effect of Cult-SDCT or Alg-SDCT varied according to milk production level before dry-off. Milk production (kilograms of liquid milk per day) before dry-off was determined using the most recent herd test before enrollment. We chose to not incorporate milk fat or protein concentrations into our definition of milk yield (e.g., kilograms of solids or FCM yield), as total milk volume is hypothesized to increase dry-period IMI risk from delayed teat end closure due to an increased intramammary pressure (
), which is unlikely to be affected by density of milk solids in the gland. For cows in the Canadian study, the most recent herd test was 1 to 65 d before dry-off (median = 22, interquartile range = 13 to 29 d). Data were not available to describe the interval between the most recent herd test and dry-off for cows in the US study. However, all herds recruited into the US study conducted DHIA testing on a monthly basis. Given that interpreting interactions with continuous predictors can be difficult, we decided to categorize milk yield into tertiles (i.e., 3 equal-sized groups in terms of cow numbers across all herds): low (<23.7 kg/d), mid (23.7 to 30.4 kg/d), and high (>30.4 kg/d). As a sensitivity analysis, we also classified milk yield into tertiles based on relative production within herd, thus creating 3 equal-sized production-level groups within each herd. We decided to only report the former approach (tertile groups that ignored herd-level clustering), as results were similar to the within-herd tertile group approach (see analysis log). Effect modification was investigated using 4 main strategies: (1) comparison of crude risks for disease outcomes for treatment groups within milk yield strata, (2) comparison of stratum-specific Kaplan-Meier failure curves, (3) evaluating P-values for interaction terms [pre-dry milk yield × DCT approach (Cult-SDCT vs. BDCT, in one analysis; and Alg-SDCT vs. BDCT, in another analysis), within interaction models], and (4) deriving pre-dry milk yield stratum-specific effect estimates and 95% confidence intervals (CI; Cult-SDCT vs. BDCT and Alg-SDCT vs. BDCT) from multivariable interaction models. For outcomes evaluated with GEE models (outlined earlier), we used alternative models (linear mixed models or generalized linear models) to calculate the P-value for interaction terms, as likelihood-based measures of model fit are not recommended for GEE. P-values were determined using the ‘Anova' function within the ‘car' package (
), which uses likelihood ratio, partial-likelihood ratio, and Wald chi-square tests for generalized linear regression, Cox proportional hazards, and linear mixed models, respectively. Regardless of P-value for the interaction term, we explored stratum-specific estimates for all outcomes by passing the GEE or Cox Proportional Hazards interaction model through the ‘contrast' function within the ‘Emmeans' package (
Randomized controlled non-inferiority trial investigating the effect of 2 selective dry-cow therapy protocols on antibiotic use at dry-off and dry period intramammary infection dynamics.
. A summary of key demographic details for cows in the final data set is shown in Table 1, which shows that treatment groups were balanced at the baseline, except that, by study design, no Canadian cows were randomized to Alg-SDCT. The prevalence of IMI at dry-off was 25.3% (1,187/4,688 quarters). The most commonly isolated pathogens at dry-off were Staphylococcus chromogenes (7.0%), unspeciated Staphylococcus (4.3%), unspeciated Corynebacterium (3.8%), and Staphylococcus xylosus (1.3%). For more information about the pathogen profile at dry-off, we suggest that readers refer to the original clinical trials and the analysis log for this study. At dry-off, the proportion of quarters receiving DCT within each treatment group were BDCT (1,893/1,893, 100%), Cult-SDCT (800/1,849, 43.3%), and Alg-SDCT (596/1,355, 44.0%). For Cult-SDCT, the proportion of quarters treated within each milk production level were 40.7% (low), 41.7% (mid), and 47.2% (high). For Alg-SDCT the proportions were 60.6% (low), 38.7% (mid), and 35.1% (high).
