Accepted Manuscript

Title: Associations between prior management of cattle andrisk of bovine respiratory disease in feedlot cattle

Author: K.E. Hay J.M. Morton M.L. Schibrowski A.C.A.Clements T.J. Mahony T.S. Barnes

PII: S0167-5877(16)30070-8DOI: http://dx.doi.org/doi:10.1016/j.prevetmed.2016.02.006Reference: PREVET 3988

To appear in: PREVET

Received date: 18-10-2015Revised date: 21-2-2016Accepted date: 23-2-2016

Please cite this article as: Hay, K.E., Morton, J.M., Schibrowski, M.L., Clements,A.C.A., Mahony, T.J., Barnes, T.S., Associations between prior management of cattleand risk of bovine respiratory disease in feedlot cattle.Preventive Veterinary Medicinehttp://dx.doi.org/10.1016/j.prevetmed.2016.02.006

This is a PDF file of an unedited manuscript that has been accepted for publication.As a service to our customers we are providing this early version of the manuscript.The manuscript will undergo copyediting, typesetting, and review of the resulting proofbefore it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers thatapply to the journal pertain.

1

Associations between prior management of cattle and risk of bovine respiratory disease in feedlot cattle

KE Haya, JM Mortonbc, ML Schibrowskia, ACA Clementsd, e, TJ Mahonya, TS Barnesab

a The University of Queensland, Queensland Alliance for Agriculture and Food Innovation,

St Lucia 4072, Queensland, Australia

b The University of Queensland, School of Veterinary Science, Gatton 4343, Queensland,

Australia

c Jemora Pty Ltd, PO Box 2277, Geelong 3220, Victoria, Australia

d The University of Queensland, Infectious Disease Epidemiology Unit, School of Population

Health, Herston 4006, Queensland, Australia

e Current address: The Australian National University, ANU College of Medicine,

Biology and Environment, Research School of Population Health, Canberra 0200, Australian

Capital territory, Australia

* Corresponding author: Dr Timothy Mahony. Tl: +61 7 33466505; fax:+61 7 3346 6501

e-mail address: t.mahony@uq.edu.au

2

Abstract

Bovine respiratory disease (BRD) is the major cause of clinical disease and death in feedlot

populations worldwide. A longitudinal study was conducted to assess associations between

risk factors related to on-farm management prior to transport to the feedlot and risk of BRD

in a population of feedlot beef cattle sourced from throughout the cattle producing regions of

Australia. Risk factors were derived from questionnaire data provided by farmers supplying

cattle (N=10,721) that were a subset of the population included in a nationwide prospective

study investigating numerous putative risk factors for BRD. Causal diagrams were used to

inform model building to allow estimation of effects of interest. Multilevel mixed effects

logistic regression models were fitted within the Bayesian framework. Animals that were

yard weaned were at reduced risk (OR: 0.7, 95% credible interval: 0.5 to 1.0) of BRD at the

feedlot compared to animals immediately returned to pasture after weaning. Animals that had

previously been fed grain (OR: 0.6, 95% credible interval: 0.3 to 1.1) were probably at

reduced risk of BRD at the feedlot compared to animals not previously fed grain. Animals

that received prior vaccinations against Bovine viral diarrhoea virus 1 (OR: 0.8, 95%

credible interval: 0.5 to 1.1) or Mannheimia haemolytica (OR: 0.8, 95% credible interval: 0.6

to 1.0) were also probably at reduced risk compared to non-vaccinated animals. The results of

this study confirm that on-farm management before feedlot entry can alter risk of BRD after

beef cattle enter feedlots.

Key words: Bovine respiratory disease, risk factors, vaccination, feedlot cattle, longitudinal

study, yard weaning

3

Introduction

Bovine respiratory disease (BRD) is the major cause of clinical disease and death in feedlot

cattle, both in Australia (Sackett et al., 2006) and elsewhere (Miles, 2009). BRD actually

comprises a heterogeneous complex of diseases involving the respiratory system of cattle. No

single factor acting in isolation is able to cause BRD; component causes consisting of

pathogenic organisms, stress, immunological susceptibility and animal management are some

of the contributors to the multifactorial aetiology. Whilst decades of research have advanced

our knowledge of many of these component causes (Fulton, 2009), BRD remains a major

problem in intensively managed cattle populations worldwide. It has become increasingly

clear that more effective management of BRD requires an improved understanding of the

contributing biological pathways and approaches to animal management that simultaneously

address multiple pathways.

An increasing body of evidence indicates that the risk of feedlot cattle developing BRD is

strongly influenced by a multitude of exposures occurring prior to their arrival at the feedlot.

