D.R. Smith*, P. Fedorka-Cray+, K.V. Brock *,
T.E. Wittum++, P.S. Morley++, R. Mohan#, K.H. Hoblet++,
and L.J. Saif *
*Food Animal Health Research Program
+Midwest Area Animal Disease Center, USDA-ARS
++Department of Veterinary Preventive Medicine
#Ohio Department of Agriculture, Division of Animal Industry
Winter dysentery is a disease of cattle characterized by diarrhea suddenly affecting many adult individuals in a herd. The disease occurs worldwide in dairy and beef animals. Bovine coronavirus (BCV) has been recovered from the feces of affected animals, and rising BCV antibody titers have been demonstrated in affected herds. However, few controlled studies have been conducted to investigate fully the role of BCV in winter dysentery outbreaks. We conducted a case-control study over a two-year period to compare the odds of exposure to various infectious agents and management factors in herds affected with winter dysentery to herds that did not experience winter dysentery in the same year. Several herd exposures appeared to be strongly associated with the occurrence of winter dysentery based on univariate logistic regression odds ratio calculations. Odds ratios for less common diseases, such as winter dysentery, approximate the relative risk of disease occurrence for a given exposure level. For every 10% increase in the proportion of animals demonstrating a four-fold or greater rise in BCV antibody titer, the risk of winter dysentery increased over 15 times (P < 0.0001). Other potential risk factors for winter dysentery included increasing animal density and the use of manure handling equipment to mix or deliver feed.
Winter dysentery is a disease of dairy cattle familiar to dairy farmers and veterinarians. The disease is characterized by a sudden onset of diarrhea, rapidly affecting many adult animals in the herd. Affected animals rarely die, but losses are incurred from the dramatic decrease in milk production and lost body condition. The disease also occurs in adult beef cattle. No diagnostic test for winter dysentery currently exists. The disease syndrome is diagnosed by considering the clinical signs of affected animals and the herd outbreak history. Other diseases can appear similar to winter dysentery, including acute bovine virus diarrhea (BVDV) infections and salmonellosis. The role infection with these agents may play in winter dysentery is not clear.
Winter dysentery was first described in 1931 (Jones, 1931). Since that time, many researchers have investigated the role of various infectious agents in the disease. Currently, the causative agent of winter dysentery definitively is not known. Researchers from many parts of the world have found bovine coronavirus (BCV) in the feces of affected animals and demonstrated BCV antibody seroconversion during outbreaks of winter dysentery (Saif, 1990; Clark, 1993); however, only a few controlled investigations have been conducted (Koopmans et al., 1991; White et al., 1989; Van Kruiningen et al., 1991; Jactel et al., 1990). If the important risk factors for the development of winter dysentery could be determined, control measures possibly could be adopted to prevent outbreaks of the disease. Such control methods could include the enhancement of immunity to specific agents through vaccination or management modifications to minimize the risk of disease occurrence.
Other studies have identified management factors as potential risk factors for the occurrence of winter dysentery in herds, although the studies have not always been in complete agreement. In a survey study of farmer-diagnosed winter dysentery, researchers in New York reported increased risk of winter dysentery with increasing herd size and outbreaks in previous years (White et al., 1989). French researchers have reported greater risk in small herds with either greater or lesser animal density than recommended (Jactel et al., 1990). This investigation was undertaken to apply epidemiological principles and diagnostic methods to test the hypothesis that BCV exposure is an important risk factor for the development of winter dysentery and to identify other potential etiologic and management risk factors. In this paper we report the results of the univariate logistic regression analysis of risk factors associated with winter dysentery in 12 affected herds during the winters of 1992-1993 and 1993-1994.
Study Design. This observational study was conducted as a case-control study, comparing herds affected with winter dysentery to control herds in the same geographical area that had not had adult cow diarrhea that season. For every case herd, data were collected from two control herds. In case herds fecal samples and blood specimens were collected from approximately 10 animals identified by the owner as currently affected and approximately 10 animals not observed to be sick. In control herds, 10 animals were chosen by the same systematic selection process and the same specimens and data collected. Blood was collected at the initial visit and approximately 3 weeks later. Within 24 hours of collection, the fecal samples were aliquoted and frozen at -70oC for later viral studies or shipped on ice for bacterial culture. Serum was separated, aliquoted, heat inactivated (56oC for 30 minutes), and then frozen. Whole blood samples were immediately submitted for BVDV virus isolation. Production and health records were collected from case and control herds. For BVDV identification in the second year, 100 milliliters of bulk tank milk were collected for PCR in lieu of whole blood. A questionnaire on herd management practices was completed during an interview of the owners during the initial visit. The median time to the initial visit to case herds was 3 days from the onset of the outbreak; convalescent samples were collected an average of 24 days later.