Table 1Cow status at dry-off for US and Canadian cows at the time of randomization to receive Culture-SDCT, Algorithm-SDCT, or Blanket DCT
Of 5,097 quarters at risk of IMI at the postcalving milk sampling, 1,120 (22.0%) were infected. The most commonly isolated pathogens were Staphylococcus sp. (4.3%), Staphylococcus chromogenes (3.9%), Staphylococcus sciuri (2.7%), unspeciated Bacillus (2.6%), and unspeciated Corynebacterium (2.6%; see previously published studies and analysis log for more information). Crude risk and model-adjusted RD estimates for early lactation IMI are shown in Table 2. When considering all cows, the RD estimates for Cult-SDCT and Alg-SDCT were +0.02 (95% CI: −0.02 to 0.07) and 0.00 (−0.03 to 0.04), respectively, as compared with BDCT. The P-values for the interaction terms with pre-dry milk yield were 0.33 (Cult-SDCT × milk yield) and 0.85 (Alg-SDCT × milk yield). Early lactation IMI RD estimates within milk production strata for Cult-SDCT were +0.05 (low), +0.00 (mid), and +0.01 (high), and for Alg-SDCT were +0.00 (low), +0.01 (mid), and −0.01 (high), when compared with BDCT. Therefore, we did not find evidence to suggest that the effect of SDCT on early lactation IMI prevalence varies as a function of pre-dry milk yield.
Table 2Quarter-level prevalence of IMI at 1–13 DIM for cows randomized to 1 of 3 DCT approaches, stratified by milk yield level (tertile) at the last test before dry-off
Risk difference estimates are derived from multivariable generalized estimating equation models (binomial family, identity link), which included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States).
Tertiles of milk production were low: <23.7 kg/d, mid: 23.7 to 30.4 kg/d, and high >30.4 kg/d.
Blanket DCT
138/706 (19.55%)
Referent
Cult-SDCT
142/582 (24.4%)
0.05 (−0.02 to 0.12)
Algorithm-guided
85/411 (20.68%)
0 (−0.04 to 0.03)
Mid production
Blanket DCT
124/621 (19.97%)
Referent
Cult-SDCT
132/636 (20.75%)
0 (−0.05 to 0.05)
Algorithm-guided
102/442 (23.08%)
0.01 (−0.04 to 0.07)
High production
Blanket DCT
131/566 (23.14%)
Referent
Cult-SDCT
150/631 (23.77%)
0.01 (−0.02 to 0.05)
Algorithm-guided
116/502 (23.11%)
−0.01 (−0.06 to 0.05)
1 Risk difference estimates are derived from multivariable generalized estimating equation models (binomial family, identity link), which included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States).
Of 1,187 quarters that were infected at dry-off, 1,045 (88.0%) cured over the course of the dry period. Cure crude risk and model-adjusted RD estimates are shown in Table 3. When considering all cows, the RD estimates for Cult-SDCT and Alg-SDCT compared with BDCT were +0.02 (95% CI: −0.02 to 0.05) and +0.03 (0.00 to 0.05), respectively. The P-values for the interaction terms between pre-dry milk yield and DCT approach were 0.26 (Cult-SDCT × milk yield) and 0.81 (Alg-SDCT × milk yield). Cure RD estimates, when compared with BDCT and within milk production strata for Cult-SDCT were +0.03 (low), −0.03 (mid), and +0.05 (high), and for Alg-SDCT were +0.05 (low), +0.01 (mid), and +0.04 (high). Therefore, we did not find evidence to suggest that the effect of SDCT on dry-period IMI cure risk varies as a function of pre-dry milk yield.
Table 3Quarter-level risk of IMI cure during the dry period for cows randomized to 1 of 3 DCT approaches, stratified by milk yield level (tertile) at the last test before dry-off
Risk difference estimates are derived from multivariable generalized estimating equation models (binomial family, identity link), which included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States).
Tertiles of milk production were low: <23.7 kg/d, mid: 23.7 to 30.4 kg/d, and high >30.4 kg/d.