Prior management history of cattle entering feedlots has long been considered important in

determining susceptibility to BRD and hence BRD incidence at the feedlot (Wieringa et al.,

1976; Schipper et al., 1989). To reduce concurrent exposure to stressors associated with the

transition from pasture-based herds to feedlots, ‘preconditioning’ programs focusing on

presale management practices have been advocated. Recommended protocols include

ensuring that stressful procedures such as weaning, castration and dehorning occur at least

one month before feedlot entry. Preconditioning may also include the introduction of

roughage and concentrates, with training to use feed bunks, parasite control and vaccination

against respiratory pathogens well before feedlot entry (Woods et al., 1973; Schipper et al.,

1989). Although some researchers have concluded that preconditioning programs are

4

associated with reduced BRD incidence and mortality rate after feedlot entry (Schipper et al.,

1989; Taylor et al., 2010), varying definitions of ‘preconditioning’ have been used and it has

often not been possible to separate the effects of the individual components of various

programs. Improved BRD outcomes were observed with either preconditioning or just

vaccination (Roeber et al., 2001; Macartney et al., 2003), and with retention of calves on the

farm of origin for 45 days after weaning and shipping direct to the feedlot (Step et al., 2008).

However preconditioning was not associated with reduction in BRD incidence in a large

retrospective study, (Sanderson et al., 2008).

In Australia, weaning normally occurs when calves are six to ten months of age. Upon

separation from their dams, ‘weaners’ (i.e. weaned calves aged under 12 months) may be

immediately returned to pasture (‘paddock weaned’) or they may be ‘yard weaned’. Yard

weaning is a practice whereby immediately following weaning, calves are held in small yards

and introduced to daily handling, water troughs, feed bunks, concentrated rations and/or

conserved forage. It is recommended that calves be kept in yards for at least the first 5 to 7

days from weaning at a recommended stocking density of about four square meters per head

for animals weighing 180-260 kg (Kahn and Cottle, 2014). An Australian study involving

two separate experimental groups, each of about 200 animals compared three weaning

methods following traditional abrupt separation: paddock weaning, yard weaning with hay or

silage and yard weaning with hay or silage plus training to use feed bunks with the

introduction of grain. Half of each group also received experimental BRD vaccines against

one or more viruses associated with BRD (Walker et al., 2007). All study cattle were

subsequently grazed on pasture before being moved to a feedlot six to nine months after

weaning. Yard-weaned animals had significantly better weight gain at the feedlot than

paddock-weaned animals, with the best performance in yard-weaned, vaccinated animals.

5

Yard-weaned animals also had lower BRD incidence, but this was not significant at the 5%

level (Walker et al., 2007).

The National Bovine Respiratory Disease Initiative (NBRDI) was a longitudinal study

conducted in Australia to investigate numerous putative risk factors for BRD in feedlot cattle.

To investigate the effects of exposures occurring prior to cattle arrival at the feedlot,

management practices employed by cattle producers who supplied study cattle directly to

feedlots were documented. Questionnaire data were collected from producers who sold

groups of 20 or more cattle that were subsequently enrolled in the NBRDI; these data were

used to derive variables describing on-farm management practices and the estimated age of

cattle at feedlot entry. The aims of the current study were to investigate associations between

these risk factors and occurrence of BRD at the feedlot.

Methods

Detailed descriptions of the study design, data sources, project population, case definition and

analysis methods used in the NBRDI have been presented elsewhere (Hay et al., 2014). The

NBRDI project population comprised 35,131 animals nested within 170 cohorts nested within

14 feedlots. Animals were enrolled in the study at feedlot induction, when animals were

individually identified and animal characteristics recorded, before they were placed in a study

cohort. Each cohort was defined as the cattle managed together in the same feedlot pen. Each

animal was monitored from induction, defined as ‘day 0’, until it left the study cohort for any

reason; the most common reasons were removal to the hospital pen or another pen separate

from the cohort, death or removal from the feedlot. The case definition of BRD was based on

clinical signs indicating respiratory system involvement at the animal’s first examination in

the feedlot’s hospital crush after induction; those with diagnoses of ‘pneumonia’,

‘respiratory’, ‘BRD’ and ‘IBR’ (infectious bovine rhinotracheitis) were classified as BRD

6

cases (Hay et al., 2014). The outcome of interest was the development of BRD during the

animal’s first 50 days following induction. The unit of analysis was the individual animal.

