Prevalence of antibodies to bluetongue virus and Anaplasma marginale in Montana yearling in Montana yearling cattle entering Alberta feedlots: Fall 2001

Abstract

A serologic survey was conducted in yearling cattle imported into Alberta feedlots from Montana during October 2001 to estimate the prevalence of antibodies to bluetongue virus (BTV) and Anaplasma marginale in Montana yearling cattle. The apparent prevalence of antibodies to BTV when the competitive enzyme-linked immunosorbent assay (cELISA) was used was 0.37% (21/5608). Test positive cELISA samples were also all positive when tested by virus neutralization (VN) and they reacted to 1 or more BTV serotypes, including 2, 10, 11, 13, and 17.

The apparent prevalence of antibodies to A. marginale when a recombinant cELISA (rcELISA) was used with a positive cutoff at 30% inhibition was 1.93% (108/5608). When the rcELISA positive cutoff was at 42% inhibition, the apparent prevalence was 0.73% (41/5608). After the reported sensitivity and specificity of the test had been accounted for, the A. marginale antibody results were consistent with a population that was either free of exposure or had a very low prevalence for A. marginale.

Introduction

Bluetongue viruses (BTV) and Anaplasma marginale are widely distributed in the United States (US), but they are most prevalent in southern and western States (1). The diseases associated with these agents are uncommon in northern States, such as Montana, Idaho, Washington, and North Dakota (1). With the exception of the Okanagan Valley, Canada is considered to be BTV free and A. marginale free, based on serosurveillance of cows and bulls at slaughter (2). Canada attempts to maintain its health status by imposing test requirements, or by requiring certain disease management activities for imported animals, based on the geographic source of the cattle and time of the year for the importation, or by both. Cattle from areas of the US considered to be “low” risk during the nonvector season enter Canada under a specialized import program called the Restricted Feeder Program (RFP) (3). Under the RFP, the import constraints during the summer and the post entry import protocols in the fall and winter are an added cost to Canadian and US cattle producers.

At present, the effect that year-round importation of US feeder cattle from “low” risk states during the summer vector season would have on the probability of transmission of BTV or A. marginale to Canadian cattle, sheep, and wildlife populations is not known with certainty. A terminal feeder (TF) risk assessment was conducted by the Canadian Food Inspection Agency (CFIA) in October 2001 to predict disease risks (4), the results predicted that importation of US feeders from “low” risk states during the summer would result in 1.4 outbreaks of bluetongue per year and 1 outbreak of anaplasmosis every 11 y (4). The risk assessment relied on seroprevalence data for Montana from 1978 to the 1990s for BTV (3.9%; 63/1605) and from 1992 to 1994 for A. marginale (0.27%; 45/16, 680) (4). These data relied on older serologic tests with poorer sensitivity and specificity. Additionally, the data were from cows, which likely have a higher seroprevalence than do calves and yearlings. The survey methodology was not clearly reported, making it difficult to assess the representativeness of the data. Clinical bluetongue has not been observed in Montana for more than 35 y (personal communication, Montana Department of Livestock) and this suggests that the occurrence of bluetongue is extremely rare or nonexistent.

The objective of this study was to provide current and representative prevalence estimates of antibodies to BTV and A. marginale in yearling cattle from Montana for use in updating the TF risk assessment to reflect current overall risks of disease following importation of yearling cattle from Montana into Canadian terminal feedlots during the summer, and to determine if the current post entry tetracycline treatment for anaplasmosis control could be eliminated from the RFP import protocols without substantial increase in risk.

Materials and methods

Target population

The target population was yearling cattle from Montana.

Sampling frame

Feeder cattle from Montana are imported into feedlots in Alberta under the RFP and can be available for serologic testing during arrival processing. In collaboration with the CFIA, the head of cattle to be sampled was set at 15 000. This sample size would allow the detection of a prevalence of 0.02% with 95% confidence, based on the assumptions of simple random sampling (5). The 0.02% prevalence is the cutoff value that the CFIA uses for health equivalency under the Health of Animal’s regionalization regulations (6). In addition, at the request of the CFIA, the sampling was to be conducted over a 3-year period in order to account for possible year to year variations in prevalence.