Blanket DCT
139/161 (86.34%)
Referent
Cult-SDCT
99/112 (88.39%)
0.03 (−0.02 to 0.08)
Alg-SDCT
60/68 (88.24%)
0.05 (−0.02 to 0.11)
Mid production
Blanket DCT
124/137 (90.51%)
Referent
Cult-SDCT
131/151 (86.75%)
−0.03 (−0.09 to 0.03)
Alg-SDCT
90/103 (87.38%)
0.01 (−0.02 to 0.03)
High production
Blanket DCT
130/152 (85.53%)
Referent
Cult-SDCT
155/171 (90.64%)
0.05 (0.02 to 0.09)
Alg-SDCT
117/132 (88.64%)
0.04 (−0.02 to 0.09)
1 Risk difference estimates are derived from multivariable generalized estimating equation models (binomial family, identity link), which included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States).
Of 4,638 quarters at risk, 965 (20.8%) developed a new IMI during the dry period. New IMI crude risk and model-adjusted RD estimates are shown in Table 4. When considering all cows, the RD estimates for Cult-SDCT and Alg-SDCT, compared with BDCT, were +0.02 (95% CI: −0.02 to 0.07) and 0.00 (−0.03 to 0.04), respectively. The P-values for the pre-dry milk yield per DCT approach interaction terms were 0.22 (Cult-SDCT × milk yield) and 0.92 (Alg-SDCT × milk yield). New IMI RD estimates when compared with BDCT and within milk production strata for Cult-SDCT were +0.05 (low), 0.00 (mid), and +0.03 (high), and for Alg-SDCT were +0.01 (low), 0.00 (mid), and 0.00 (high). Therefore, we did not find evidence to suggest that the effect of SDCT on dry-period new IMI risk varies as a function of pre-dry milk yield.
Table 4Quarter-level risk of new IMI during the dry period for cows randomized to 1 of 3 DCT approaches, stratified by milk yield level (tertile) at the last test before dry-off
Risk difference estimates are derived from multivariable generalized estimating equation models (binomial family, identity link), which included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States).
Tertiles of milk production were low: <23.7 kg/d, mid: 23.7 to 30.4 kg/d, and high >30.4 kg/d.
Blanket DCT
123/653 (18.84%)
Referent
Cult-SDCT
130/546 (23.81%)
0.05 (−0.03 to 0.13)
Alg-SDCT
80/375 (21.33%)
0.01 (−0.02 to 0.04)
Mid production
Blanket DCT
106/569 (18.63%)
Referent
Cult-SDCT
109/581 (18.76%)
0 (−0.05 to 0.04)
Alg-SDCT
83/401 (20.7%)
0 (−0.04 to 0.05)
High production
Blanket DCT
106/499 (21.24%)
Referent
Cult-SDCT
131/566 (23.14%)
0.03 (−0.02 to 0.07)
Alg-SDCT
97/448 (21.65%)
0 (−0.07 to 0.06)
1 Risk difference estimates are derived from multivariable generalized estimating equation models (binomial family, identity link), which included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States).
Effect of SDCT on Risk for Clinical Mastitis and Risk for Removal from the Herd During the First 120 DIM
Of the 1,485 cows at risk for clinical mastitis during 1–120 DIM, 203 (13.7%) developed at least one case. Kaplan-Meier survival analysis estimated the cumulative incidence of clinical mastitis at 30, 60, 90, and 120 d to be 3.8, 7.6, 10.8, and 14.3%, respectively. Failure curves (1 − survival) showing the cumulative incidence of clinical mastitis for each DCT approach among low, mid, and high producing cows are shown in Figure 1. Crude risks and hazard ratio (HR) estimates for clinical mastitis during 1–120 DIM for cows of low, mid, high, and all production levels are shown in Table 5. When considering all cows, the HR for clinical mastitis for Cult-SDCT and Alg-SDCT were 0.73 (95% CI: 0.52 to 1.02) and 0.80 (95% CI: 0.6 to 1.06), respectively, when compared with BDCT. The P-values for the pre-dry milk yield per DCT approach interaction terms were 0.31 (Cult-SDCT × milk yield) and 0.45 (Alg-SDCT × milk yield). Hazard ratio estimates within milk production strata, and as compared with BDCT, for Cult-SDCT were 0.52 (low), 1.02 (mid), and 0.72 (high) and for Alg-SDCT, the HR estimates were 0.99 (low), 0.87 (mid), and 0.62 (high). Therefore, we did not find evidence to suggest that the effect of SDCT on postcalving clinical mastitis varies as a function of pre-dry milk yield.
