A case-control study of Nocardia mastitis in Nova Scotia dairy herds

A case-control study of Nocardia mastitis in Nova Scotia dairy herds Lyn Ferns, Ian Dohoo, Alan Donald Abstract A case-control study was conducted to identify herd production, housing, and hygienic and therapeutic factors associated with a diagnosis of Nocardia mastitis in dairy herds in Nova Scotia. The data were collected by on-farm interviews with owners of 54 case and 54 control herds. Logistic regression was used to study risk factors. The use of dry cow products containing neomycin, including two specific dry cow products, was strongly associated with a diagnosis of Nocardia mastitis in a herd. Other factors which increased the risk of Nocardia mastitis were higher levels of production, larger herd size, and a large percentage of cows treated with dry cow products. These results are compared to results from a similar study carried out in Ontario. Resume Etude de cas avoc groupos temoins de mammite A Nocardla chez les bovins laitiers en Nouvelle-Ecosse Une etude de cas avec groupes temoins a ete realisee pour identifier les facteurs therapeutiques, d'hygiene, d'hebergement et de regie associes au diagnostic de mammite a Nocardia chez les bovins laitiers en Nouvelle-Ecosse. Les donnees ont ete obtenues lors d'entrevues effectuees sur la ferme avec les proprietaires et regroupant 54 cas et 54 troupeaux temoins. Une analyse par regression logistique a ete effectuee pour etudier les facteurs susceptibles. L'utilisation des produits pour vache tarie renfermant de la neomycine, incluant deux produits specifiquement pour vache tarie, a ete fortement associee au diagnostic de mammite a Nocardia chez les troupeaux. Parmi les autres facteurs susceptibles d'augmenter le nombre des mammites a Nocardia, des taux de production plus eleves, des troupeaux de plus grande envergure et un pourcentage eleve d'animaux traites avec des produits pour vache tarie ont ete incrimines. Les resultats sont compares 'a ceux d'une etude semblable effectuee en Ontario. (Traduit par Dr Therse Lanthier) Can Vet J 1991; 32: 673-677 Nova Scotia Department of Agriculture, P.O. Box 550, Truro, Nova Scotia B2N 5E3 (Ferns) and Department of Health Management, Atlantic Veterinary College, University of Prince Edward Island, Charlottetown, Prince Edward Island CIA 4P3 (Dohoo, Donald) Funding was provided by the Nova Scotia Department of Small Business Development, Public Sector Employment Program, and the Nova Scotia Department of Agriculture and Marketing. Introduction Nocardia spp. were first reported as a cause of Ilmastitis in Canada in 1956 (1). A 1958 report established the pathogenicity of Nocardia asteroides (2), and it has been detected as a sporadic cause of mastitis since then. Outbreaks have been reported and attributed to contamination of drugs and equipment used for intramammary therapy (3-5). Nocardia spp. are widely distributed soil-borne organisms. In 1988 there was a sudden increase in the number of Nocardia isolations in milk samples cultured at the Veterinary Pathology Laboratory of the Nova Scotia Department of Agriculture and Marketing in Truro. A similar increase was seen in many other parts of Canada (6). In Nova Scotia there had been no recent changes in culture techniques or in the number of samples cultured when the increase was initially detected, although such changes did occur after mid 1988. This study was designed to identify herd characteristics such as production and hygienic and therapeutic factors which could be associated with Nocardia mastitis in Nova Scotia dairy herds, and to compare these findings to those obtained from a recent Ontario study (7). Materials and methods The herd was chosen as the unit for analysis. Case herds were initially selected from herds submitting milk samples for culture to the laboratory in Truro between January 1988 and June 1989. In this way, Nocardia was isolated from 60 herds, but nine herds were excluded from the study. Owners of five of the latter herds had recently purchased an affected cow prior to diagnosis of Nocardia mastitis, one herd no longer existed when the interviews were