Bijlage A: MODEL VOOR DE TITELPAGINA VAN HET PROEFSCHRIFT

Impact of bovine subclinical mastitis and effect of lactational treatment

Bart van den Borne 2010

Van den Borne, Bart H. P. 2010. Impact of bovine subclinical mastitis and effect of lactational treatment. Dissertation Faculty of Veterinary Medicine, Utrecht University, Utrecht. the Netherlands -with summaries in English and DutchISBN: 978-90-393-5378-3 Photos: Bart van den Borne Cover design: Anjolieke Dertien, Multimedia Division, Faculty of Veterinary Medicine, Utrecht University, the Netherlands Layout: Harry Otter, Multimedia Division, Faculty of Veterinary Medicine, Utrecht University, the Netherlands Printed by: Ridderprint, Ridderkerk, the Netherlands

Impact of bovine subclinical mastitis and effect of lactational treatment Impact van subklinische mastitis en het effect van lactatiebehandelingen bij melkkoeien (met een samenvatting in het Nederlands)

Proefschrift ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de rector magnificus, prof.dr. J.C. Stoof, ingevolge het besluit van het college voor promoties in het openbaar te verdedigen op dinsdag 12 oktober 2010 des middags te 12.45 uur door Bart Henricus Philippus van den Borne geboren op 15 februari 1980 te Reusel

Promotor:

Prof.dr. J. A. Stegeman

Co-promotoren:

Dr. M. Nielen Dr.ir. G. van Schaik

This study was part of the 5-year mastitis program of the Dutch Udder Health Centre (UierGezondheidsCentrum Nederland) and was financially supported by the Dutch Dairy Board. Printing of this thesis was financially supported by the Department of Farm Animal Health of the Faculty of Veterinary Medicine of Utrecht University and Boehringer Ingelheim.

All the good ideas I ever had came to me while I was milking a cow Grant Wood, American painter (1981-1942)

Contents Chapter 1

General introduction

1

Chapter 2

Variation in herd level mastitis indicators between primi- and multiparae in Dutch dairy herds

11

Chapter 3

Quantification of the relationship between high somatic cell counts and subsequent clinical mastitis in dairy cows with two analytical methods

27

Chapter 4

Therapeutic effects of antimicrobial treatment during lactation of recently acquired bovine subclinical mastitis: Two linked randomized field trials

49

Chapter 5

Host adaptation of bovine Staphylococcus aureus seems associated with bacteriological cure after lactational antimicrobial treatment

79

Chapter 6

Bioeconomic modeling of lactational antimicrobial treatment of new bovine subclinical intramammary infections caused by contagious pathogens