Figure 1Kaplan-Meier curves showing the cumulative incidence (1 − survival) for clinical mastitis over the first 120 d of lactation for cows randomized to blanket dry-cow therapy (DCT; red line), cultured-guided selective dry-cow therapy (SDCT; blue line), and algorithm-guided SDCT (green line), stratified by milk yield level (tertile) at the last test before dry-off.
Table 5Cow-level incidence of clinical mastitis during 1 to 120 DIM for cows randomized to 1 of 3 DCT approaches, stratified by milk yield level (tertile) at the last test before dry-off
Hazard ratio estimates are derived from multivariable Cox proportional hazards models, which included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States).
Tertiles of milk production were low: <23.7 kg/d, mid: 23.7 to 30.4 kg/d, and high >30.4 kg/d.
Blanket DCT
37/203 (18.23%)
Referent
Cult-SDCT
17/171 (9.94%)
0.52 (0.22 to 1.23)
Alg-SDCT
19/121 (15.7%)
0.99 (0.59 to 1.66)
Mid production
Blanket DCT
25/186 (13.44%)
Referent
Cult-SDCT
25/182 (13.74%)
1.02 (0.69 to 1.5)
Alg-SDCT
14/127 (11.02%)
0.87 (0.54 to 1.39)
High production
Blanket DCT
29/166 (17.47%)
Referent
Cult-SDCT
22/183 (12.02%)
0.72 (0.39 to 1.31)
Alg-SDCT
15/146 (10.27%)
0.62 (0.37 to 1.03)
1 Hazard ratio estimates are derived from multivariable Cox proportional hazards models, which included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States).
Of the 1,485 cows at risk, 148 (10.0%) were culled or died during 1–120 DIM. Kaplan-Meier survival analysis estimated the cumulative incidence of removal from the herd at 30, 60, 90, and 120 d to be 3.8, 6.1, 8.0, and 10.0%, respectively. Failure curves showing the cumulative incidence of removal for each DCT approach among low, mid, and high producing cows are shown in Figure 2. Crude risks and hazard ratio estimates for removal from the herd during 1–120 DIM for cows of low, mid, high, and all production levels are shown in Table 6. When considering all cows, the HR for removal from the herd for Cult-SDCT and Alg-SDCT, when compared with BDCT, were 1.02 (0.72 to 1.43) and 1.06 (0.70 to 1.62), respectively. The P-values for the pre-dry milk yield per DCT approach interaction terms were 0.92 (Cult-SDCT × milk yield) and 0.53 (Alg-SDCT × milk yield). Hazard ratio estimates within milk production strata, when compared with BDCT for Cult-SDCT, were 1.02 (low), 0.92 (mid), and 1.12 (high) and for Alg-SDCT, the HR estimates were 1.28 (low), 0.80 (mid), and 1.06 (high). Therefore, we did not find evidence to suggest that the effect of SDCT on removal from the herd varies as a function of pre-dry milk yield.
Figure 2Kaplan-Meier curves showing the cumulative incidence (1 − survival) for culling during the first 120 d of lactation for cows randomized to blanket dry-cow therapy (DCT; red line), cultured-guided selective dry-cow therapy (SDCT; blue line), and algorithm-guided SDCT (green line), stratified by milk yield level (tertile) at the last test before dry-off.
Table 6Incidence of culling or death during 1 to 120 DIM for cows randomized to 1 of 3 DCT approaches, stratified by milk yield level (tertile) at the last test before dry-off
Hazard ratio estimates are derived from multivariable Cox proportional hazards models, which included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States).
Tertiles of milk production were low: <23.7 kg/d, mid: 23.7 to 30.4 kg/d, and high >30.4 kg/d.
