Materials and Methods

Animals—Holstein calves on 12 commercial dairy farms (herds 1 through 12, respectively) located in western Minnesota (n = 9 farms) and eastern Wisconsin (3 farms) were used in the study. Herd participation was entirely voluntary. Criteria for herd inclusion were producer willingness to comply with study protocols, participation in the Minnesota (or Wisconsin) Dairy Herd Improvement Association milk production records system, and evidence that JD was endemic in the herd. Endemic status was initially assigned on the basis of a history of culling cattle because of JD during the year preceding the start of the study, although confirmatory testing to determine herd infection status was subsequently performed. Before the start of the study, producers completed a questionnaire describing their routine colostrum management programs and general herd information. The study was approved (protocol No. 0704A06362) by the Institutional Animal Care and Use Committee of the University of Minnesota.

Study design—A prospective randomized clinical trial was conducted to evaluate the risk of developing MAP infection in heifer calves fed raw bovine MC (control group, n = 261 calves) or CR (treatment group, 236 calves) at birth. Heifer calves born between July and October 2003 were enrolled in the study. Collection of samples and data were completed for the study cohort in December 2007.

Each enrolled heifer calf was separated from its dam within 30 to 60 minutes after birth and was not allowed the opportunity to suckle its dam. By use of a systematic allocation procedure (ie, alternating such that a calf was assigned to one group and the subsequent calf was assigned to the other group), calves were assigned to be fed raw bovine MC (control group) or a single dose of CR^a (treatment group). The MC fed to a calf originated from its dam, although there were instances in which an insufficient quantity of MC was available from a particular dam, which necessitated that refrigerated colostrum from other dams was fed instead. The CR contained 125 g of IgG/dose and was mixed with 2 L of warm water, in accordance with the manufacturer's instructions, prior to feeding.

Calves in 11 of the 12 participating herds were fed 3.8 L of MC at the first feeding, whereas calves in the other herd were fed 1.9 L of MC during the first feeding. Calves in 5 herds were routinely provided a single meal of MC followed by a commercial milk replacer thereafter. In the other 7 herds, calves were routinely provided a second meal consisting of 1.9 L of MC within 8 to 12 hours after the first colostral feeding. In those herds, calves assigned to the treatment group were fed a second meal consisting of 1.9 L of a commercial milk replacer supplemented with a single dose of a commercially available colostrum-supplement product^b within 8 to 12 hours after the first CR meal was fed. In contrast to CR, colostrum-supplement products are intended to provide < 100 g of IgG/dose when fed to calves and are not formulated to replace MC. The colostrum-supplement product used in this study contained 45 g of IgG/dose, and the CR and colostrum-supplement products both originated from the same manufacturer. Preweaning diets (up to 56 days of age) for all calves consisted of commercial milk replacer and unrestricted amounts of water and calf starter pellets.

Preweaning samples and data collection—For each enrolled calf, calf managers recorded the identification of the dam, identification of the calf, birth date, number of calves born (singleton or twins), type of colostrum fed (CR or MC), calving score, and interval from birth until feeding of colostrum. Paired 20-mL samples of MC fed to each calf in the MC group were collected into sterile sample tubes that were subsequently labeled with the respective identification of the calf and farm identification prior to freezing at −20°C until subsequent analysis to determine contamination with MAP. The intent was to establish whether calves were exposed to MAP via MC.

The short-term objective of this study was to evaluate the effect of feeding CR versus MC on serum IgG and total protein concentrations in young (1- to 8-day-old) calves and morbidity and mortality risk for calves between birth and weaning. Findings for the short-term objective of this study have been reported elsewhere.⁸

Postweaning samples and data collection—Calves were weaned at approximately 56 days of age and monitored until adulthood. Calves were tested for MAP infection at approximately 30, 42, and 54 months of age. Data on calves were not officially collected by the investigators between weaning and the first testing event at 30 months of age. However, records of deaths, sale of animals, or culling events during this period were retrieved from commercially available dairy management software^c installed on participating farms. Alternately, some of the death and culling events were accessed from records maintained manually by the producers. Despite this effort, some heifers were lost from the study during this period. These heifers were censored at the last known date within the herd (weaning date). There were no losses to follow-up monitoring in the period between the first testing event (30 months of age) and conclusion of the study (54 months of age). Prior to the start of testing of the adult study cows for MAP infection, producers completed a questionnaire describing details of their routine JD management programs.