97

Chapter 7

General discussion

117

Summary

147

Samenvatting

153

Dankwoord

163

Curriculum Vitae

167

List of publications

171

V{tÑàxÜ


D


Mastitis is the inflammation of the mammary gland tissue. Mastitis in dairy cows is generally caused by infection of bacteria, but sterile inflammation due to chemical, physical or mechanical trauma may also occur. Intramammary infections (IMI) generally occur when pathogens pass the teat canal and penetrate the mammary gland. Thereafter, the cow’s immune system initiates a cellular response to kill the pathogens by phagocytosis and bactericidal activity of neutrophils and macrophages (Sordillo et al., 1997). Mastitis is considered to be clinical when inflammation is accompanied with visible alterations of milk or udder gland. Clinical signs of mastitis include clotting and discoloration of milk, redness and swelling of the udder, fever and even death. Mastitis is considered to be subclinical when IMI is present but no visible signs occur. Subclinical mastitis can be detected only through laboratory techniques such as somatic cell counts (SCC; Pyörälä, 2003; Schukken et al., 2003). SCC measure the number of cells (mainly inflammatory cells during an immune response) (Leitner et al., 2000; Sarikaya et al., 2006) and indicate IMI when elevated (Dohoo et Leslie, 1991; Reksen et al., 2008). SCC are used as a diagnostic tool to monitor subclinical mastitis in dairy herds worldwide (Schukken et al., 2003). Losses because of mastitis Bovine mastitis is associated with milk yield losses (Halasa et al., 2009; Schukken et al., 2009), changes in milk composition (Halasa et al., 2009), extra labor (Halasa et al., 2007), costs of antimicrobial treatment, culling and, occasionally, death (Bar et al., 2008). Mastitis may also result in financial penalties when bulk milk SCC are too high. The average costs of a clinical mastitis case were estimated to be €210 under Dutch circumstances, varying from €164 to €235 depending on the month of lactation (Huijps et al., 2008). An indication of the occurrence of mastitis can be extrapolated from Sampimon et al. (2009), who identified 22% of dairy cows to have an elevated composite SCC in a random sample of 49 Dutch dairy herds. A further indication would be the yearly incidence rate of clinical mastitis; estimated as 26.2 cases per 100 cows in the Netherlands in 1993 and 1994 (Barkema et al., 1998). More recent estimates are unavailable. Altogether, for the Dutch dairy industry, bovine mastitis results in an economic loss of about 100 million euros annually (van der Zwaag et al., 2005). Moreover, clinical mastitis distresses the animal (Willeberg, 1994; Kemp et al., 2008) and frustrates the farmer’s milking routine (Jansen et al., 2009). Subclinical mastitis can occur before clinical onset. Elevated composite SCC have been observed before clinical mastitis occurred (de Haas et al., 2002). The relation between increased SCC and the occurrence of clinical mastitis has been quantified (e.g. Peeler et al., 2003; Green et al., 2004; Whist and Østerås , 2007) and subclinical mastitis identified by bacterial culture was associated with increased probabilities of clinical mastitis (Reksen et al., 2006; Whist et al., 2009).

3

Chapter 1

Pathogens and strains Many pathogens can cause mastitis, but the majority of IMI is caused by a few bacterial species. The most important major pathogens involved with bovine mastitis worldwide are Staphylococcus aureus, Streptococcus uberis, Streptococcus dysgalactiae, Streptococcus agalactiae, Escherichia coli and Klebsiella spp. (Barkema et al., 1998; Gianneechini et al., 2002; Karimuribo et al., 2006; Bradley et al., 2007; Olde Riekerink et al., 2008). Coagulase negative staphylococci (CNS) and Corynebacterium bovis, two other highly prevalent pathogens, are historically considered to be of limited importance and are therefore often described as minor pathogens. The impact of CNS is increasing (Pyörälä and Taponen, 2009), probably because prevalences of major pathogens are decreasing (Sampimon et al., 2009). Some pathogens, e.g. Strep. agalactiae and Staph. aureus, are considered to be contagious (Keefe, 1997; Barkema et al., 2009), but environmental Staph. aureus IMI may also occur (Zadoks et al., 2000, 2002). E. coli and Klebsiella spp. have mainly an environmental origin (Nemeth et al., 1994; Munoz et al., 2007). Other pathogens have both routes of infection. Strep. uberis IMI originate mainly from the environment (Pullinger et al., 2006), but can also behave contagious (Zadoks et al., 2001, 2003). Strep. dysgalactiae behaves intermediate between contagious and environmental transmission (Baseggio et al., 1997). For CNS, both environmental and contagious IMI occur (Taponen et al., 2008; Taponen and Pyörälä, 2009). Environmental or contagious transmission is assumed to depend on strain characteristics (or species and strain characteristics in the case of CNS). Additionally, clinical characteristics differ within bacterial species. For example, Staph. aureus strain characteristics have been associated with persistence, within-herd prevalence, somatic cell count, milk production and severity of mastitis (Zadoks et al., 2000; Middleton and Fox, 2002; Smith et al., 2005; Graber et al., 2009). To determine strain characteristics, genotypic techniques are preferred above phenotypic techniques because of phenotypic misclassification (Zadoks and Watts, 2009). Treatment of subclinical mastitis Treatment of subclinical mastitis with antimicrobials is one option to improve udder health in dairy herds and can be applied at drying off or during lactation. Curing subclinical mastitis may decrease the probabilities of clinical mastitis flare-ups and culling and also reduce SCC and milk yield losses within the cow (Barkema et al., 2006). Treating subclinical mastitis might also prevent additional losses in other cows by reducing the transmission of contagious pathogens within the herd (Barlow et al., 2009). Lactational treatment of subclinical mastitis seems economically beneficial in some circumstances, depending on the probability of cure and the economic value of the cow (Yamagata et al., 1987; Swinkels et al., 2005a,b; Steeneveld et al., 2007). In contrast to the widespread application of blanket dry cow treatment in the Netherlands (87% of the dairy