To ensure a representative sample from across ecologic regions (ecoregion) in Montana, sampling was weighted and stratified by ecoregion (Table 1). Montana was divided into 6 ecoregions, which were defined by characteristics based on United States Department of Agriculture (USDA) plant hardiness zone maps, altitudes, temperature, climatic conditions, and geographic/physical characteristics that may affect the habitat of vectors (Figure 1). The January 2001 beef cow inventory in Montana from the USDA National Agricultural Statistics Service for each ecoregion was used to allocate the overall sample size (Table 1).

Table 1.

January 2001 beef cow population in Montana by ecologic region (ecoregion)

Ecoregion	January 2001 beef cow inventory	Population (%)	Sampling allocation
1	85 600	6	300
2	188 300	12	600
3	295 000	19	950
4	81 000	5	250
5	499 200	33	1650
6	381 900	25	1250
Total	1 531 000	100	5000

Figure 1.

Six ecologic regions defined within Montana, based on habitat for bluetongue and anaplasmosis vectors.

Sampling was restricted to yearling cattle that arrived from Montana in October 2001 under the RFP and would be slaughtered by March 2002 (Table 2). All members of the Alberta Cattle Feeders’ Association were contacted by facsimile to request their participation in the survey. Nine feedlots indicated that they would be importing yearling cattle from Montana in October 2001 and all 9 participated in the survey. These feedlots together have a one-time holding capacity of approximately 300 000 head of cattle.

Table 2.

Comparison of the percent distribution of beef cows from Montana, Restricted Feeder Program (RFP) imports, October imports, tested overall, and tested in populations in Alberta by ecologic region (ecoregion)

	Ecoregion
Group	1	2	3	4	5	6	Number of animals
Montana beef cows	6	12	19	5	33	25	1 531 000
All RFP importsa	4	33	7	1	42	13	49 287
October importsb	3	34	8	0	3	12	46 843
Tested overallc	5	18	17	5	32	22	5608
Tested in Albertad	6	20	15	0	35	24	5163

^aYearling cattle imported into Alberta from Montana under the RFP from October 1, 2001, to March 31, 2002

^bYearling cattle imported into Alberta under the RFP during October 2001

^cYearlings from Montana tested in Alberta and Montana

^dYearlings from Montana tested in Alberta

The feedlots provided an estimate of the number of yearling cattle that they would be importing each week during October. Based on this weekly prediction, the number of working days, and the number of feedlots, a weighted estimate was made of the number of cattle that had to be sampled from each feedlot at each feedlot visit. For example, feedlot A estimated importing 10 000 head out of a total of 50 000 (20%) imports in October; 20% × 5000 target sample size over 3 feedlot visits = 333 test animals per visit. This weighting was done, since it was not known if feedlots purchased different types of cattle and whether this might influence the findings.

Only 1 team was available to collect blood samples from the participating feedlots. Thus, only 1 feedlot was visited each sampling day, except on 1 occasion on October 18, when 1 feedlot was visited in the morning and another feedlot was visited in the afternoon.

Selection method

Sampling was a combination of probability sampling in the 1st wk and nonprobability sampling in the following 3 wk. In the 1st wk, feedlots were visited on a systematic rotated basis; for example, feedlots A, B, C, A, B, C for Monday, Tuesday, Wednesday, Thursday, Friday, Saturday, and so on. Cattle were systematically selected for sampling, based on the number of yearling cattle from Montana to be processed that day and the required number of samples from that feedlot that week. For example, if the daily sample size was 300 animals and the feedlot was processing 600 head, then every other animal in the chute was bled.

During the 2nd wk, following the initial rush of imported cattle, feedlots were visited if they were processing yearling cattle from Montana and this resulted in a different feedlot being visited each day. The cattle were systematically sampled in the chute as described above. Following week 2, the number of cattle bled per ecoregion was tabulated and compared with the sample allocations by ecoregion (Table 1). Shortages in the target number of cattle to bleed per ecoregion were identified. In the subsequent 2 wk, feedlots were contacted for cattle originating from those ecoregions from which there were shortages of cattle to meet the allocation. Technicians went to the feedlot when the feedlot was processing yearling cattle from those ecoregions and cattle were systematically sampled until the desired allocation was met.