Participating feedlots were requested to contact cattle producers (‘vendors’) that supplied

cattle directly to the feedlots to determine if they were agreeable to being approached by

project staff to respond to the questionnaire. Based on data provided by feedlot personnel, an

estimated 82% of the study population were sourced directly from cattle producers and 3%

were sourced from saleyards immediately prior to being placed ‘on feed’, defined as being

fed a feedlot ration in a feedlot pen, at the feedlot. The remaining 15% were sourced from

either saleyards or directly from cattle producers, but were ‘pre-assembled’ which involved

being assembled on pasture close to the vicinity of the feedlot, typically for several weeks

prior to being placed on feed at the feedlot (Hay et al., 2014). ‘Vendor questionnaires’ were

sent to all cattle producers that supplied ‘arrival groups’ of 20 or more animals from the same

farm at the same time and for which contact details were provided by feedlot staff. The

questionnaire aimed to ascertain arrival group-level details of on-farm management and the

timing of those exposures prior to movement to the feedlot. The questionnaire comprised

three sections; one section was common to all cattle and the other sections differentiated

between animals born on the vendor’s farm (i.e. ‘vendor bred’) and animals that had been

purchased from another producer either directly or through saleyards. A summary of the

questionnaire data collected and how it was used to derived variables for use in analysis is

provided in Appendix 1. Vendor questionnaires received as hard copy or administered over

the telephone were entered into the online questionnaire by project staff so that all responses

had equivalent electronic records. Vendor questionnaire data were periodically downloaded,

checked and compiled in Microsoft® Excel, before importing, saving and cross checking

using Stata® (version 12).

7

A number of questions included both the timing (i.e. month, year) of events such as dates of

purchase of the study cattle and dates of management procedures, and the average age of the

study animals at that time. These data were compiled and used to estimate the average age of

animals in the group at induction. The group-level categorical variable (‘Age at induction’: <

16, 16 to < 22 and ≥ 22 months) was derived, although there was a moderate amount of

missing data (9.2%) for this variable.

Cattle that were born on the vendor’s farm comprised the ‘vendor-bred subset’. Data about

marking/branding, commingling on the farm, weaning management and feeding history were

collected for these animals. This subset was used to analyse the effects of yard weaning as

well as prior feeding history and on-farm mixing. The effects of yard weaning were analysed

using the categorical variable (‘Weaning detail’: paddock weaned, yard weaned for < 7 days,

and yard weaned ≥ 7 days); categories were subsequently collapsed to form a binary variable:

‘Yard weaning’ (yes, no). Prior feeding history was described using two variables. ‘Prior

grain feeding’ (yes, no) indicated whether the animals had ever been fed grain and ‘Prior

supplementary feeding’ (yes, no) indicated whether the animals had ever been fed conserved

forage (e.g. lucerne hay or silage) or other supplements (e.g. mineral licks, molasses). ‘On-

farm mixing’ (yes, no) indicated whether the group of animals had been mixed with other

groups on the vendor’s farm.

Animals that had been purchased by the vendor were often purchased as weaners either

directly from other producers or through saleyards, so the respondent did not know the

animals’ lifetime histories, including their lifetime vaccination history. At the time the study

was conducted, there were only two vaccines available in Australia for on-farm use against

BRD pathogens. Pestigard® was an inactivated whole virus vaccine against Bovine viral

diarrhoea virus 1 (BVDV-1); two doses four to six weeks apart are recommended.

Bovilis MH® was an inactivated whole-cell vaccine registered to protect against Mannheimia

8

haemolytica; two doses three to four weeks apart with the first at least four weeks before

induction, are recommended. It was unlikely that animals would have received these

vaccinations before ten months of age and this was supported by the responses for vendor-

bred animals. Hence, for animals purchased by the vendor, we assumed that none were

vaccinated before this age and expected this to cause minimal classification errors. Thus, the

‘prior-vaccination subset’ (those for which lifetime vaccination history was assumed to be

known) comprised animals purchased by the vendor prior to 10 months of age as well as

those in the vendor-bred subset. Binary variables (vaccinated at least once/never vaccinated)

were used for both vaccines because few animals were vaccinated. About 15% of animals

had received Bovilis MH® prior to induction, of which 57% had received two doses; ‘Prior

BovilisMH vaccination’ (yes, no) indicated the administration of one or more doses of the

vaccine prior to 14 days before induction. About 12% of animals had received Pestigard®, of

which 69% had received two doses; ‘Prior Pestigard vaccination’ (yes, no), indicated the

administration of one or more doses of this vaccine prior to 14 days before induction.