During each sampling event at approximately 30, 42, and 54 months of age, 10 mL of blood was collected from the coccygeal vein of each cow by use of a 20-gauge, 1-inch needle into a 16 × 100-mm blood collection tube. Approximately 10 g of feces was collected manually from the rectum of each cow by use of a disposable plastic examination sleeve and stored in sterile plastic sample containers prior to shipment to the Minnesota Veterinary Diagnostic Laboratory located at the University of Minnesota College of Veterinary Medicine for processing and testing.

Serologic testing for MAP antibodies—Serum derived from blood samples was tested for antibodies against MAP by use of a commercially available serologic ELISA kit.^d Seropositive results for MAP were indicated by use of an optical density value. For each sample, values for MAP antibodies were reported as an S/P. Samples with an S/P < 0.25 were classified as having negative results, whereas those with an S/P ≥ 0.25 were classified as having positive results for MAP antibodies. In a validation study,⁹ sensitivity of this test varied from 15.5% to 88.1%, depending on the clinical stage and amount of MAP fecal shedding of each infected cow.

Bacterial culture for MAP—The 72-hour sedimentation culture method for fecal samples was used to culture MAP from fecal samples. The detailed protocol for this method has been described elsewhere.⁴ Briefly, fecal samples (2 g) were mixed with 40 mL of sterile distilled water for 30 minutes and then placed on a horizontal surface and allowed to settle for 1 hour. An aliquot (5 mL) was aspirated from the top of the suspension and transferred into a 50-mL tube that contained 35 mL of 0.9% hexadecyl pyridinium chloride. The tube was thoroughly mixed (manually shaken), and the suspension was allowed to settle for 72 hours. Sediment (2 mL) was then collected from the bottom of the tube and mixed with 0.1 mL of 50 mg of amphotericin B rehydrated in 10 mL of sterile distilled water. Four tubes of Herrold egg medium containing Mycobactin J and control tubes of Herrold egg medium without Mycobactin J were each inoculated with 0.2 mL of the suspension. Caps were loosely applied to the tubes, and tubes were then incubated in a horizontal position at 37°C for 1 to 3 weeks. When there was a dry appearance on the surface of the media at that time, caps were tightened, and tubes were incubated for an additional period. Beginning at approximately 6 weeks after start of incubation, tubes were examined on a weekly basis to detect growth of MAP. Results were reported as negative, positive-low shedder (1 to 10 colonies/tube), positive-moderate shedder (10 to 50 colonies/tube), positive-high shedder (51 to 100 colonies/tube), or positive–very high shedder (> 100 colonies/tube) per gram of fecal sample.

Estimation of herd MAP infection prevalence—The initial evidence used to indicate that selected herds had MAP infection was based on the recollections of the producers with regard to culling of cows infected with MAP during the year preceding the start of the study. To confirm that these herds were infected with MAP at the conclusion of the study in 2007, the investigators estimated the true within-herd prevalence of MAP for each herd. The apparent prevalence of MAP infection in each herd was first determined by testing 30 cows (other than those in the study cohort) that were in their second or greater lactation; these cows were randomly selected from each herd during the last sampling event conducted at 54 months for the detection of antibodies against MAP. Blood samples were collected as described previously and tested by use of a commercial ELISA kit^d with an estimated specificity of 98.6% and sensitivity of 50%.⁹ For each herd, true prevalence estimates for MAP were calculated from the estimated apparent prevalence by use of the following equation¹⁰:

View Expanded

where p (MAP) was the true MAP prevalence and AP the apparent MAP prevalence for each herd, respectively.

The Minnesota Veterinary Diagnostic Laboratory maintained an electronic database with records of test results for all samples submitted by veterinary clinics serving many dairies in Minnesota and western Wisconsin. Test records for MAP infection (serologic ELISA [based on testing ≥ 30 cows and ≥ 2 MAP-positive outcomes/record]) were retrieved from the database for each of the 6 herds (herds 2, 3, 5, 9, 11, and 12) in which the estimate of apparent MAP prevalence (as determined on the basis of results for 30 randomly tested cows at the last sampling event) was 0%. These test records were used to estimate apparent prevalence prior to calculating true prevalence estimates for these herds.