Blanket DCT
24/203 (11.82%)
Referent
Cult-SDCT
21/171 (12.28%)
1.02 (0.5 to 2.1)
Alg-SDCT
20/121 (16.53%)
1.28 (0.82 to 1.99)
Mid production
Blanket DCT
17/186 (9.14%)
Referent
Cult-SDCT
16/182 (8.79%)
0.92 (0.49 to 1.73)
Alg-SDCT
10/127 (7.87%)
0.8 (0.44 to 1.46)
High production
Blanket DCT
13/166 (7.83%)
Referent
Cult-SDCT
15/183 (8.2%)
1.12 (0.45 to 2.79)
Alg-SDCT
12/146 (8.22%)
1.06 (0.29 to 3.79)
1 Hazard ratio estimates are derived from multivariable Cox proportional hazards models, which included the following covariates: parity, SCC before dry-off, and study location (Canada or the United States).
Effect of SDCT on Test-Day Milk Yield and SCC During 1–120 DIM
The estimated differences in logeSCC during 1–120 DIM for Cult-SDCT minus BDCT and Alg-SDCT minus BDCT were +0.03 (95% CI: −0.10 to 0.15) and +0.03 loge cells/mL (95% CI: −0.11 to 0.16), respectively (Table 7 and Figure 3). The P-values for the pre-dry milk yield per DCT approach interaction terms were 0.56 (Cult-SDCT × milk yield) and 0.56 (Alg-SDCT × milk yield). Estimated differences in logeSCC within milk production strata for Cult-SDCT were −0.01 (low), −0.05 (mid), and +0.14 (high) and for Alg-SDCT, the estimated differences were +0.15 (low), −0.09 (mid), and +0.04 (high) loge cells/mL. Therefore, we did not find evidence to suggest that the effect of SDCT on postcalving SCC varies as a function of pre-dry milk yield.
Table 7Differences in loge SCC during 1 to 120 DIM, for cows randomized to 1 of 3 DCT approaches, stratified by production level (tertile)
Differences in average loge SCC estimates are derived from multivariable generalized estimating equation models (Gaussian family, identity link), which included the following covariates: parity, DIM at herd test, SCC before dry-off, and farm ID.
Tertiles of milk production were low: <23.7 kg/d, mid: 23.7 to 30.4 kg/d, and high >30.4 kg/d.
Blanket DCT
Referent
Cult-SDCT
−0.01 (−0.22 to 0.21)
Alg-SDCT
0.15 (−0.09 to 0.39)
Mid production
Blanket DCT
Referent
Cult-SDCT
−0.05 (−0.26 to 0.16)
Alg-SDCT
−0.09 (−0.31 to 0.14)
High production
Blanket DCT
Referent
Cult-SDCT
0.14 (−0.08 to 0.36)
Alg-SDCT
0.04 (−0.19 to 0.26)
1 Differences in average loge SCC estimates are derived from multivariable generalized estimating equation models (Gaussian family, identity link), which included the following covariates: parity, DIM at herd test, SCC before dry-off, and farm ID.
Figure 3Estimated milk SCC concentration (cells/mL) in the subsequent lactation and as a function of DIM for cows randomized to blanket dry-cow therapy (DCT; red line), cultured-guided selective dry-cow therapy (SDCT; blue line), and algorithm-guided SDCT (green line), stratified by milk yield level (tertile) at the last test before dry-off.
The estimated differences in milk yield during 1–120 DIM in the subsequent lactation for Cult-SDCT minus BDCT and Alg-SDCT minus BDCT were −0.18 (95% CI: −0.99 to 0.62) and −0.98 kg/d (95% CI: −1.87 to −0.09), respectively (Table 8 and Figure 4). The P-values for the pre-dry milk yield per DCT approach interaction terms were 0.56 (Cult-SDCT × milk yield) and 0.57 (Alg-SDCT × milk yield). Estimated differences in subsequent lactation 1–120 DIM daily milk yield (kg/d) within pre-dry milk production strata for Cult-SDCT were −0.09 (low), +0.46 (mid), and −0.93 (high) and for Alg-SDCT, the estimated differences were −1.70 (low), +0.20 (mid), and −1.54 (high) kg/d, when compared with BDCT. Therefore, we did not find evidence to suggest that the effect of SDCT on postcalving milk yield varies as a function of pre-dry milk yield.