As previously reported, causal diagrams were used to inform model building in the NBRDI

(Hay et al., 2014). Causal diagrams were constructed based on a priori postulated causal

pathways, represented by unidirectional arrows, linking exposures of interest with each other

and with the BRD outcome. A direct effect was indicated by an arrow that directly linked an

exposure with the outcome. Indirect pathways were those linking the exposure to the outcome

via intervening variables. The total effect of an exposure on an outcome was the sum of the

direct effect (if one was postulated) and all indirect effects; causal diagrams facilitate the

selection of models to estimate both the total and direct effects (Dohoo et al., 2009). The

diagram illustrated in Figure 1 displays pathways relevant to analyses described in the current

study. Exposures of interest, indicated in bold, were those described above that were derived

from the vendor questionnaire data. Other variables were included in the diagram if they were

9

included in any of the models used to estimate the effects reported in the current study. Both

dentition (assessed at the animal level but only a crude measure of age) and age (estimated at

the group-level based on survey responses) were included in the diagram because they were

measured at different levels. The definitions and categories of the covariates are reported in

Appendix 2. Several of these covariates were significantly associated with risk of BRD in the

NBRDI project population (Hay et al.; Hay et al.; Hay et al., 2014).

The causal diagram displayed in Figure 1 was reproduced within the DAGitty® software user

interface (Textor et al., 2011) to determine appropriate sets of covariates to include in models

to estimate the total effects of exposures being investigated on the BRD outcome. Separate

models were used to estimate direct effects where these were also of interest. To simplify

analyses and to avoid incorrectly assuming linearity of associations with the logit of BRD by

day 50, all continuous predictors were categorised. The Stata® statistical software package

was used for all data management, preliminary analyses and to run the multilevel modelling

software package, MLwiN® (version 2.27) which was used for modelling.

The 50-day cumulative incidences of BRD in different subsets of the study population were

adjusted for clustering by feedlot. Three-level Bayesian logistic regression models were fitted

with random effects for feedlot and cohort nested within feedlot. Non-informative prior

distributions were used. For each exposure of interest, a model was fitted using second-order

penalised quasi-likelihood methods to produce starting values for a second model using

Bayesian Markov chain Monte Carlo (MCMC) methods. Convergence was assessed by

inspecting diagnostic trajectory plots and summary statistics (Browne, 2012). Final models

were run for between 50,000 and 100,000 iterations after a burn-in of 500 iterations.

To further investigate the association between age and BRD, the crude odds ratio and odds

ratios after fitting just the random effects, and after adjusting for induction season alone and

10

source region alone (the variables in the total effects model) without fitting random effects

were compared. The adjustment set of variables used for estimating the direct effects for age

was modified slightly from that specified by DAGitty®, because the original model failed to

converge. Hence, the variable describing the timing of the move to the feedlot was collapsed

into four categories (Appendix 2) and the variable describing exposure to saleyards within the

period day -27 to day 0 was excluded because no animals in the current study had been

through saleyards in that period.

Results

Questionnaires were sent to the vendors of 579 (46%) of 1,257 arrival groups enrolled into

the NBRDI; these vendors supplied 75% of the enrolled animals. Of the 579 groups for which

questionnaires were sent, responses were received for 238, giving a group-level response rate

of 41%. Vendors completed an online questionnaire (58%), returned a hard copy (by post,

fax, or email: 36%) or participated in a telephone interview (6%). The population for the

current study (i.e. the vendor questionnaire subset) comprised 31% (10,721/35,131) of

animals from the NBRDI project population. Within this vendor questionnaire subset, 47%

(5,063/10,721) of animals were vendor-bred. After adjusting for clustering by feedlot, the 50-

day cumulative incidence of BRD in the vendor questionnaire subset (2,006/10,721) was

18.7% (95% CI: 7.9 to 29.5%), compared to 17.6% (95% CI: 8.8 to 32.2%) in the full project

population. The 50-day cumulative incidence of BRD at 21.4% (95% CI: 12.1 to 35.0%) in

the vendor-bred subset (1,084/5,063) contrasted with 15.8% (95% CI 7.4 to 30.6%) observed

in animals purchased by the vendor (874/5,528). The subset for which prior vaccination

status was known (i.e., vendor-bred animals and animals that were purchased by 10 months

of age), comprised 80% (8,580/10,721) of animals in the current study population. The 50-

11

day cumulative incidence of BRD amongst animals purchased by 10 months of age

(513/3,517) was 14.6% (95% CI: 6.0 to 31.2%).

Of animals in the vendor questionnaire subset with sufficient data to estimate age (N=9,731),

55% were aged 16 to < 22 months at induction (Table 1). Cattle aged at least 22 months were

at moderately increased risk of BRD compared to those aged 16 to < 22 months (OR 1.6,

95% credible interval: 1.3 to 2.1,Table 1). This was in contrast to a crude odds ratio of 0.7 for

animals ≥ 22 months relative to 16 to < 22 months of age. After fitting just induction season

(and no random effects), the odds ratio was 0.7, and after fitting just source region (and no

random effects), the odds ratio was 0.9. In contrast, after fitting just the random effects, the

odds ratio for this comparison was 1.5 (95% cred int 1.2 to 1.8), very similar to the adjusted

odds ratio (1.6, Table 1) after fitting the random effects, induction season and source region.