Disease outcome definitions—The primary outcome for this study was subclinical MAP infection indicated by a positive test result for the serologic ELISA or bacterial culture of MAP in feces. A third outcome variable was derived by applying a parallel diagnostic test interpretation criterion to both test outcomes (ie, cows were considered infected with MAP when they had positive test results for the serologic ELISA, bacterial culture of feces, or both diagnostic tests).

Statistical analysis—Descriptive and multivariable analyses were conducted by use of standard statistical software.^e For all analyses, values of P ≤ 0.05 were considered significant.

The proportion of cows with positive test results for MAP at each sampling event (approx 30, 42, and 54 months of age) and the cumulative proportion positive for MAP (all months combined) were summarized. The difference in the risk of developing MAP infection between treatment groups (CR vs MC) was tested by fitting logistic regression models to the data while adjusting for random effects of herd (n = 12).

For each cow in the study, the number of months of observation was calculated from the time of enrollment (calf birth date) to the sampling date when the first positive test result for MAP was established or when the study concluded at approximately 54 months of followup monitoring, at which time all surviving cows were censored. Cows that died or were culled for any reason during the 54-month follow-up period were censored at the date of their death or culling. Cows lost during the follow-up monitoring period between weaning and the first testing event (approx 30 months of age) were assumed to have exited the study at weaning and were therefore censored from the analysis at date of weaning (approx 56 days after birth).

Three separate Weibull hazard regression models were fit to the data to evaluate the effect of feeding CR versus MC on the hazard of a positive MAP diagnosis (as determined on the basis of the serologic ELISA, bacterial culture of feces, or both diagnostic tests) while adjusting for herd effect by inclusion of an inverse-Gaussian shared-herd frailty term.^11,12 The Weibull distribution provides a variety of monotonically increasing or decreasing shapes of the hazard function determined by the estimated parameter, ρ. For ρ = 1, it implies a constant hazard function. For ρ < 1, it suggests a monotonically decreasing hazard function, whereas ρ > 1 is indicative of a monotonically increasing hazard function.¹²

In the study reported here, the Weibull distribution was chosen to model the hazard of MAP infection in CR (vs MC) treatment groups for 2 reasons. The first is that it provided a good fit to the data in comparison to other known hazard distributions (eg, exponential, Gompertze, log-normal, and log-logistic), as determined on the basis of the AIC computed for each model with treatment group (CR vs MC) as the only predictor. Detailed description for computing the AIC has been reported elsewhere,¹² and per this criterion, the best-fitting model is the one that yields the smallest AIC value. The second reason was that the long incubation period for MAP infections in cattle (assumed to be ≥ 2 years) meant that with the passage of time, the hazard rate of detecting infection increases. Because the diagnostic tests used in the study reported here have a sensitivity that increases with progression of JD from a subclinical to a clinical state,^9,13 in our opinion, the Weibull hazard distribution provided the most suitable description for the natural course of MAP infection in cattle for the characteristics of the diagnostic tests used to detect MAP infections in the study.

Similarly, the choice between Gamma and inverse-Gaussian distributions to model the shared-herd frailty effect was based on their respective fit for the data, with a better fit indicated by a higher maximized log-likelihood of each full model (ie, models with the frailty terms included) and how well the shape of the frailty distributions fit the theoretic description of the herd effect. The Gamma frailty effect assumes that with passage of time, the effect of model variables on the population hazard decreases in favor of the frailty effect (ie, the herd effects would completely dominate in the long term). On the other hand, an inverse-Gaussian frailty effect assumes that the effects of model variables decrease with time, although they never completely disappear.¹¹

The goodness of fit of the final models was evaluated by generating plots of cumulative hazard function against Cox-Snell residuals with a cumulative hazard approximating a 45° line with a zero intercept, suggesting a good fit of the model to the data.^10,12

The possibility of a treatment-herd interaction was examined, but there was no evidence for such an interaction. Nonetheless, the data were stratified by herd, and estimates of the treatment effect (CR vs MC) on the hazard of MAP infection within each herd were calculated to examine whether individual herd-level responses to feeding of CR were consistent with the average population-wide estimates from the herd-adjusted models.