4


herds), lactational treatment of subclinical mastitis is employed less frequently (47% of the dairy herds; Sampimon et al., 2008). The success of subclinical mastitis treatment depends on the pathogen species and on strain, treatment and cow factors (Barkema et al., 2006). Streptococci and CNS subclinical mastitis have a relatively high probability of cure after treatment (Wilson et al., 1999; Deluyker et al., 2005; Salat et al., 2008), whereas those for Staph. aureus are relatively low (Sol et al., 1997; Deluyker et al., 2005; Sandgren et al., 2008). CNS species show no difference in their probability of cure (Taponen et al., 2006). The most well-known strain characteristic to affect probability of cure is probably the presence of antimicrobial resistance (Owens et al., 1997; Sol et al., 1997, 2000). Other strain characteristics have also been identified as affecting cure of subclinical mastitis at drying off (Dingwell et al., 2006) or cure of clinical mastitis (Milne et al., 2005; Haveri et al., 2007). Treatment factors includes factors such as antimicrobial compounds, the combination of compounds, route of application and duration of treatment. Duration of treatment seems the most important treatment factor, as reviewed by Barkema et al. (2006) for Staph. aureus. Cow factors that affect cure are parity, stage of lactation, number and location of the infected quarter and SCC (Sol et al., 1997; Deluyker et al., 2005; Salat et al., 2008; Sandgren et al., 2008). Intervention of recently acquired subclinical mastitis Probability of bacteriological cure is believed to improve when lactational treatment of subclinical mastitis is used after a short period of infection (Barkema et al., 2006). Probabilities of cure were high in a pilot study with cows in early lactation (Beggs and Wraight, 2006) and in experimentally infused animals (Milner et al., 1997; Owens et al., 1997). Cure rate of subclinical mastitis decreased with an increasing number of culturepositive samples pre-treatment during lactation (Sol et al., 1997) and at drying off (Sol et al., 1994; Dingwell et al., 2003). Udder health in dairy herds may improve by decreasing the impact of subclinical mastitis through early intervention. Hence, there is a need for knowledge on the epidemiologic and economic effects of lactational treatment of recently acquired subclinical mastitis. Objectives The main objectives of this dissertation were to estimate the impact of subclinical mastitis in the Netherlands and to explore the beneficial effects of early intervention of subclinical mastitis by antimicrobial therapy during lactation. To reach these objectives, several studies were conducted with the following goals: • To estimate the occurrence of clinical and subclinical mastitis in primiparae and multiparae in a random sample of Dutch dairy herds (Chapter 2). • To quantify the relationship between elevated composite SCC and clinical mastitis later in lactation (Chapter 3).

5

Chapter 1

•

To quantify therapeutic effects of early lactational treatment of subclinical mastitis and to identify factors related to therapeutic success in a randomized field trial (Chapter 4). • To investigate the relation between genotypic Staph. aureus strain characteristics and the probability of cure after lactational treatment of subclinical mastitis (Chapter 5). • To simulate the direct and indirect epidemiologic and economic effects of early lactational intervention of subclinical IMI caused by contagious pathogens (Chapter 6). The work presented in Chapters 2-6 is summarized and discussed in Chapter 7 and is complemented with unpublished results from additional studies. Chapter 7 concludes with recommendations about how to apply the results from this dissertation into dairy practice.