After including intervening (e.g. induction weight, mixing summary) and additional

confounding variables in the model, the direct effect estimates for age were similar to the

total effects (Table 1). The distribution of induction weights for cattle aged at least 22 months

at induction (median: 446 kg, interquartile range: 416 to 474 kg) was similar to the

distribution for animals aged 16 to < 22 months (median: 446 kg, interquartile range: 422 to

468) and overlapped the distribution of induction weights for cattle aged < 16 months

(median: 410 kg, interquartile range: 380 to 440).

The majority of study animals that were born on the vendor’s farm or were purchased by the

vendor prior to 10 months of age had not been vaccinated with Bovilis MH® (85%) or

Pestigard® (88%) prior to day -14 (Table 1). Prior vaccination with Bovilis MH® was

associated with a reduced risk of BRD (OR 0.8, 95% credible interval: 0.6 to 1.0, Table 1).

There was some evidence that prior vaccination with Pestigard® was also associated with a

reduced risk of BRD (OR 0.8, 95% credible interval: 0.5 to 1.1, Table 1).

12

Weaning method, prior feeding history and on-farm mixing were evaluated in the vendor-

bred subset comprising animals born on the vendors’ farms. The majority of these animals

had been yard weaned (80%, Table 2). For cattle that were yard weaned, the length of time

they were held in yards before being returned to pasture ranged from 2 to 21 days, with a

mean and median of 7 days, and an interquartile range from 5 to 10 days. Yard weaning was

associated with a decreased risk of BRD (OR 0.7, 95% credible interval: 0.5 to 1.0, Table 2).

The direct effect estimate was similar (Table 2), indicating that the protective effect of yard

weaning was not mediated through prior feeding of grain or other supplementary feeding.

The majority of animals with vendor questionnaire data that were born on the vendor’s farm

had not been fed grain (77%, Table 2) but had been fed conserved forage or supplement

before leaving the vendor’s farm (84%, Table 2). Total effect estimates were suggestive of a

probable decrease in the risk of BRD associated with prior feeding of grain, but the estimates

were imprecise (OR 0.6, 95% credible interval: 0.3 to 1.1, Table 2). There was no evidence of

a large effect on the risk of BRD associated with prior feeding of conserved forage or other

supplements (Table 2). The majority of animals in the vendor-bred subset had been mixed on

the farm (94%, Table 2). There was no evidence of a large effect of this on the risk of BRD

(Table 2).

Discussion

The practice of yard weaning was associated with a significantly reduced risk of BRD in this

study. The similar total and direct effect sizes indicated that the effect was not mediated via

our postulated intervening variables (i.e. prior grain feeding or prior conserved forage or

supplement feeding). The observed beneficial effect of yard weaning is consistent with

previous work (Walker et al., 2007). Further, in examining the practice in a large population

of animals sourced from a wide geographical area entering numerous cohorts, our study had

13

sufficient power to demonstrate a significant protective effect against BRD. The observed

effect sizes were similar for cattle held less than 7 days or 7 days or more. Animals that were

yard weaned may have been better accustomed to yards, feed bunks, water troughs, more

crowded conditions and handling so that entry to a feedlot pen was associated with less stress

compared to animals that were not yard weaned. Prior feeding of grain was also associated

with reduced risk of BRD, but this did not explain the effect of yard weaning. Thus, both

practices appeared to be independently beneficial in reducing the likelihood of cattle

acquiring BRD.

The effect of prior vaccination against agents causing respiratory disease was evaluated in the

subset of animals that were vendor-bred or purchased by 10 months of age. Total effect

estimates indicated that prior vaccination with Bovilis MH®, which is registered to protect

against M. haemolytica, was associated with a reduced risk of BRD. Our results are

consistent with those of a recent meta-analysis which concluded that prior vaccination against

M. haemolytica was potentially beneficial (Larson and Step, 2012).

Pestigard® is an inactivated BVDV-1 vaccine with label claims to reduce reproductive loss

and assist in the reduction of losses due to BRD. BVDV-1 was commonly circulating in

cohorts of animals enrolled in NBRDI (Hay et al.). Overall, 66% of animals were in cohorts

in which BVDV-1 was detected by qPCR analyses of samples collected from animals in the

same cohort (indicating either transient or persistent infection) (Hay et al.). Further, in a

nested case-control study conducted within the NBRDI study population, at least one animal

seroincreased for BVDV-1 between induction and follow-up (between 35 and 60 days after

inductions) in 88% (142/161) of cohorts that contributed animals to the case-control study

population (Hay et al.). The total effects estimate for prior vaccination with Pestigard®

vaccine against BVDV-1 provided some evidence that vaccination was associated with a

14

reduced risk of BRD. However, the effectiveness of prior Pestigard® vaccination under

feedlot conditions with a high level of challenge with BVDV-1 requires further investigation.