Definition of Disease. For the purpose of this study, case herds were defined as follows:
In addition, at least one of the following conditions must have existed:
Case Selection. To increase awareness of the study, letters were sent to veterinarians involved in dairy practice in Ohio and to state Cooperative Extension agents. Information about the study was included in the Ohio DHIA newsletter, and many extension agents included information in county newsletters mailed to area farmers. To be included in the study, the herd must have been in Ohio, met the case definition, and the disease outbreak must have been early in its course.
Control Selection. After a case herd was identified, two control herds were randomly selected from a frame of dairies in the same area using a random number generator. If the case herd was on DHIA test, then the frame for control selection was the list of DHIA dairies of the same breed in the same area. Two case herds were not on DHIA test; their controls were chosen from a list of non-DHIA dairies in the same county provided by area veterinarians. Control herds must not have had any herd problem with diarrhea in adult cows in that winter season.
Diagnostic Assays. Bovine coronavirus exposure was determined from two diagnostic assays. The first method, a double sandwich antigen capture ELISA (Smith et al., 1995), detected the presence of BCV antigens in the feces. The second, a serum BCV IgG ELISA, measured the antibody response from paired serum samples. A significant rise in antibody titer was defined for each individual as a four-fold (two dilutions) rise in antibody titer from acute and convalescent serum samples.
Herd exposure to BVDV was determined by one of two methods to identify the presence of virus. A serologic assay was used to identify the virus neutralizing antibody responses in paired sera. Virus isolation was conducted on whole blood buffy coat cells from each individual in the first 27 herds. The final nine herds were tested for BVDV using a bulk tank milk sample collected at the first farm visit and a polymerase chain reaction assay. A virus neutralization assay was conducted on each paired serum sample.
Herd exposure to Salmonella spp. was determined by selective bacterial culture of the feces of each individual. Positive salmonella cultures were serotyped for further identification. Exposure status for each herd to any infectious agent was expressed as 10 times the proportion of positive individuals to individuals tested for each herd.
Statistical Analysis. Univariate odds ratios were calculated for each variable using logistic regression. Conditional logistic regression, which considers matching in the study design, was also used to calculate odds ratios for those variables representing herd exposure to infectious agents.
Twelve case herds and 24 control herds were studied. The case herds ranged in size from 25 to 225 adult animals; control herds ranged from 15 to 202. Most herds were Holstein breeding, but half of the cows in one case herd were Jerseys, and another case herd had several breeds represented. One control herd was a Jersey herd. Samples for infectious agent diagnostics were collected from 469 animals. The attack rate in case herds ranged from 15 to 100%.
Bovine viral diarrhea virus was not isolated from the buffy coat cells of any individuals in any of the 27 herds tested, and only one herd (a case herd) of the nine tested was positive by PCR. Therefore, these results were not included in the logistic analysis.
The univariate odds ratios, with 95% confidence intervals, are summarized in Figure 1. Most animals in case and control herds had measurable antibody titers to BCV in the acute samples. However, in some animals the acute BCV antibody titer was below the level of detection. In these cases the titer was reported as the inverse of of the lowest sample dilution tested. The odds ratio for 10% increase in prevalence of the herd that demonstrated a > 4-fold rise in BCV antibody titer was 15.48 (P < 0.0001). Cow density expressed in the number of cows/100 square feet had an odds ratio of 6.36 (P = 0.043). Herds that used manure handling equipment to handle feed had an odds ratio of 5.5 (P = 0.0654).
By conditional logistic regression, the odds ratio for each 10% increase in the herd prevalence of animals with a > 4-fold rise in BCV antibody titer was 13.03 (P = 0.0006).
Figure 1. Univariate odds ratios with 95
percent confidence intervals showing, for each variable, the odds of
exposure in 12 winter dysentery herds to the odds of exposure in 24
non-affected herds. Odds ratios greater than 1 indicate an association
of the variable with winter dysentery, odds ratios less than 1 indicate
an association of the variable with no winter dysentery.