Table 8Differences in average daily milk production during 1 to 120 DIM, for cows randomized to 1 of 3 DCT approaches, stratified by production level (tertile)
Differences in average loge SCC estimates are derived from multivariable generalized estimating equation models (Gaussian family, identity link), which included the following covariates: parity, DIM at herd test, SCC before dry-off, milk yield before dry-off, and farm ID.
Tertiles of milk production were low: <23.7 kg/d, mid: 23.7 to 30.4 kg/d, and high >30.4 kg/d.
Blanket DCT
Referent
Cult-SDCT
−0.09 (−1.45 to 1.28)
Alg-SDCT
−1.7 (−3.39 to 0.00)
Mid production
Blanket DCT
Referent
Cult-SDCT
0.46 (−0.82 to 1.75)
Alg-SDCT
0.2 (−1.11 to 1.5)
High production
Blanket DCT
Referent
Cult-SDCT
−0.93 (−2.45 to 0.59)
Alg-SDCT
−1.54 (−3.11 to 0.04)
1 Differences in average loge SCC estimates are derived from multivariable generalized estimating equation models (Gaussian family, identity link), which included the following covariates: parity, DIM at herd test, SCC before dry-off, milk yield before dry-off, and farm ID.
Figure 4Estimated daily milk yield (kg/d) in the subsequent lactation and as a function of DIM for cows randomized to blanket dry-cow therapy (DCT; red line), cultured-guided selective dry-cow therapy (SDCT; blue line), and algorithm-guided SDCT (green line), stratified by milk yield level (tertile) at the last test before dry-off.
Findings from this study indicate that quarter-level culture- and cow-level Alg-SDCT can be successfully performed in cows of all production levels. We found that crude risks for new IMI, IMI prevalence at calving, IMI cure, and clinical mastitis and removal from the herd during 1–120 DIM were similar between all treatment groups within pre-dry milk yield strata, and effect estimates derived from multivariable models (i.e., RD and HR) were close to the null (0 and 1, respectively). This was also found for postcalving logeSCC and milk yield. The homogeneity of the effect of SDCT within milk yield strata for the aforementioned measures of health and productivity is further supported by the finding that P-values for interaction terms (Cult-SDCT × milk yield, Alg-SDCT × milk yield) were >0.05 in all models. Finally, the stratified survival analysis curves illustrating time to clinical mastitis and time to removal from the herd also failed to find evidence to support the hypothesis that the safety of SDCT may vary according to milk production level.
To our knowledge, this study is the first to use data from clinical trials to investigate the potential role of milk yield at dry-off as an effect modifier for SDCT efficacy. However, a recent observational study of cows (n = 1,514) that did not receive DCT as part of an Alg-SDCT program in 36 New Zealand herds found that higher producing cows (>15 kg/d at the last herd test) had substantially higher odds of developing clinical mastitis during 1–60 DIM of the subsequent lactation (odds ratio = 4.8) when compared with lower producing cows (<10 kg/d). However, no cows received BDCT in that study. Consequently, it was not possible to determine if administration of BDCT would have mitigated these milk yield-associated udder health risks. Observational studies have found that this association can also exist in cows that receive BDCT. For example, increasing milk yield before dry-off was associated with greater new IMI risk during the dry-period (
). However, it should be noted that in those studies, no untreated cows were included for comparison, so it is not possible to identify if outcomes would have been worse if cows had not been received BDCT.