Power analysis—Power refers to the probability that a significant treatment effect of CR (vs MC) against MAP infection of a certain magnitude will be found when it truly exists.¹⁰ In the study reported here, we tested the null hypothesis that calves fed MC at birth were at no greater risk of being infected with MAP than were calves fed CR at birth. The sample size needed to test this hypothesis was not determined a priori because the initial study⁸ was designed to compare rates for failure of passive transfer and preweaning health (morbidity and deaths) among calves fed MC versus CR. Therefore, we conducted a post hoc power analysis to determine whether the sample of heifer calves enrolled provided the study reported here with sufficient power to enable rejection of the null hypothesis or to mitigate a type II error. The variables for this analysis were estimated from the data and included the observed treatment effect estimated from the model for positive outcomes to the serologic ELISA, bacterial culture, or both tests, a 2-sided α ≤ 0.05, and the observed cumulative incidence of MAP infection (serologic ELISA, bacterial culture, or both tests) among study cohorts.

Results

Herd description, JD management, and herd infection status—Number of cattle in participating herds ranged from 190 to 1,500 cows (median, 594 cows). Fifty-two percent of dairy herds in Minnesota¹⁴ and 54% of dairy herds in Wisconsin¹⁴ are within this range for number of cows in the herd. Lactating cows were housed in free stalls at all participating dairies. Mean milk production (rolling herd average) for each farm ranged between 9,773 and 13,636 kg/y (21,500 and 30,000 lb/y), with a median of 11,727 kg/y (25,800 lb/y). Bulk-tank somatic cell count ranged between 180,000 and 280,000 cells/mL (median, 240,000 cells/mL).

Interval from birth to feeding of the first colostral meal ranged from 36 to 138 minutes (median, 81 minutes). Colostrum was bottle-fed to calves in 6 herds, whereas the remaining 6 herds fed colostrum to calves via an esophageal tube feeder. Calves were raised onsite at 6 farms, whereas the other 6 farms transported young (1- to 3-day-old) calves to be raised off-site at the facilities of 1 of 2 professional heifer growers.

Distribution of estimates for true within-herd MAP prevalence determined on the basis of testing 30 randomly selected cows at the last sampling event in 2007 by use of a serum ELISA or past ELISA test records (herds 2, 3, 5, 9, 11, and 12) obtained from the Minnesota Veterinary Diagnostic Laboratory database was summarized (Figure 1). All participating herds were infected with MAP, with the burden of infection varying among herds. The estimated true MAP prevalence for each herd ranged between 0.43% and 64% (median, 8%).

View Full Size

Distribution for estimated true MAP prevalence of 12 herds participating in a study to evaluate the effect of feeding of CR versus MC to heifer calves at birth on subsequent development of MAP infection.

Citation: Journal of the American Veterinary Medical Association 234, 9; 10.2460/javma.234.9.1167

• Download Figure
• Download figure as PowerPoint slide

Of the 9 farms in Minnesota, 8 (herds 4, 5, 6, 7, 8, 9, 10, and 11) were participants in the Minnesota Voluntary JD Control Program.¹⁴ Only 1 (herd 12) of the 3 farms enrolled from Wisconsin reportedly participated in a similar voluntary JD control program. Testing strategies for MAP varied among study herds and were not mutually exclusive because some of the herd owners reported use of > 1 MAP testing strategy. Five herd owners (herds 1, 2, 4, 10, and 12) reported that they tested only cattle that had clinical signs of JD, whereas 1 herd owner (herd 7) reported that testing was conducted only on cattle purchased from external sources. Four herd owners (herds 5, 6, 7, and 8) reported that they tested cattle only at the end of lactation. Whole-herd testing for MAP was reported for 2 herds (herds 8 and 9). For 5 herds (herds 1, 4, 5, 10, and 11), the decision to cull a cow suspected of being infected with MAP was based on both a positive diagnostic test result and detection of clinical signs of JD. On the other hand, the culling decisions for 6 herds (herds 2, 3, 6, 8, 9, and 12) were based on clinical signs of MAP alone, whereas in only 1 herd (herd 7) were MAP-suspect cows culled solely on the basis of positive diagnostic test results for MAP.

Descriptive analyses—A total of 497 Holstein heifer calves born between July and October 2003 were enrolled in the study; 261 of these received MC, and 236 received CR. The slight imbalance in number of calves enrolled between the MC versus CR groups was attributed to the fact that early during the study, personnel on some participating dairies who were working weekend shifts only fed MC to newborn calves. This was not likely to have introduced a bias for the study findings.