Results from North American studies indicate that weaning at least several weeks prior to

sending cattle to the feedlot is beneficial (Macartney et al., 2003; Step et al., 2008), but these

associations may have been confounded by several management practices occurring

simultaneously such as the administration of vaccine, bunk feeding and commingling of

animals. In the current study, we have extended knowledge by comparing weaning methods

and separately estimating the effects of several different components of pre-conditioning

programs. Combined with our previously reported results regarding the effects of prior

mixing, saleyard exposure, group dynamics and the timing of the move to the feedlot (Hay et

al., 2014), these results add important information to the body of knowledge about the

beneficial effects of prior management practices.

Based on our a priori hypothesis, we expected older animals to be at reduced risk of BRD,

because they would be expected to be less likely to be immunologically naïve to the major

respiratory pathogens due to the increased opportunity for exposure compared to younger

cattle. Crude incidences of BRD differed markedly between age categories with the highest

crude incidence in cattle aged 16 to < 22 months and the lowest crude incidence in the

youngest cattle (< 16 months). The increased risk of BRD in animals older than 22 months at

induction compared to younger animals, as indicated by the adjusted odds ratio estimates was

unexpected. Comparisons of results from models with and without each of random effects,

source region and season of induction indicated that the factors associated with the random

effects were the major confounders of the relationship between age and BRD. Of the random

effects, the feedlot-level component was the largest. The effect of age on BRD risk persisted

in the direct effects model indicating that the effect was over and above the effects of our

postulated intervening variables (i.e. induction weight, mixing history, saleyard exposure).

15

This association could have been confounded by unmeasured factors that resulted in cattle

entering the feedlot at a later age because they had been slower to attain the desired weight

for age (e.g., factors causing lower growth rates, including nutritional, genetic,

immunological and parasitic infectious factors) or because of differences in immune status at

induction depending on the timing of when cattle were previously mixed. The observation

that these animals’ median induction weight was similar to that of the younger animals in the

reference category provided some evidence that their growth rates prior to induction may

have been lower than animals entering feedlots at younger ages but unfortunately, body

condition scores were not available for the cattle in the study. Many lower-level exposures in

the study population were clustered at the feedlot level. For example, feedlot-level

management decisions regarding purchasing of cattle would be expected to result in feedlot-

level clustering of weight for age (an animal-level exposure). The observed association may

have also been subject to bias due to measurement error because we relied on retrospective

reports by vendors. In assigning average ages estimated at the group level to individual

animals, there may have been further misclassification of individual animals’ ages. But to

explain our results, these errors would have to have been differential with respect to BRD (or

to other risk factors for BRD).

Although vendors of about 75% of the cattle were requested to complete the survey,

responses were received for only 31% of the NBRDI population. Vendors that completed the

questionnaire may have differed in important ways from those that did not return the

questionnaire. For example, prevalences of exposure to practices such as yard weaning and

prior vaccination in the target population may have been overestimated. However, because

non-response to the vendor questionnaire was likely to be a surrogate for other unmeasured

factors, any resulting bias would be expected to be minimal (Dohoo, 2014).

16

The results from the current study combined with our previously reported results indicating

the importance of prior mixing and group dynamics (Hay et al., 2014) highlight the

importance of prior management of cattle before they are placed at a feedlot in reducing risk

of BRD at the feedlot. In the Australian context, these findings could be used to provide an

industry-wide framework to direct efforts at substantially reducing the impact of BRD. Yard

weaning is already commonly practiced and educating producers about the advantages of

yard weaning, introduction of grain and prior vaccination against BVDV-1 and M.

haemolytica in reducing risk of BRD at the feedlot would be expected to be beneficial.

Conclusion

Findings from this study demonstrate that several prior management exposures are effective

in reducing risk of BRD after beef cattle enter feedlots. Each of the management-related

interventions, yard weaning, prior feeding of grain, prior vaccination against BVDV-1 and

prior vaccination against M. haemolytica reduce risk of BRD in feedlot cattle.

Acknowledgements

This study was supported by grant B.FLT.0224 from Meat and Livestock Australia with

matching funds provided by the Australian Government.

17

References

Browne, W.J., 2012. MCMC Estimation in MLwiN, Version 2.26. Centre for Multilevel Modelling, University of Bristol, Bristol.

Dohoo, I., Martin, W., Stryhn, H., 2009. Veterinary Epidemiologic Research. VER Inc Charlottetown, Canada.