conducted, and three herds contained 12 cows or less. Control herds were selected randomly from a list of dairy producers in the same county using computergenerated random numbers. This list included only herds with more than 12 cows, thus the exclusion of smaller herds from the case herds. Sample size calculations (8) indicated that a study containing approximately 50 case and 50 control herds would have an 80% probability of detecting a 50% reduction in the prevalence of exposure among control herds compared to case herds, provided 50% or more of case herds were exposed. On this basis it was felt that it was worthwhile to pursue the study with the pool of case material available. Data were collected by means of a personal on-farm interview carried out by a single interviewer between May and August of 1989. The possibility of interviewer bias was minimized by hiring an interviewer who was not aware of the Nocardia mastitis outbreak at the time of hiring. The interviewer was asked to avoid reading 673

Table 1. Variables analyzed for associations with Nocardia caselcontrol status of the dairy herd AUTO BREED BSCC88 DI-D23 DBARN DBED DCABRUPT DCPCT DCPREP DOUT DRYTWL Ll-L18 MBARN MBED MILKSYS MOUT MP MPBED NUMCOW PLC88 PROD TDAFTER TDAPPL TDBEFOR TDCALV TDFREQ WASHDIS WASHTW Use of automatic takeoff units Main breed of cattle in the herd Average bulk tank SCC for first six months of 1988 Intramammary dry cow mastitis products used on the farm Type of barn housing dry cows Type of bedding used for dry cows Method of drying off cows at end of lactation Percentage of cows dry treated Method of preparation of teat ends prior to dry treatment Type of outdoor area available to dry cows during the summer Type of material used to dry cows' udders Intramammary lactating cow mastitis products used on the farm Type of barn housing milking cows Type of bedding used for milking cows Type of milking system Type of outdoor area available to milking cows during the summer Use of maternity pen for calvings Type of bedding used in maternity pens Number of cows milked Average plate loop count for first six months of 1988 Average milk production per cow per day (kg) Use of teat dip after milking Method of applying teat dip Use of teat dip before milking Use of teat dip before calving Frequency of changing teat dip in applicator Use of disinfectant in udder wash water Type of material used to wash cows' udders articles about Nocardia mastitis, and was not informed of rumors regarding the cause of the outbreak. At the start of the interview, farmers were specifically cautioned not to mention whether Nocardia mastitis had been diagnosed in their herd until the end of the interview. Farmers were also asked not to discuss their theories about the cause of the outbreak with the interviewer, although they did not always comply. The survey questions pertained to the variables listed in Table 1. Questions covered production, udder health, housing, milking system, milking sanitation techniques, dry cow therapy techniques, and intramammary antibiotic products used for dry and lactating cows. Management practice changes in the previous two years were noted. The time of change was recorded. The procedures used prior to the first diagnosis of Nocardia mastitis on the farms were recorded for analysis. Survey forms are available to readers on request. During the interview, the bulk tank was agitated, and, after the interview, a milk sample was collected from it with a sterile syringe. Occasionally, the interviewer could not be at the farm when the bulk tank was full, so the farmer was instructed in sampling, and sent in the sample. The milk was inoculated with a 0.01 ml loop onto trypticase-soy agar plates supple- 674 Table 2. Continuous variables associated (p c 0.15) with the Nocardia mastitis caselcontrol status of dairy herds Case herds Control herds Variable' Mean SD Mean SD p NUMCOW 54.9 36.2 43.3 21.0 0.000 PROD 24.2 3.2 22.6 4.5 0.021 BSCC88 ('000) 179.8 90.5 196.1 115.0 0.086 DCPCT 87.5 26.3 63.7 42.9 0.001 mented with 5%70 citrated sheep blood and gentamycin sulfate 25 mg/l (Gentocin, Schering Canada, Pointe Claire, Quebec). The