In light of our results and the existing research in this area, we conclude that treatment with an antibiotic of high producing cows that are identified as healthy (i.e., culture negative or algorithm negative) may be an inefficient and uneconomical use of antibiotics. In our opinion, which is based on our experience in DCT research and extension, the following recommendations are effective steps that can be taken to reduce SDCT-associated health risks for all cows, regardless of level of milk production: (1) use of a validated SDCT screening approach, (2) use a teat sealant in all cows, (3) establish standard operating procedures for dry-off and ensure that staff receive regular training and supervision, and (4) practice excellent hygiene when performing intramammary infusions.
Furthermore, we also wish to remind readers of general management strategies recommended by the National Mastitis Council to reduce IMI risk during the dry period, which include reducing milk yield before dry-off by reducing dietary energy density in late lactation, use of teat disinfectants after the dry-off procedure, providing adequate nutrition during the dry period, and maintenance of a clean, dry, and comfortable environment during the dry period.
Study Strengths and Limitations of This Study
Internal Validity
We encourage readers to consider the strengths and limitations of this study before generalizing our findings. One important strength of this study was the use of multiple measures of health and productivity, that have relevance to farmers (e.g., removal from the herd, clinical mastitis, SCC), but also the capacity to identify underlying mechanisms (e.g., new IMI risk, IMI cure risk). Furthermore, the large number of cows (n = 1,485) and quarters (n = 5,097) enrolled in this study enabled us to stratify our analysis by milk production group. However, it should be noted that for some outcomes (e.g., removal from the herd), the stratum-specific effect estimates were imprecise, as evidenced by wide 95% CI. This lack of precision may be due to the sample size used in this study. We hope that these analyses can be replicated from other clinical trials so that effect estimates can be incorporated into meta-analyses, which will ultimately lead to a more precise effect estimate. Confounding and selection were unlikely sources of bias in the original studies, as participants were randomized to their treatment group. However, it is possible that herd-level confounding may have occurred after combining data sets from the 2 studies, especially with Alg-SDCT, which was not performed in Canadian herds. This means that Alg-SDCT cows from US herds would have been compared with BDCT cows in both Canadian and US herds, which could lead to confounding. To combat this, we adjusted for herd and country in models. Furthermore, cows in Canadian and US herds were efficiently balanced by the randomization process for IMI prevalence, parity, and SCC at dry-off (see analysis log), indicating that these potential risk factors for postcalving udder health were unlikely to be associated with farm origin and, thus, unlikely to cause confounding. Other important limitations that should be mentioned are the risks of measurement error and consequential information bias, which can occur when imperfect diagnostic tests are used to determine outcome status. In this study, a laboratory-aerobic culture, which has been estimated to have sensitivity and specificity values of 70 and 90% respectively (
), was used to determine postcalving IMI prevalence, dry-period new IMI risk, and IMI cure risk. Furthermore, it should be noted that aerobic culture methods differed slightly between the US and Canadian laboratories, including the use of different agar plates (Columbia and Tryptic Soy agar). Given that both medias are commonly used nonselective medias, we think that this difference in laboratory methods is unlikely to cause significant bias in our study. Another important consideration is the large number of associations evaluated in this study (8 per outcome, total = 56). We did not use a correction method for multiple comparisons (
), which means that some findings of ‘statistical significance' may be due to random error within the data, rather than true causal relationships. For example, we found that dry-period cure risk for Alg-SDCT was slightly higher than BDCT when considering all production levels (RD = +0.03, 95% CI: 0.00 to 0.05) and that Cult-SDCT had slightly higher cure risk than BDCT within high production cows (RD = 0.05, 95% CI: 0.02 to 0.09), which raises the hypothesis that SDCT could potentially improve cure risk. Furthermore, we found that milk production during 1–120 DIM was less for Alg-SDCT cows, when compared with BDCT (−0.98 kg/d, 95% CI: −1.87 to −0.09). This finding raises the possibility that Alg-SDCT could impair udder health and subsequently reduce milk yield in the subsequent lactation. These hypotheses require further investigation, as they are at odds with other findings within our study.