At the end of October, no cattle had been sampled in ecoregion 4 and the study was short of 179 head in ecoregion 3. Beef herd managers in Montana were contacted for potential on-farm sampling of animals in ecoregions 3 and 4 to reach the allocation per ecoregion. Yearling heifers selected as a convenience sample from 4 Montana herds in ecoregion 3 and from 4 Montana herds in ecoregion 4 were bled to reach the desired allocation.

Serologic testing

Serum samples were sent by courier to the Montana Veterinary Diagnostic Laboratory in Bozeman, Montana. A commercial competitive enzyme-linked immunosorbent assay (cELISA) was used in testing for antibodies to BTV (Veterinary Medical Research and Development [VMRD], Pullman, Washington, USA). Test sera were positive if they produced an optical density of less than 50% of the negative control. Samples that were positive for BTV with the cELISA were sent to the USDA: APHIS:VS National Veterinary Services Laboratories in Ames, Iowa, where virus neutralization (VN) testing was performed.

A commercial recombinant cELISA (rcELISA) was used in testing for antibodies to A. marginale (VMRD, Pullman) (7). Test sera were declared positive with a percent inhibition ≥ 30%. All samples that were positive for A. marginale with the rcELISA on first screening were sent to VMRD and retested as a cross-check. A confirmatory nested polymerase chain reaction (PCR) test could not be done on the rcELISA positives because whole blood was not available.

Statistical analysis

The apparent prevalence was calculated as the number of test positive samples divided by the total number of samples tested. The reported test sensitivity and specificity from VMRD for the BTV cELISA is 99%. The reported test sensitivity from VMRD for the A. marginale rcELISA is 94.8% and the specificity is 97.6%, if a 30% cutoff value for test positives is used. The level of uncertainty (standard error [sχ̄]) in these reported test characteristics was not reported by the laboratory.

The apparent prevalence for the cELISA tests was adjusted for the reported test sensitivity and specificity by the following standard equation (5): Adjusted

For A. marginale, cutoff values of both 30% inhibition , and 42% inhibition were used.

Anaplasma exposure simulation modeling

If the apparent prevalence is lower than (1.00-specificity of the test), as it was in the case of the A. marginale results, the standard equation (5) can return a negative value for the adjusted prevalence. Essentially, the failure of the equation to transform the apparent prevalence into a true prevalence is caused by the occurrence of fewer test positives than would have been expected, on average, given the stated fixed specificity value.

Retesting of the cELISA positive BTV sera with the VN test was possible to eliminate false positives. However, retesting was not possible for A. marginale test positive sera, which leaves open to question the true prevalence of A. marginale in the imported cattle.

To address the issue of the imperfect A. marginale test, a simulation model was constructed to allow for incorporation of some variability in the A. marginale test characteristics. The modeling approach used a Monte Carlo simulation of sampling both a hypothesized “disease free” population and an infected population. The infected population had a prevalence that ranged from 0.0025 to 0.03, while the “free” population had a prevalence of 0. Prevalence variation was modeled by using a Beta-pert distribution, which is defined by the minimum, maximum, and most likely values. The prevalence range represented the most likely value. The maximum and minimum values were calculated, respectively, as the most likely value ± 0.6 times the most likely value (Table 6). Beta-pert distributions were also used to model variation in sensitivity (minimum = 0.94, maximum = 0.990, most likely = 0.948) and specificity (minimum = 0.970, maximum = 0.999, most likely = 0.976). For simplicity, the model assumed a random sample of the population, thus ignoring the stratified design of the sampling protocol. Ten thousand simulations were run at each prevalence level to develop a distribution of positive test results for the infected and “free” populations. The probability of observing results at least as extreme as was observed in the serological survey was determined from these distributions. Additionally, the ratio of the tail areas of these distributions (infected/“free”) was calculated to provide the likelihood that the results came from the infected population.

Table 6.