plates were incubated for 48 hours at 35 C and, if negative for Nocardia spp., held for at least seven days at room temperature. Nocardia was cultured from 11 case herd bulk tank samples and three potential control herd samples. The latter herds were then considered as case herds, and new control herds were selected randomly. Nocardia was subsequently isolated from individual cows in two of these three herds. One herd did not submit any individual milk sample for testing. Two owners of herds initially selected as controls chose not to participate in the study. These were replaced by other randomly selected control herds. Fifty-four case herds and 54 control herds were ultimately included in the study. The data were analyzed using the SAS (9) and BMDP (10) statistical software packages. Differences between case and control herds with regard to continuously distributed variables were assessed using the Student's t-test. Contingency table analyses utilizing the chisquare statistic were used to assess differences between case and control herds with regard to categorical variables. All variables unconditionally associated with the case/control status of the herd at a p value of <0.15 were selected for inclusion in logistic regression analyses. All variables were entered initially into the logistic model and backwards elimination was used to remove variables not significant at p = 0.05. New variables based on the antibiotic composition of dry cow products used on each farm were created. These were DNEO (one or more neomycin-containing dry cow products used on the farm prior to the diagnosis of Nocardia mastitis), DCLOX (a cloxacillincontaining dry cow product used) and DOTH (dry cow products containing other antibiotics). Unconditional odds ratios for these variables were computed and then logistic regression analyses were repeated using these variables instead of individual dry cow product variables. Results A total of 12 of the variables examined originally were associated with the case/control status of the herd at p ' 0.15 (Tables 2 and 3). Case herds tended to contain more milking cows and have greater milk production per cow per day. Most of the management variables related to risk of Nocardia mastitis were associated with the dry cow treatment program in use on the farm. This treatment program included the method of teat end preparation (DCPREP), the percent

Table 3. Categorical variables associated (p c 0.15) with the Nocardia mastitis case/control status of dairy herds Number of Number of Odds Variable' case herds control herds p ratio BREED Holstein 53 49 0.093 5.4 Other 1 5 DOUT Pasture 51 50 0.135 N/A Yard/Drylot 0 3 None 3 1 AUTO Yes 13 7 0.137 2.1 No 41 47 DCPREP No preparation 1 1 0.091 N/A Wash only 3 2 Wash and disinfect 50 45 No dry cow used 0 6 D5 (Biotef) Yes 10 4 0.086 2.8 No 44 50 D8 (Co-op Dry Cow) Yes 27 11 0.001 3.9 No 27 43 D18 (Orbenin Dry Cow) Yes 8 19 0.015 0.3 No 46 35 L3 (Co-op Lactating) Yes 18 10 0.079 2.2 No 36 44 DNEO Yes 49 25 0.001 11.4 No 5 29 DCLOX Yes 8 19 0.015 0.3 No 46 35 DOTH Yes 26 34 0.121 0.6 No 28 20 N/A = Not applicable - odds ratios for variables with three or more categories have not been calculated because there is no obvious category to be used as a baseline of cows dry treated (DCPCT), and the products used (D5, D8, D18 and L3). The constructed variable DNEO was also significantly associated with the risk of Nocardia mastitis. The constructed variable DCLOX was identical to D18 in that there was only one cloxacillin-containing dry cow product on the market. None of the six herds in which dry cow therapy was not used, and only one of 13 herds in which less than 10% of the cows were given dry cow therapy, were case herds. The logistic regression analysis identified three dry cow products that were significantly associated with Nocardia mastitis (Table 4). D5 (Biotef, The Upjohn Company, Orangeville, Ontario) and D8 (CO-OP Dry Cow, Interprovincial Cooperatives Ltd., Mississauga, Ontario) were positively associated with Nocardia mastitis (odds ratios 3.09 and 2.54, respectively); D18 (Orbenin Dry Cow, Ayerst Laboratories, Saint Laurent, Quebec) had