REFERENCES Bar, D., Y. T. Gröhn, G. Bennett, R. N. González, J. A. Hertl, H. F. Schulte, L. W. Tauer, F. L. Welcome, and Y. H. Schukken. 2008. Effects of repeated episodes of generic clinical mastitis on mortality and culling in dairy cows. J. Dairy Sci. 91:2196-2204. Barkema, H. W., M. J. Green, A. J. Bradley, and R. N. Zadoks. 2009. Invited review: The role of contagious disease in udder health. J. Dairy Sci. 92:4717-4729. Barkema, H. W., Y. H. Schukken, T. J. G. M. Lam, M. L. Beiboer, H. Wilmink, G. Benedictus, and A. Brand. 1998. Incidence of clinical mastitis in dairy herds grouped in three categories by bulk milk somatic cell counts. J. Dairy Sci. 81:411-419. Barkema, H. W., Y. H. Schukken, and R. N. Zadoks. 2006. Invited review: The role of cow, pathogen, and treatment regimen in the therapeutic success of bovine Staphylococcus aureus mastitis. J. Dairy Sci. 89:1877-1895. Barlow, J. W., L. J. White, R. N. Zadoks, and Y. H. Schukken. 2009. A mathematical model demonstrating indirect and overall effects of lactation therapy targeting subclinical mastitis in dairy herds. Prev. Vet. Med. 90:31-42. Baseggio, N., P. D. Mansell, and G. F. Browning. 1997. Strain differentiation of isolates of streptococci from bovine mastitis by pulsed-field gel electrophoresis. Mol. Cell. Probes 11:349-354. Beggs, D. S. and M. D. Wraight. 2006. Pilot study - parenteral treatment of recently acquired subclincal mastitis during lactation. Aust. Vet. J. 84:50-52. Bradley, A. J., K. A. Leach, J. E. Breen, L. E. Green, and M. J. Green. 2007. Survey of the incidence and aetiology of mastitis on dairy farms in England and Wales. Vet. Rec. 160:253-258. de Haas, Y., H. W. Barkema, and R. F. Veerkamp. 2002. The effect of pathogen-specific clinical mastitis on the lactation curve for somatic cell count. J. Dairy Sci. 85:1314-1323. Deluyker, H. A., S. N. Van Oye, and J. F. Boucher. 2005. Factors affecting cure and somatic cell count after pirlimycin treatment of subclinical mastitis in lactating cows. J. Dairy Sci. 88:604-614. Dingwell, R. T., K. E. Leslie, T. F. Duffield, Y. H. Schukken, L. DesCoteaux, G. P. Keefe, D. F. Kelton, K. D. Lissemore, W. Shewfelt, P. DIck, and R. Bagg. 2003. Efficacy of intramammary tilmicosin and risk factors for cure of Staphylococcus aureus infection in the dry period. J. Dairy Sci. 86:159-168. Dingwell, R. T., K. E. Leslie, P. Sabour, D. Lepp, and J. Pacan. 2006. Influence of genotype of Staphylococcus aureus, determined by pulsed-field gel electrophoresis, on dry-period elimination of subclinical mastitis in Canadian dairy herds. Can. J. Vet. Res. 70:115-120.