Anaplasma marginale modeling results comparing the probability of observing the survey results (≤108 positives) given an infected population with the probability of observing the survey results given a “free” population

Disease prevalence
Minimum	Most likely	Maximum	Probability of ≤ 108 positives	Probability of ≤ 108 positives given “freedom”	Likelihood ratio of Probability of probability with disease divided by probability without disease
0.001	0.0025	0.004	22.7	34.6	0.66
0.002	0.005	0.008	13.84	35.1	0.39
0.003	0.0075	0.012	7.3	34.8	0.21
0.004	0.01	0.016	3.82	34.9	0.11
0.005	0.0125	0.02	1.8	34.7	0.05
0.006	0.015	0.024	0.9	35.3	0.03
0.007	0.0175	0.028	0.046	35.3	0.00
0.008	0.02	0.032	0.018	35	0.00
0.009	0.0225	0.036	0.008	34.6	0.00
0.01	0.025	0.04	0.004	35.1	0.00
0.011	0.0275	0.044	0	35	0.00
0.012	0.03	0.048	0	35.2	0.00

Clustering of disease was explored using the analysis of variance (ANOVA) estimator of the intracluster correlation (14). The number and size of the beef herds in Montana from which the imported yearling cattle originated were not known. Brands on the imported cattle could not be related to USDA eartags, so there was no way to assess clustering by herd. Instead, clustering was evaluated by using the processing lot as the cluster, since similarly sourced cattle arrive together in truckloads and make up processing lots

Results

Bluetongue

The overall apparent prevalence of antibodies to BTV with the cELISA was 0.37%, varying from 0% to 0.83% among the 6 ecoregions (Table 3). During the 1st wk of random sampling, the prevalence was 0.37% (7/1902, 95% CI = 0.15% to 0.76%). When the cELISA results were adjusted for reported sensitivity and specificity by using the standard equation (5), the adjusted prevalence was 0%.

Table 3.

Antibodies to bluetongue virus determined by competitive enzyme-linked immunosorbent assay (cELISA) in cattle from different ecologic regions (ecoregion) of Montana

Ecoregion	Number tested	Apparent prevalence
1	299	0%
2	1021	0.20%
3	953	0%
4	266	0%
5	1815	0.83%
6	1254	0.32%
Total	5608	0.37%

All 19 cELISA positive samples that were available for further testing were VN positive (Table 4). There was reactivity to multiple serotypes (2, 10, 11, 13, and 17) in the samples. Two samples were positive for 1 serotype, 2 samples were positive for 2 serotypes, 1 sample was positive for 3 serotypes, 13 samples were positive for 4 serotypes, and 1 sample was positive for 5 serotypes.

Table 4.

Number of bluetongue virus (BTV) competitive enzyme-linked immunosorbent assay-positive (cELISA) animals (n = 19) that tested positive for BTV serotypes on virus neutralization (VN)

BTV Serotype	1:10 VN titer	1:20 VN titer	1:40 VN titer	1:80 VN titer
2	5	6	4	0
10	1	5	7	4
11	3	4	5	4
13	2	5	5	5
17	0	1	0	0

Anaplasma marginale

With a cutoff of 30% inhibition, the overall apparent prevalence of antibodies to A. marginale was 1.93%, varying from 0% to 3.78% among the 6 ecoregions (Table 5). When the cutoff of 42% inhibition was used, the apparent prevalence decreased to 0.73%, varying from 0% to 1.47% in the 6 ecoregions. During the 1st wk of random sampling, the apparent prevalence at 30% inhibition was 3.2% (61/1902); at 42% inhibition it was 1.1%. The adjusted prevalence in week 1 at 30% inhibition was 0.87% (95% CI: 0.5% to 1.4%) and at 42% inhibition was 0%.

Table 5.

Prevalence of antibody to Anaplasma marginale in cattle from different ecologic regions (ecoregion) of Montana determined by the recombinant competitive enzyme-lined immunoassay

Ecoregion	Number tested	Apparent prevalence 30% cutoff	Apparent prevalence 42% cutoff
1	299	0.87%	0.33%
2	1021	1.86%	0.29%
3	953	3.78%	1.47%
4	266	0%	0 %
5	1815	1.93%	0.88%
6	1254	1.28%	0.56%
Total	5608	1.93%	0.73%

The overall adjusted prevalence for A. marginale based on the standard equation (5) for either cutoff was less than 0%, indicating that the test characteristics used to adjust the prevalence were incorrect.