a negative association (odds ratio 0.37). The percentage of cows treated with dry cow therapy (DCPCT) had a slight but positive association with Nocardia mastitis, as did herd size (NUMCOW) and level of production (PROD) (increased risk of Nocardia Table 4. Important predictors of Nocardia mastitis identified by logistic regression Regression Odds Predictor' coefficient p ratiob PROD 0.161 0.045 1.17c NUMCOW 0.022 0.021 1.02 DCPCT 0.032 0.001 1.03 Biotef 2.068 0.008 7.91 CO-OP Dry Cow 2.202 0.001 9.04 Orbenin Dry Cow -2.561 0.000 0.08 bthe odds ratio (OR) and the regression coefficient are related by the formula OR = e(regression coefficient) CThe odds ratio for a continuous variable is the increased risk associated with an increase of one unit in the variable mastitis with greater proportion of cows treated with dry cow therapy, in larger herds, and in herds with higher levels of production). A further logistic regression analysis showed neomycin containing products to be positively associated with Nocardia mastitis (Table 5). The main effect of cloxacillin was not significant but there was a significant 675

negative interaction between neomycin and cloxacillin. The odds ratio associated with the use of neomycin containing products on farms that did not also use cloxacillin was 28.91. However, this was reduced to 1.58 if the farm had also used cloxacillin. As in the previous logistic regression analysis, the percent of cows treated with dry cow therapy (DCPCT), the herd size (NUMCOW), and the level of milk production (PROD) all had small positive associations with the risk of Nocardia mastitis. Discussion A personal on-farm interview by a single interviewer was chosen as the best means of getting cooperation, complete data, and results that were consistent and reliable. All possible case herds were included in the study in order to maximize the power of the study. Control herd status was verified by bulk tank culture. This resulted in three control herds being changed to case herds. While environmental contamination of the bulk tank milk is possible, it was considered unlikely since two of the herds subsequently had Nocardia mastitis confirmed in individual cows. The third herd owner did not submit any individual cow samples for culture. The sensitivity of bulk tank culture for Nocardia is questionable, although it has recently been reported to be approximately 750/o if multiple samples are tested (11). Consequently, it is possible that some control herds did contain cows infected with Nocardia. This small possibility of misclassification bias would be unlikely to have a major effect on the strong associations found in the analyses (12,13). Five case herds which admitted milking Nocardia-infected cows had negative bulk tank cultures. This would not affect survey results, however, as these herds were already known to have had Nocardia mastitis. Variables were included in this study because of their accepted relationship with risk of mastitis in general or because of their hypothesized association with the risk of Nocardia mastitis (7). Since the outcome of interest was the risk of a herd developing Nocardia mastitis in one or more cows, it was important to deter- 676 Table 5. Effect of neomycin-containing and cloxacillin-containing intramammary dry cow treatments on Nocardia mastitis diagnosis in a herd as determined by logistic regression analysis Regression Odds Predictore coefficient p ratiob PROD 0.158 0.040 1.17c NUMCOW 0.019 0.040 1.02 DCPCT 0.032 0.001 1.03 Neomycin 3.364 0.000 28.91 Cloxacillin 0.172 0.875 1.19 Interactiond - 3.080 0.022 0.05 bthe odds ratio (OR) and the regression coefficient are related by the formula OR = e(regression coefficient) CThe odds ratio for a continuous variable is the increased risk of Nocardia mastitis associated with an increase of one unit in the variable dsee text for an explanation of the meaning of the interaction between neomycin and cloaxcillin mine which factors were involved prior to the first diagnosis of Nocardia mastitis in the herd. To this end, the time of any change in a management practice was noted and compared to the date of