The number of cattle lost to follow-up monitoring between weaning and the first testing event at 30 months of age was more than twice as high in the CR group (12/236 [5.1%]), compared with that in the MC group (6/261 [2.3%]), although these values did not differ significantly (χ² = 3.33; P = 0.07). Cumulative incidence for losses during the follow-up period was 20 of 497 (4.0%). By assuming that the calves lost to follow-up monitoring were censored from the study at weaning, bias attributable to a loss to follow-up monitoring, which would have disproportionately affected the CR group, may have been avoided in this analysis.

The proportion of cows in the MC and CR groups that had positive results for MAP for the first, second, and third testing events was determined (Table 1). The cumulative proportion of cows with positive results for MAP by use of the ELISA was 25 of 497 (5.0%), with 16 of 261 (6.1%) in the MC group versus 9 of 236 (3.8%) in the CR group. Similarly, the cumulative proportion of cows with positive results for MAP on the basis of results of bacterial culture of feces was 46 of 497 (9.3%), with 29 of 261 (11.1%) in the MC group versus 17 of 236 (7.2%) in the CR group. When the positive outcomes for the serologic ELISA and bacterial culture were combined via a parallel test interpretation criterion, the cumulative proportion of cows with positive test results for MAP was 49 of 497 (9.9%), with 31 of 261 (11.9%) in the MC group versus 18 of 236 (7.6%) in the CR group.

Multivariable logistic regression analysis—Results from the logistic regression models adjusted for random herd effects and comparing the risk of developing MAP infection in CR versus MC groups for the first, second, and third testing events were summarized (Table 1). When a positive outcome for the serologic ELISA was considered the only test criterion for interpreting MAP infection status, cows fed CR at birth were 55% (OR = 0.488; P = 0.118) less likely to be infected with MAP, compared with the likelihood for cows fed MC at birth. Similarly, when a positive outcome for bacterial culture was considered the only test criterion for interpreting MAP infection status, cows fed CR at birth were 46% (OR = 0.538; P = 0.067) less likely to be infected with MAP, compared with the likelihood for cows fed MC at birth. When positive outcomes for the serologic ELISA and bacterial culture were combined via a parallel test interpretation criterion, cows fed CR at birth were 48% (OR = 0.523; P = 0.051) less likely to be infected with MAP, compared with the likelihood for cows fed MC at birth.

Multivariable time-to-MAP positive model analysis—The AIC was computed for each hazard distribution. For the ELISA model, Weibull (AIC = 174), exponential (199), Gompertze (179), log-normal (171), and log-logistic (174) were calculated. For the bacterial culture model, Weibull (AIC = 251), exponential (305), Gompertze (259), log-normal (246), and log-logistic (250) were calculated. For the both-tests model, Weibull (AIC = 257), exponential (318), Gompertze (266), log-normal (253), and log-logistic (257) were calculated.

In the analysis for the study reported here, the inverse-Gaussian shared-herd frailty distribution resulted in models with a better fit than that for the Gamma shared-herd frailty distribution (log-likelihood for the serologic ELISA, −71.446 vs −72.249 for inverse-Gaussian and Gamma shared-herd frailty, respectively; bacterial culture, −111.731 vs −111.691 for inverse-Gaussian and Gamma shared-herd frailty, respectively; and combination of serologic ELISA and bacterial culture, −113.418 vs −113.461 for inverse-Gaussian and Gamma shared-herd frailty, respectively). Thus, the inverse-Gaussian shared-herd frailty models were preferred. Evaluation of the shapes of plots of cumulative hazard function against Cox-Snell residuals suggested that the final models provided relatively good fits for the data (results not provided).

Predicted hazard functions for each final model comparing CR versus MC groups were plotted (Figure 2). In all cases, a hazard of close to zero was detected up to approximately 30 months, which reflected the duration of the follow-up monitoring period when no MAP testing was implemented. The hazard rate of MAP infection gradually increased in both groups up to approximately 54 months, but the increase was more rapid in the MC group, which suggested that cows in the MC group had a greater likelihood of developing MAP infection with the passage of time, compared with the likelihood for cows in the CR group.

View Full Size

Weibull hazard functions for the analysis of interval until a positive test outcome for MAP, as determined by use of a serologic ELISA (A), bacterial culture of fecal samples (B), or a combination of both diagnostic tests (C), in cows fed CR (solid lines) or raw bovine MC (dotted lines) at birth.