The results of the simulation modeling of a sample of 5000 animals indicated that there was a 35% probability of observing 108 or fewer positive results for A. marginale, given that there was no disease in the population (Table 6, Figure 1). Conversely, there was a 65% probability that there would be 109 or more positive tests from sampling in a disease free population. At a low prevalence of 0.25%, the probability of observing 108 or fewer test positives was only 23%; conversely, 77% of the time 109 or more positives would be expected from a population with this prevalence. The probability of the observed results with respect to prevalence in the sampled population is shown in Figure 2.

Figure 2.

Results of simulation when prevalence levels of the alternative hypothesis are varied.

The intracluster correlation coefficient (rho) for BTV was 0.0185 and for A. marginale it was 0.0142 at the 30% inhibition cutoff and 0.0059 at the 42% inhibition cutoff.

Discussion

Bluetongue

Since all positive cELISA tested positive on VN, it suggests that the specificity of the cELISA is higher than reported and probably close to 100%, if it is assumed that the VN test is the “gold standard” and has a 100% sensitivity and specificity. Although the sensitivity and specificity of the VN test is unknown, it is assumed to be very high.

Cross reactivity is common among BTV serotypes when the VN test is used (personal communication, Dr. William C. Wilson, USDA, Agricultural Research Services, Laramie, Wyoming, USA). It may indicate exposure to multiple serotypes, cross reaction from common epitopes among genetically related serotypes, or exposure to another virus with common epitopes, such as an epizootic hemorrhagic disease virus (EHDV) serotype. Bluetongue virus serotypes 10, 11, and 13 are related genetically, as are serotypes 2 and 17.

The prevalence results from this study are much lower than the 3.9% (63/1605) reported from 1978 to 1990 for cattle from Montana, based on agar gel immunodiffusion (AGID) test seropositivity (3,4). The AGID test can cross-react with EHDV, epizootic hemorrhagic disease that has been observed in wildlife in Montana (1). Relative to the cELISA, the specificity of the AGID test is reported to be only 57.7% (3,4). A high rate of false positives may explain some of the difference noted between the previous estimate of prevalence and the results of this survey. The difference also may be explained by changes in disease dynamics over time. The age of the animals may also be a factor in explaining the difference in prevalence estimates, since cows were tested in the 1978 to 1990 survey and yearling cattle were tested in this study. It is believed that antibodies to BTV may persist for life; thus, cows would have more opportunity for exposure than yearling cattle due to their age, since the latter would have only 2 summers of potential exposure. Yearling cattle are the age group of interest in import feeder programs.

The very low prevalence estimate of 0.37% in these yearling cattle from Montana that were tested is consistent with the absence of clinical disease. If infection in cattle were present at a higher level, one would expect to see clinical signs of bluetongue in Montana, because there are large sheep and wildlife populations present (15).

Anaplasma marginale

The test manufacturer (VMRD) suggests a 30% inhibition cutoff for test positives. The CFIA conducted its serological study in Canadian cattle in 1998 using the VMRD test with a 42% inhibition cutoff (2,8). The specific impact of changing the test positive cutoff value on test sensitivity and specificity is unknown, although it is likely that sensitivity is decreased and specificity is increased. Thus, both test positive cutoff results were calculated and reported here.

The A. marginale results in this survey were consistent with a population that either was free of exposure or had a very low prevalence of infection. The previous A. marginale prevalence estimates that were used in the RFP and TF risk assessments were based on test results on exports from Montana from 1992 to 1994 (mean = 0.0027 with a 99% CI of 0.0018–0.0040) (3,4). Information regarding serological test procedures and sampling methodology was not provided. Until recently, the complement fixation (CF) test has been used in import/export programs and related serosurveillance, including the 1998 to 1999 survey that Canada carried out to maintain its disease freedom status (2). The CF test was recently reported to have a sensitivity of only 20% in carrier cattle (7), thus potentially resulting in false negatives. If the prevalence data for cattle from Montana from 1992 to 1994 is adjusted for the inadequacies of the CF test by using a sensitivity of 20% and a specificity of 98% (7) in the standard equation (5), the adjusted prevalence is less than zero. The findings from 1992 to 1994 are consistent with the results in this survey, which indicate either that the prevalence of anaplasmosis is very low or that the population is disease free.