Dohoo, I.R., 2014. Bias--is it a problem, and what should we do? Preventive Veterinary Medicine 113, 331-337.

Fulton, R.W., 2009. Bovine respiratory disease research (1983-2009). Animal Health Research Reviews 10, 131-139.

Hay, K.E., Ambrose, R.C.K., Morton, J.M., Horwood, P.F., Gravel, J.L., Waldron, S., Commins, M.A., Fowler, E.V., Clements, A.C.A., Barnes, T.S., Mahony, T.J., Effects of exposure to Bovine viral diarrhoea virus 1 on risk of bovine respiratory disease in Australian feedlot cattle. Preventive Veterinary Medicine. http://dx.doi.org/10.1016/j.prevetmed.2016.01.025

Hay, K.E., Barnes, T.S., Morton, J.M., Clements, A.C.A., Mahony, T.J., 2014. Risk factors for bovine respiratory disease in Australian feedlot cattle: Use of a causal diagram-informed approach to estimate effects of animal mixing and movements before feedlot entry. Preventive Veterinary Medicine 117, 160-169.

Hay, K.E., Barnes, T.S., Morton, J.M., Gravel, J.L., Commins, M.A., Horwood, P.F., Ambrose, R.C., Clements, A.C.A., Mahony, T.J., Associations between exposure to viruses and bovine respiratory disease in Australian feedlot cattle. Preventive Veterinary Medicine. http://dx.doi.org/10.1016/j.prevetmed.2016.01.024

Hay, K.E., Morton, J.M., Mahony, T.J., Clements, A.C.A., Barnes, T.S., Associations between animal characteristic and environmental risk factors and bovine respiratory disease in Australian feedlot cattle. Preventive Veterinary Medicine. http://dx.doi.org/10.1016/j.prevetmed.2016.01.013

Kahn, L., Cottle, D., 2014. Beef Cattle Production and Trade. CSIRO Publishing, Melbourne. Larson, R.L., Step, D.L., 2012. Evidence-based effectiveness of vaccination against Mannheimia

haemolytica, Pasteurella multocida, and Histophilus somni in feedlot cattle for mitigating the incidence and effect of bovine respiratory disease complex. Veterinary Clinics of North America - Food Animal Practice 28, 97-106.

Macartney, J.E., Bateman, K.G., Ribble, C.S., 2003. Health performance of feeder calves sold at conventional auctions versus special auctions of vaccinated or conditioned calves in Ontario. Journal of the American Veterinary Medical Association 223, 677-683.

Miles, D.G., 2009. Overview of the North American beef cattle industry and the incidence of bovine respiratory disease (BRD). Animal Health Research Reviews 10, 101-103.

Roeber, D.L., Speer, N.C., Gentry, J.G., Tatum, J.D., Smith, C.D., Whittier, J.C., Jones, G.F., Belk, K.E., Smith, G.C., 2001. Feeder Cattle Health Management: Effects on Morbidity Rates, Feedlot Performance, Carcass Characteristics, and Beef Palatability. The Professional Animal Scientist 17, 39-44.

Sackett, D., Holmes, P., Abbott, K., Jephcott, S., Barber, M., 2006. Assessing the economic cost of endemic disease on the profitability of Australian beef cattle and sheep producers. Final Report: Project AHW.087. Sydney.

Sanderson, M.W., Dargatz, D.A., Wagner, B.A., 2008. Risk factors for initial respiratory disease in United States' feedlots based on producer-collected daily morbidity counts. Canadian Veterinary Journal 49, 373-378.

Schipper, C., Church, T., Harris, B., 1989. A review of the Alberta certified preconditioned feeder program (1980-1987). The Canadian Veterinary Journal 30, 736-741.

Step, D.L., Krehbiel, C.R., DePra, H.A., Cranston, J.J., Fulton, R.W., Kirkpatrick, J.G., Gill, D.R., Payton, M.E., Montelongo, M.A., Confer, A.W., 2008. Effects of commingling beef calves from

18

different sources and weaning protocols during a forty-two-day receiving period on performance and bovine respiratory disease. Journal of Animal Science 86, 3146-3158.

Taylor, J.D., Fulton, R.W., Lehenbauer, T.W., Step, D.L., Confer, A.W., 2010. The epidemiology of bovine respiratory disease: what is the evidence for preventive measures? Canadian Veterinary Journal 51, 1351-1359.

Textor, J., Hardt, J., Knuppel, S., 2011. DAGitty A Graphical Tool for Analyzing Causal Diagrams. Epidemiology 22, 745-745.

Walker, K.H., Fell, L.R., Reddacliff, L.A., Kilgour, R.J., House, J.R., Wilson, S.C., Nicholls, P.J., 2007. Effects of yard weaning and training on the behavioural adaptation of cattle to a feedlot. Livestock Science 106, 210-217.