Citation: Journal of the American Veterinary Medical Association 234, 9; 10.2460/javma.234.9.1167

• Download Figure
• Download figure as PowerPoint slide

Results for the Weibull regression models adjusted for the shared-herd frailty effect determined on the basis of records for 497 heifer calves in 12 herds were summarized (Table 2). The estimated shape parameters for the models were as follows: ELISA, ρ = 3.54; bacterial culture, ρ = 3.65; and both diagnostic tests, ρ = 3.77. These values suggested a monotone increasing hazard of MAP infection for both the CR and MC groups. The shared-herd frailty effects were significant (P < 0.001) in all 3 models, which suggested the existence of true unmeasured herd effects.

The models in this study accounted for herd as a random-effect variable (frailty), and the fixed effects for treatment (CR vs MC) were interpreted as population average estimates (ie, treatment effect estimates had herd-specific independent interpretations).¹⁵ Cows in the CR group had a 53% (HR = 0.474; P = 0.081) reduction in the hazard of MAP infection (as determined on the basis of results for the ELISA), compared with the hazard for cows in the MC group. For cows that had positive results for bacterial culture, a 43% (HR = 0.572; P = 0.076) reduction in hazard of MAP infection was detected for the CR group, compared with the hazard for the MC group. When outcomes for the ELISA and bacterial culture were combined via a parallel test interpretation criterion, a 44% (HR = 0.558; P = 0.056) reduction in the hazard of MAP infection was detected for cows in the CR group, compared with the hazard for cows in the MC group.

Estimates of individual herd treatment effects—Estimates of the treatment effect on MAP infection (modeled for serologic ELISA, bacterial culture, or both) stratified on the basis of herd was determined (Table 3). Cows in 4 herds (herds 7, 9, 10, and 11) did not have a failure event (ie, no positive test results for MAP in the study cohort during the 54-month study period). Therefore, the effect of treatment within these 4 herds could not be estimated. There was a significant protective effect of feeding CR for herd 5 (HR = 0.24; P < 0.007). Although numeric estimates of treatment effect revealed that sometimes treatment was protective (herds 3, 4, and 6) or not protective (herds 1, 2, and 8), there was no significant difference between treatment groups for these herds. These herd-stratified data should be interpreted with caution because no overall herd-by-treatment interaction was detected and because stratification of the data by herd would have significantly reduced the analytic power for each estimate of within-herd treatment effect.

Power analysis—Variables for the power analysis included the observed treatment effect estimated from the model for outcome of the ELISA, bacterial culture, or both tests (Table 2), a 2-sided α ≤ 0.05, and the observed cumulative incidence (49/497 [9.9%]; Table 1) of MAP infection (ELISA, bacterial culture, or both diagnostic tests) among study cohorts. This analysis revealed that a sample of this size (n = 497 heifer calves [assuming calves were evenly allocated between groups]) provided the current study with a statistical power ≥ 80% to enable the rejection of the null hypothesis at P ≤ 0.05 if the true reduction in hazard of MAP infection had been 51% (HR = 0.487) in the population of cows fed CR versus MC at birth (compared with the 44% [HR = 0.558] risk reduction estimated from the current data).

Discussion

To our knowledge, the study reported here is the first prospective, long-term evaluation of the role of MC in the epidemiology of MAP and the efficacy of a CR product for prevention of MAP transmission in preweaned Holstein dairy calves. These results provide evidence that feeding a CR product reduced the risk of developing MAP infection in Holstein calves.

A major strength of the study was its design (ie, a prospective randomized clinical trial conducted in 12 herds with JD). This study provided us with the ability to potentially identify cause-and-effect relationships between the 2 colostrum-feeding programs (CR vs MC) and risk of developing MAP infection, thus providing dairy producers with evidence needed to judge the performance of alternative colostrum-feeding programs for the prevention of MAP transmission within commercial dairy herds. An additional strength of the study was the fact that data analysis by use of Weibull and inverse-Gaussian shared-herd frailty models allowed for the adjustments of herd effects, thus enabling the possible generalization of study results to a wider population of herds with management systems similar to those included in this study because the estimates of the fixed effects of treatment (CR vs MC) had a population average interpretation¹⁵ (ie, HR estimates calculated on the basis of the hazard of MAP infection among cows in the CR group vs hazard of MAP infection among cows in the MC group were not herd-specific effects and were therefore interpreted independent of the herd effects).