Some potential biases may exist in this study that could affect whether the data are representative of cattle in Montana and inferences can be made to the cattle population of Montana. These biases are discussed below.

Sampling was restricted to October because of the need by the CFIA to have any test-positive cattle slaughtered before the next vector season in Alberta. A potential bias could exist if the cattle imported in October were different from those imported during the rest of the vector season. Based on a review of US health permits and brand certificates, 95% of the imported cattle from Montana (46 843/49 287) to Canada were imported in October (Table 2). There were no obvious differences in the imported cattle that might be related by time to disease risks that would be expected to influence the overall results.

If the cattle tested did not originate initially from Montana, then bias may exist. Trace-back of all the sampled cattle by the Montana Department of Livestock indicated that 95% (5323/5603) of the yearling cattle that were exported to and tested in Canada were born or raised in Montana. Two hundred and eighty-five cattle originated from Idaho under the 60-day residency allowance in the RFP and were sold in Montana. Given that the cattle from Idaho were present for some of the vector season for bluetongue and anaplasmosis in Montana, and given that they only represent 5% of the total sample, it is unlikely that they influenced the overall findings significantly. Since these cattle were mixed in loads of cattle originating from Montana and since USDA eartags were not correlated with Montana brands on health certificates, it was not possible to pull these 285 animals from Idaho from the test results. If these cattle from Idaho had a higher prevalence of antibodies to BTV or A. marginale, they would have shifted the estimates of prevalence in Montana upwards.

Some cattle in ecoregion 3 (179/953) and all cattle in ecoregion 4 (n = 266) were conveniently sampled in a limited number of herds in Montana, because they were not available to test in feedlots in Alberta. During October, only 158 yearling cattle (0.3%) from ecoregion 4 were imported into Alberta feedlots (Table 2) and they were not sampled at the feedlot, suggesting that they arrived during the first 2 wk of sampling, when ecoregion was not considered in sample selection. The cattle sampled in Montana in ecoregion 4 were from only 4 herds and thus may not be fully representative of ecoregion 4. It is hoped that cattle from more herds in that ecoregion will be imported in years 2 and 3 of the study, so that they can be tested on a systematic basis to reduce any potential bias that the convenience sampling from a few herds may have caused.

Clustering of disease among originating beef herds was not accounted for in this study. Clustering may affect the precision of the prevalence estimates (9). The relatively low intracluster correlation coefficient values for BTV and for A. marginale at the 30% and 42% inhibition offered little evidence of clustering within processing lots. Although these values were determined from the artificial shipping lot clusters, recent research conducted by USDA Animal Plant Health Inspection Services, VS Centers for Epidemiology and Animal Health, suggests that in low prevalence areas, such as is to be expected in Montana, clustering is not evident with bluetongue.

The largest constraint in this study was the practical inability to implement fully randomized sampling, which is necessary to evaluate sampling error and determine the reliability of the findings for inferences to the target population. A sampling frame for imported yearling cattle from Montana by ecoregion was not available prior to the initiation of the study.

The prevalence estimates for antibodies to BTV and A. marginale from this study appear representative of yearling cattle following possible summer exposure in Montana. Potential biases have been investigated and no large biases that would substantially affect the interpretation of the data have been identified. These data support clinical observations regarding these 2 diseases, either a lack of reports (bluetongue) or rare reports (anaplasmosis).

The current data should be very useful in updating the previous CFIA feeder risk assessments (3,4), since the data are current, are derived from the most sensitive and specific serologic testing procedures available, and represent the age of imported animals of interest. The prevalence data collected here can be used in the import feeder risk assessments, either through 1st order or 2nd order approaches (10,13). Both variability and uncertainty can be modeled by using these new data and the distributions of the variables included in the risk assessments. These newly acquired empirical data can be combined with existing information to improve the estimate of the prevalence of infection used to characterize the distribution through Bayesian inference calculation (10). To improve the value of the 1970 to 1990 historical BTV and A. marginale prevalence data, a Markov-chain model can be used to estimate the present value of the disease information collected in the past (12).