Herds were classified to exposure as a proportion (%) of the herd testing positive. This method was used as a means to fairly apply individual cow testing results to a herd level diagnosis. Because false positive test results can occur with any test, we have more confidence that a herd is truly exposed when a greater proportion of animals test positive. By assigning each herd a proportion positive value for the individual diagnostic tests, we give more weight to agent exposure in herds with many animals testing positive.
Odds ratios represent the odds of exposure in diseased herds compared to the odds of exposure in control herds. For diseases that occur infrequently, such as winter dysentery, the odds ratio is a reasonable estimate of the relative risk of a disease, comparing different exposure levels. The odds ratio for each 10% increase in prevalence of animals showing four-fold or greater increases in BCV antibody titers was 15.48, suggesting that for every 10% increase in the prevalence of animals demonstrating an antibody response to BCV in a herd, the risk for winter dysentery in a herd increased over 15 times. This ratio is a strong magnitude of effect and adds to the body of evidence implicating BCV as an important etiologic agent in winter dysentery. Similarly, an increase in cow density of one cow per 100 square feet increased the risk of winter dysentery occurring in a herd by over 6 times. The risk of winter dysentery occurring in a herd was 5.5 times greater if the herd used manure handling equipment (e.g., skid steer loaders) to handle feed.
Herds were matched by time and geographical location, because it was possible that there may have been a clustering of winter dysentery in different locations at different times, possibly due to different infectious agents (White et al., 1989). Conditional logistic regression considers the matched study design in the analysis. A disadvantage of matching, when the matching variable is not related to the exposure and to the disease, is the potential to lose power to measure the true level of effect. However, the odds ratio estimations for the prevalence of BCV titer increases were similar using either logistic regression or conditional logistical regression, indicating that the matching variable was independent of the disease status given these variables (Breslow et al., 1978). Therefore, the matching was disregarded, and the odds ratios were estimated by logistic regression.
The results of the univariate analysis of this case-control study suggested a strong association of the prevalence of BCV exposure, as evidenced by > 4-fold rising antibody titers, with the occurrence of winter dysentery in a herd. Other risk factors, such as increasing animal crowding and the use of manure handling equipment to mix or deliver feed, also appeared to be associated with winter dysentery in this study. Taken together, these risk factors make sense in terms of what we know about the biology of BCV. Because BCV is transmitted by both fecal-oral and respiratory routes (Saif et al., 1986), the spread of infection could be enhanced by both greater crowding of animals and manure contamination of their feed. Multivariate modeling, which considers potential confounding of the variables, may further clarify the role of these exposure variables to winter dysentery.
Breslow, N.E., N.E. Day, K.T. Halvorsen, R.L. Prentice, and C. Sabai. 1978. Estimation of multiple relative risk functions in matched case-control studies. Am. J. Epidemiol. 108:299.
Clark, M.A. 1993. Bovine coronavirus. Br. Vet. J. 149:51.
Jactel, B., J. Espinasse, M. Viso, and H. Valiergue. 1990. An epidemiological study of winter dysentery in fifteen herds in France. Vet. Res. Comm. 14:367.
Jones, F.S., and R.B. Little. 1931. The etiology of infectious diarrhea (winter scours) in cattle. J. Exp. Med. 53:835.
Koopmans, M., L. van Wuijckhuise-Sjouke, Y.H. Schukken, H. Cremers, and M.C. Horzinek. 1991. Association of diarrhea in cattle with torovirus infections on farms. Am. J. Vet. Res. 52:1769.
Smith, D.R., H. Tsunemitsu, R.A. Heckert, and L.J. Saif. 1995. Evaluation of two antigen capture ELISAs using polyclonal or monoclonal antibodies for the detection of bovine coronavirus. J. Vet. Diagn. Invest. 8:99.
Saif, L.J. 1990. A review of evidence implicating bovine coronavirus in the etiology of winter dysentery in cows: An enigma resolved? Cornell Vet .80:303.
Saif, L.J., D.R. Redman, P.D. Moorhead, and K.W. Theil. 1986. Experimentally induced coronavirus infections in calves: Viral replication in the respiratory and intestinal tracts. Am. J. Vet. Res. 47:1426.
Van Kruiningen, H.J., V.P. Castellano, A. Torres, and R.L. Sharpee. 1991. Serologic evidence of coronavirus infection in New York and New England dairy cattle with winter dysentery. J. Vet. Diagn. Invest. 3:293.
White, M.E., Y.H. Schukken, and B. Tanksley. 1989. Space-time clustering of, and risk factors for, farmer-diagnosed winter dysentery in dairy cattle. Can. Vet. J. 30:948.