the first diagnosis of Nocardia mastitis. Odds ratios were used to assess the relationships between dichotomous factors and Nocardia mastitis. An odds ratio greater than 1.0 indicated that herds in which a certain management practice was used were more likely to be case herds than herds in which the management practice had not been used. Since Nocardia mastitis is relatively rare, the odds ratio can be interpreted in the same way as a relative risk (8). Consequently, the odds ratio of 2.8 for D5 (Table 3) suggests that herds in which D5 had been used were approximately 2.8 times as likely to be a case herd than herds in which D5 had not been used. The statistical significance of these unconditional relationships was assessed by the use of the chi-square statistic. Student's t-tests were used to assess unconditional relationships between variables measured on a continuous scale and Nocardia mastitis. The purpose of all unconditional analyses (both odds ratios and t-tests) was simply to identify potential risk factors for inclusion in the logistic regression analyses. Logistic regression was used to simultaneously evaluate the effects of all the potential risk factors identified by the univariate analyses. It is a statistical technique which measures the effects of several variables on a single dichotomous outcome variable which, in this study, was the herd's Nocardia mastitis status. Odds ratios obtained from the logistic regression analyses have been adjusted to remove the confounding effects of other possible risk factors. These odds ratios can be obtained for both dichotomous and continuously distributed factors. The odds ratio for a continuous variable is the increased risk associated with an increase of one unit in the variable. For example, using the coefficient for NUMCOW in Table 5 (0.019), when comparing the risk of Nocardia mastitis in a 40 cow herd to the risk in a 30 cow herd, the 40 cow herd would have an increased risk (odds ratio) of e0.019(40-30)= 1.21. As with the Ontario case-control study (7), Nocardia mastitis was associated with low herd somatic cell counts in simple univariate analyses. However, the difference in BSCC88 between case and control herds in this study was small and it became nonsignificant once other variables were controlled in the logistic regression analyses. It seems unlikely that herds with a low cell count are at higher risk of developing Nocardia mastitis. Of the four specific products identified as risk factors in the univariate analyses (Tables 2 and 3), only one, the lactating cow product L3 (CO-OP Lactating), failed to remain as a significant predictor in the logistic regression analyses. D5 (Biotef) was the product identified as "Product A" in the previous study (Stark personal communication). D18 (Orbenin Dry Cow) was the product identified as "Product B" in that study. D8 (CO-OP Dry Cow) was the same as D4 (Biodry, The Upjohn Company, Orangeville, Ontario). Those two products are manufactured by the same company and marketed under the two labels.

They were both similar to Biotef except that they did not contain a corticosteroid. It is interesting to note that D4 (Biodry) was not significantly associated with the risk of Nocardia mastitis in our study. The logistic regression analysis supported the observations from the univariate analysis. D5 (Biotef) and D8 (CO-OP Dry Cow) were both significantly and positively associated with Nocardia mastitis. D18 (Orbenin Dry Cow) exhibited a significant protective effect. Since the active agent in both Biotef and CO-OP Dry Cow is neomycin and the active ingredient in Orbenin Dry Cow is cloxacillin, an obvious second step was to pool the neomycin-containing products and use logistic regression analysis to examine their effect and that of the cloxacillin-containing product on the herds' Nocardia status. Three other infrequently used products (Antimast, Coopers Agropharm Inc., Ajax, Ontario; Mastitis Care, Pfizer Canada Inc., Pointe Claire- Dorval, Quebec; and Neospan, rogar/stb Inc., Pointe Claire-Dorval, Quebec) were included in the variable DNEO. The results support the hypothesis that neomycin-containing