We were not able to test the dams for MAP infection prior to birth of each calf. The possibility of in utero MAP transmission has been reported^16,17; thus, some calves in the study may have been infected prior to receiving their first colostral meal. However, we are confident that the potential risk of MAP transmission to calves through in utero exposure was balanced between treatment groups by way of systematic allocation of calves to the treatment groups at the time of enrollment. Therefore, the potential for in utero MAP transmission is unlikely to have biased the results of this study.

Herd was included in the final models as a random-effect variable, but initial adjustment for herd as a fixed-effect variable yielded treatment-effect estimates (ELISA or bacterial culture [or both diagnostic tests] model, b = −0.603 [P = 0.049] and −0.582 [P = 0.056] for fixed-effect herd adjustment and random-effect herd adjustment, respectively) that did not alter the final conclusions derived from the models in which herd was considered a random effect. Because the models in which herd was a fixed effect were not of interest in this study (and because there were a large number of herds [n = 12 herds], which may have caused parameter estimates to become unstable) and were considered nuisance parameters meant to adjust for within-herd clustering,¹⁸ we decided to model herd as a random-effect variable, although herds that participated in the study were a convenience sample (not a random sample) from an underlying population of dairy herds in Minnesota and Wisconsin.

Although adjusting for herd-level clustering is a necessity in a study such as this, a potential limitation of adjusting for the herd effect as a random variable is that it typically amalgamates the effect of treatment in larger herds (herds that enrolled more calves in the study), which yields average population-effect estimates with less variation, making it difficult to discern those herds likely to benefit (or not benefit) from the intervention. For this reason, although there was no evidence for a treatment-herd interaction in this study, we analyzed the data after stratifying by herd to examine treatment responses for each herd. Heterogeneity of treatment response was detected at the herd level. For example, there was a numeric increase in risk of developing MAP infection in cows fed CR at birth, compared with the risk for infection in cows fed MC at birth, in 4 herds, but this apparent increase was not significant. We initially suspected that owners of these herds may have made a conscious effort to test cows at the end of lactation and then withheld feeding of colostrum from MAP-positive dams at the time of enrollment. The consequence of this could have been a potential reduction in risk of developing MAP infection among calves enrolled in the MC group, thus biasing the treatment effect toward the null. However, investigations by the authors yielded no evidence to suggest that testing of adult cows for MAP was undertaken in study herds with the intended purpose of withholding the feeding of suspect colostrum from the calves of that herd. Additionally, a numeric reduction in risk of developing MAP infection in cows fed CR at birth, compared with the risk for cows fed MC at birth, was detected in 3 herds, but this probable protective effect was significant in only 1 herd (herd 5).

Possible explanations for the variability in treatment response among herds included random error caused by chance, although it is entirely possible that a CR-feeding program provided little to no added protection to Holstein calves against MAP infection in some herds but offered absolute benefits in others; stratification of the data by herd, which may have resulted in a type II error (ie, the relatively small sample size [mean, 43 calves enrolled/herd; range, 14 to 83 calves enrolled/herd] within herds may have compromised the power to detect a significant treatment effect and increased the variability in treatment response at the herd level); and a short observation period for time at risk, relative to the long incubation period of MAP infection (which is believed to range from 24 to 72 months from exposure to disease onset). The cumulative median time at risk for cows in these herds was approximately 41 months (39 months for the CR group vs 42 months for the MC group). Infection with MAP in cattle is believed to have a prolonged course, with clinical manifestation of the disease at ≥ 24 months of age in calves exposed to MAP within the first 6 months after birth. In view of this limitation, and given that the mean ± SD censoring rate per herd was approximately 98 ± 3.0% between birth and 54 months of age in the follow-up period, it is possible that several MAP-infected cows were censored from these herds for reasons unrelated to the treatment prior to being tested or before having a positive result for MAP during testing. Moreover, the diagnostic tests used to detect MAP infection in this study have limited and unreliable sensitivity during the early phases of the disease in cattle.^9,13 Consequently, the risk of developing MAP infection within herds may have been underestimated, which may have further exacerbated the type II error for herd-level treatment effects. The lack of a significant treatment effect with regard to the risk of developing MAP infection in some herds may have also been attributable to insufficient concentrations of MAP in the MC to result in effective transmission and consequent establishment of infection in calves. The latter may have been a result of a low within-herd burden of JD for some herds (herds 1, 2, 6, 7, and 10) and therefore a reduction in overall risk of developing MAP infection. These factors, in addition to the limited period of exposure to MC (calves were fed MC once or twice within the first 24 hours after birth) and limited quantity of MC fed (approx 2 or 4 L) to calves in these herds, it is more likely that calves in the MC group may not have been infected despite possible exposure to MAP-containing raw MC.