External Validity
We believe that our research is relevant to a large proportion of North American dairy farms because cows were enrolled from 16 commercial herds located in multiple dairy farming regions of the United States and Canada. However, it should be noted that herds were selected by convenience, and enrollment criteria were used. These included an average bulk tank SCC <250,000 c/mL, on a monthly DHIA testing schedule, and having a relationship with the study investigators (US herds only). A more detailed description of the participating herds can be found in the original papers for the US (
Randomized controlled non-inferiority trial investigating the effect of 2 selective dry-cow therapy protocols on antibiotic use at dry-off and dry period intramammary infection dynamics.
) trials. Another important consideration is that study technicians performed the dry-off protocols (not farm staff) in the US study, including collection of milk samples, rapid culture setup and reading, and administration of intramammary treatments. It is, therefore, possible that outcomes could be different if these procedures were performed by farm workers. Farm workers performed the dry-off routine in the Canadian study.
Future Research
More studies are needed to identify if findings from this study can be replicated, particularly from herds in other management systems such as pasture-based dairy farms.
ACKNOWLEDGMENTS
Fieldwork conducted in the United States funded by the United States Department of Agriculture – NIFA (Award # 2018-67015-28298) and was supported by an in-kind donation of product (Spectramast DC, Orbeseal) from Zoetis (Parsippany, NJ). Fieldwork in Canada was supported by Agriculture and Agri-Food Canada (Ottawa, Ontario, Canada), and by additional contributions from Dairy Farmers of Canada (Ottawa, Ontario, Canada), the Canadian Dairy Network (Guelph, Ontario, Canada), and the Canadian Dairy Commission (Ottawa, Ontario, Canada) under the Agri-Science Clusters Initiative, through the Canadian Bovine Mastitis and Milk Quality Research Network (Saint Hyacinthe, Quebec, Canada) research program, by Zoetis (Kirkland, Quebec, Canada), and by one of the authors (Dufour) NSERC-Discovery grant funds (RGPIN/435637-2013, Saint Hyacinthe, Quebec, Canada). Fidele Kabera was also supported by the NSERC-CREATE in Milk Quality program and the Fonds de recherche du Québec-Nature et technologies (FRQNT). As per the research agreement, aside from providing financial support, the funders have no role in the design and conduct of the studies, data collection and analysis, or interpretation of the data. The authors wish to acknowledge the important contributions of co-authors from the US and Canadian studies: Pat Gorden (Iowa State University), Alfonso Lago (Dairy Experts Inc., California), Amy Vasquez (Danone, New York), Erin Royster (University of Minnesota), Jennifer Timmerman (University of Minnesota), Mark Thomas (Dairy Health and Management Services, New York), Greg Keefe (University of Prince Edward Island, Canada), and Marguerite Cameron (University of Prince Edward Island, Canada). The Minnesota Easy 4Cast plate is manufactured by the University of Minnesota (St. Paul, MN). However, the study investigators have no financial interest in the sale of this plate. The authors have not stated any other conflicts of interest.
REFERENCES
Berry E.A.
Hillerton J.E.
The effect of selective dry cow treatment on new intramammary infections.
Evaluation of selective dry cow treatment following on-farm culture: Risk of postcalving intramammary infection and clinical mastitis in the subsequent lactation.
Risk factors for clinical or subclinical mastitis following infusion of internal teat sealant alone at the end of lactation in cows with low somatic cell counts.
Randomized controlled non-inferiority trial investigating the effect of 2 selective dry-cow therapy protocols on antibiotic use at dry-off and dry period intramammary infection dynamics.
Randomized controlled trial investigating the effect of 2 selective dry-cow therapy protocols on udder health and performance in the subsequent lactation.
Postcalving udder health and productivity in cows approaching dry-off with intramammary infections caused by non-aureus Staphylococcus, Aerococcus, Enterococcus, Lactococcus, and Streptococcus species.
Monitoring udder health on routinely collected census data: Evaluating the short- to mid-term consequences of implementing selective dry cow treatment.
00013-9&id=gr1.jpg){kind=link}
00013-9&id=gr2.jpg){kind=link}
00013-9&id=gr3.jpg){kind=link}
00013-9&id=gr4.jpg){kind=link}