dry cow products are a risk factor for Nocardia mastitis (odds ratio = 28.91). The significant interaction between neomycin-containing products and cloxacillin-containing products suggests that the use of cloxacillin-containing products in herds in which neomycin-containing products had also been used, greatly reduced the risk associated with the use of the neomycin-containing products (odds ratio = 1.58). This may simply have been due to a reduction in the amount of neomycin-containing products used. The positive association of risk of Nocardia mastitis with the percentage of cows dry treated (DCPCT) should not be interpreted as a suggestion that the use of all dry cow therapies be curtailed. This study was unable to determine the amount of all products used on the farms. The increased risk associated with higher levels of dry cow therapy usage may have been due entirely to increased use of neomycin-containing products. Subsequent studies once neomycin-containing products have been eliminated will be required to investigate this. During the course of the Nocardia mastitis outbreak, the manufacturers of the neomycin-containing dry cow products voluntarily stopped their production. Since mid-1989, the number of cases of Nocardia mastitis has declined, although cases are still at a much higher level than they were prior to 1988 (14). Acknowledgments We thank Heather Prudence for conducting the interviews and bulk tank sampling; Pat Brown and the staff at the Nova Scotia Veterinary Pathology Laboratory for handling the bulk samples; Nova Scotia Department of Small Business Development, Public Sector Department of Agriculture and Marketing for financial assistance; and the farmers who participated in the study. cvj References 1. Barnum DA, Fuller DS. A report on the isolation of two species of Nocardia from bovine mastitis. Proc Northeast Mastitis Council, 1956. 2. Pier AC, Gray DM, Fossatti BS. Nocardia asteriodes - A newly recognized pathogen of the mastitis complex. Am J Vet Res 1958; 19: 319-331. 3. Pier AC, Willers EH, Mejia BS. Nocardia asteriodes as a mammary pathogen of cattle. II. The sources of Nocardial infection and experimental reproduction of the disease. Am J Vet Res 1961; 22: 698-703. 4. Bushnell RB, Pier AC, Fichtner RE, Beaman BL, Boos HA, Salman MD. Clinical and diagnostic aspects of herd problems with nocardial and mycobacterial mastitis. Proc 22nd Am Assoc Vet Lab Diagnost 1979: 1-12. 5. Hibbs CM, Hopson JH, Ferguson R, Thilsted JP. Nocardia mastitis in a large dairy herd. Proc 23rd Am Assoc Vet Lab Diagnost 1980: 73-78. 6. Dohoo IR. Nocardia spp. mastitis in Canada. Can Vet J 1989; 30: 969. 7. Stark DA, Anderson NG. A case-control study of Nocardia mastitis in Ontario dairy herds. Can Vet J 1990; 31: 197-201. 8. Martin SW, Meek AH, Willeberg P. Veterinary Epidemiology: Principles and Methods. Ames: Iowa State University Press, 1987: 44-46. 9. SAS Procedures Guide. Release 6.03 Edition 1988. SAS/STAT Guide for Personal Computers, Version 6 Edition 1987. Cary, North Carolina: SAS Institute Inc, 1987. 10. BMDP Statistical Software Manual 1985. Berkeley, California: University of California Press, 1985. 11. Schoonderwoerd M, McFadzen LL, Manninen KI, Ollis GW. Culturing of bulk tank milk for the presence of Nocardia spp. Can Vet J 1990; 31: 453-454. 12. Kleinbaum DG, Kupper LL, Morgenstern H. Selection bias and information bias. In: Epidemiologic Research. Principles and Quantitative Methods. New York: Van Nostrand Reinhold Company, 1982: 194-241. 13. Lilienfeld AM, Lilienfeld DE. Observational Studies: 1. Retrospective and cross-sectional studies. In: Foundations of Epidemiology, 2nd ed. New York: Oxford University Press, 1980: 191-225. 14. Dohoo IR. Update on Nocardia sp. mastitis. Can Vet J 1991; 32: 116. How can we Help you? 45M 1-800-567-CVMA z i1-800-567-2862 t;<, Cj? +} 339 Booth Street Ottawa, Ontario KlR 7K1 ~; (613) 236-1162 Fax (613) 236-9681 Another Membership Service from the CVMA Comment est-ce qu'on peut vous aider? 1-800-567-2862 1 339, rue Booth o FTvEgs,Fvs ( Ottawa (Ontario) KlR 7K1 95s (613) 236-1162 Fax (613) 236-9681 Un autre service aux membres de l'acv 677