Notwithstanding the aforementioned explanations, the variability and inconsistency in the treatment response among herds should have been expected. In the opinion of the authors, the variable responses to the treatment among herds suggested that we managed to include diverse herd populations in the study. If there truly were herd-level circumstances that may have affected the efficacy of CR product, the authors were unable to identify in this study the factor or factors that may have predicted herds with a beneficial effect or herds without any apparent beneficial effect. Future studies should investigate this issue. However, given the significant reduction in analytic power resulting from stratification of the data by herd, results of this stratified analysis should be interpreted with caution.

All 12 herds enrolled in the study were confirmed to be infected with MAP and were medium (100 to 499 cows) to large (≥ 500 cows) dairy herds. However, these herds were not randomly selected from the entire population of dairy herds in Minnesota and Wisconsin. Study herds were representative of 52% and 54% of Minnesota and Wisconsin dairy herds, respectively, as determined on the basis of herd size.¹⁴ As such, caution must be used when extrapolating results to smaller (< 100 cows) dairy herds.

The post hoc power analysis conducted for the study reported here revealed that a sample of this size (n = 497 heifer calves [assuming calves were evenly allocated between groups]) provided the current study with a statistical power ≥ 80% to enable the rejection of the null hypothesis at P ≤ 0.05 if the reduction in the hazard of MAP infection had been 51% (HR = 0.487) in the population of cows fed CR versus MC at birth (instead of the 44% [HR = 0.558] hazard reduction estimated from the data for this study). Although post hoc analysis is a less appealing method, in comparison with methods that estimate sample size and the associated statistical power a priori, in the opinion of the authors, this exercise may provide information for other investigators attempting to address a similar concern.

Despite the variable responses to feeding of CR among the herds, our findings revealed that the average population estimate of the treatment effect against MAP infection in this study nonetheless provided reasonable indications of the probable benefits that may accrue in some herds as a result of feeding programs that incorporate CR products. In the study reported here, the incidence of MAP infection was higher in calves fed MC at birth, compared with the incidence in calves fed CR at birth. For all herds, it was estimated that feeding CR instead of MC reduced the risk for MAP infection in calves by 44%. The most plausible explanation for the protective effect of feeding CR at birth is that the CR did not contain infective MAP, whereas fresh bovine MC served as an important source of transmission for MAP in Holstein calves, as suggested in another study.⁶ Although the study reported here revealed that programs that involve feeding CR products may have a role in JD control programs, the overall cost:benefit analysis of adopting this management tool has yet to be determined. Analyses are being conducted to evaluate the effect of feeding CR versus MC on survivorship in the herd, milk production, and reproductive performance, as well as overall results of cost:benefit analysis of adopting programs that involve feeding CR versus MC.

Because the findings of this study suggested that feeding raw MC from herds endemically infected with JD may represent a risk factor for MAP transmission, these results indicate the need to investigate other management strategies for reducing transmission of MAP through MC. First, it would seem sensible to advise producers to withhold feeding colostrum from MAP-positive dams to heifer calves. Also, research is ongoing to evaluate whether on-farm heat treatment of MC may provide an alternative method of controlling MAP transmission via colostrum. Laboratory studies^19,20 have determined that heat treatment of bovine colostrum at 60°C for 60 minutes is sufficient to eliminate MAP and other important pathogens without damaging important colostral immunoglobulins. Furthermore, a field study²¹ revealed that calves fed heat-treated colostrum (60°C for 60 minutes) had enhanced passive transfer of IgG. Research is being conducted to investigate the short- and long-term impact of feeding heat-treated colostrum on health, productivity, and risk for developing MAP infection in adult dairy cows.

The study reported here revealed that feeding CR at birth (vs feeding raw MC at birth) reduced the risk of developing MAP infection in Holstein calves in JD-endemic dairy herds. Results of this study also suggested that feeding of raw bovine MC may represent a risk factor for MAP transmission in Holstein calves.