Wieringa, F.L., Curtis, R.A., Willoughby, R.A., 1976. The influence of preconditioning on plasma corticosteroid levels, rectal temperatures and the incidence of shipping fever in beef calves. The Canadian Veterinary Journal 17, 280-286.

Woods, G.T., Mansfield, M.E., Webb, R.J., 1973. A three year comparison of acute respiratory disease, shrink and weight gain in preconditioned an non-preconditioned Illinois beef calves sold at the same auction and mixed in a feedlot. Can. J. comp. Med 37, 6.

19

Figure 1: Causal diagram depicting pathways relevant for the determination of total and direct effects of putative risk factors for bovine respiratory disease (BRD) in feedlot cattle. a At least one vaccination against Bovine viral diarrhoea virus 1 administered at least two weeks prior to induction b At least one vaccination against Mannheimia haemolytica administered at least two weeks prior to induction c Prior feeding of grain d Prior feeding of conserved forage or supplements (e.g. molasses) e No cattle in the current study were exposed to saleyards during the interval day -27 to day 0

20

Table 1: Distributions and estimated odds ratios for the total and direct effects of estimated age at induction and total effect of prior vaccination as measured in vendor questionnaire subsets on the risk of bovine respiratory disease (BRD) by day 50a

Variable & category

Number of animals

(% of animals)

Crude BRD 50-day cumulative incidence (%)

Adjusted odds ratio

95% credible interval

Total effects Ageb (months)

< 16 1,598 (16.4) 12.5 1.0 0.7 to 1.3 16 to < 22 5,326 (54.7) 23.3 Ref. cat.

≥ 22 2,807 (28.9) 17.1 1.6 1.3 to 2.1

Direct effects Agec (months)

< 16 0.8 0.6 to 1.2 16 to < 22 Ref. cat.

≥ 22 1.6 1.2 to 2.1

Total effects Prior Bovilis MH® vaccinationd,e

No 6,840 (85.0) 19.2 Ref. cat. Yes 1,205 (15.0) 15.4 0.8 0.6 to 1.0

Prior Pestigard®

vaccinationd,f No 7,063 (87.8) 19.0 Ref. cat.

Yes 982 (12.2) 16.1 0.8 0.5 to 1.1 a Three-level mixed effects logistic models based on causal diagram (Figure 1) b Covariates for total effects model: Induction season, Source region, N=9,731 c Covariates for direct effects model: Cohort fill pattern, Number of animals in cohort, Number of animals in group-13, Induction weight, Dentition, Breed, Sex, Saleyard exposure Induction season, Induction year, Mixing summary, Timing of move to the feedlot, Source region, N=9,508

d No additional covariates; N=8,045 e At least one dose of Bovilis MH® vaccine (against Mannhaemia haemolytica) at least 14 days before induction f At least one dose of Pestigard® vaccine (against Bovine viral diarrhoea virus 1) at least 14 days before induction

21

Table 2: Distributions and estimated odds ratios for the total effects of yard weaning, prior feeding history and on-farm mixing on the risk of bovine respiratory disease by day 50 a

Variable & category

Number of animals

(% of animals)

Crude BRD 50-day cumulative incidence (%)

Adjusted odds ratio

95% credible interval

Total effects Yard weaningb

No 983 (20.4) 31.2 Ref. cat. Yes, <7 days 1,788 (37.0) 23.8 0.7 0.4 to 1.0 Yes, ≥7 days 2,059 (42.6) 13.0 0.7 0.5 to 1.0

Yard weaningb

No 983 (20.4) 31.2 Ref. cat. Yes 3,847 (79.7) 18.0 0.7 0.5 to 1.0

Direct effects Yard weaningc

No Ref. cat. Yes 0.7 0.5 to 1.0

Total effects Prior graind

No 3,082 (76.6) 24.9 Ref. cat. Yes 940 (23.4) 16.4 0.6 0.3 to 1.1

Prior conserved forage or supplementd

No 659 (16.4) 28.8 Ref. cat. Yes 3,363 (83.6) 21.7 1.5 0.6 to 3.3

On-farm mixinge

No 322 (6.4) 27.3 Ref. cat. Yes 4,711 (93.6) 20.9 1.1 0.7 to 1.6

a Three-level mixed effects logistic models based on causal diagram (Figure 1); models restricted to animals in the vendor-bred subset b No additional covariates in total effects model, N=4,830 c Covariates: Induction year, Source region, Prior grain, Prior supplementary feeding, N=3,789 d Covariates: Yard weaning, Induction year, Source region; N=3,789 e No additional covariates; N=5,033