6

General introduction Dohoo, I. R. and K. E. Leslie. 1991. Evaluation of changes in somatic cell counts as indicators of new intramammary infections. Prev. Vet. Med. 10:225-237. Gianneechini, R., C. Concha, R. Rivero, I. Delucci, and J. Moreno López. 2002. Occurence of clinical and sub-clinical mastitis in dairy herds in the West Littoral Regin in Uruguay. Acta Vet. Scand. 43:221-230. Graber, H. U., J. Naskova, E. Studer, T. Kaufmann, M. Kirchhofer, M. Brechbühl, W. Schaeren, A. Steiner, and C. Fournier. 2009. Mastitis-related subtypes of bovine Staphylococcus aureus are characterized by different clinical properties. J. Dairy Sci. 92:1442-1451. Green, M. J., P. R. Burton, L. E. Green, Y. H. Schukken, A. J. Bradley, E. J. Peeler, and G. F. Medley. 2004. The use of Markov chain Monte Carlo for analysis of correlated binary data: patterns of somatic cell counts in milk and the risk of clinical mastitis in dairy cows. Prev. Vet. Med. 64:157-174. Halasa, T., K. Huijps, O. Østerås, and H. Hogeveen. 2007. Economic effects of bovine mastitis and mastitis management: A review. Vet. Q. 29:18-31. Halasa, T., M. Nielen, A. P. W. de Roos, R. van Hoorne, G. de Jong, T. J. G. M. Lam, T. van Werven, and H. Hogeveen. 2009. Production loss due to new subclinical mastitis in Dutch dairy cows estimated with a test-day model. J. Dairy Sci. 92:599-606. Haveri, M., A. Roslöf, L. Rantala, and S. Pyörälä. 2007. Virulence genes of bovine Staphylococcus aureus from persistent and nonpersistent intramammary infections with different clinical characteristics. J. Appl. Microbiol. 103:993-1000. Huijps, K., T. J. G. M. Lam, and H. Hogeveen. 2008. Costs of mastitis: facts and perception. J. Dairy Res. 75:113-120. Jansen, J., B. H. P. van den Borne, R. J. Renes, G. van Schaik, T. J. G. M. Lam, and C. Leeuwis. 2009. Explaining mastitis incidence in Dutch dairy farming: the influence of farmers' attitudes and behaviour. Prev. Vet. Med. 92:210-223. Karimuribo, E. D., J. L. Fitzpatrick, C. E. Bell, E. S. Swai, D. M. Kambarage, N. H. Ogden, M. J. Bryant, and N. P. French. 2006. Clinical and subclinical mastitis in smallholder dairy farms in Tanzania: Risk, intervention and knowledge transfer. Prev. Vet. Med. 74:84-98. Keefe, G. P. 1997. Streptococcus agalactiae: A review. Can. Vet. J. 38:429-437. Kemp, M. H., A. M. Nolan, P. J. Cripps, and J. L. Fitzpatrick. 2008. Animal-based measurements of the severity of mastitis in dairy cows. Vet. Rec. 163:175-179. Leitner, G., E. Shoshani, O. Krifucks, M. Chaffer, and A. Saran. 2000. Milk leucocyte population patterns in bovine udder infection of different aetiology. J. Vet . Med. B Infect. Dis. Vet. Public Health 47:581-589. Middleton, J. R. and L. K. Fox. 2002. Influence of Staphylococcus aureus strain on mammary quarter milk production. Vet. Rec. 150:411-413. Milne, M. H., A. M. Biggs, D. C. Barrett, F. J. Young, S. Doherty, G. T. Innocent, and J. L. Fitzpatrick. 2005. Treatment of persistent intramammary infections with Streptococcus uberis in dairy cows. Vet. Rec. 157:245-250. Milner, P., K. L. Page, and J. E. Hillerton. 1997. The effects of early antibiotic treatment following diagnosis of mastitis detected by a change in the electrical conductivity of milk. J. Dairy Sci. 80:859-863. Munoz, M. A., F. L. Welcome, Y. H. Schukken, and R. N. Zadoks. 2007. Molecular epidemiology of two Klebsiella pneumoniae mastitis outbreaks on a dairy farm in New York State. J. Clin. Microbiol. 45:3964-3971. Nemeth, J., C. A. Muckle, and C. L. Gyles. 1994. In vitro comparison of bovine mastitis and fecal Escherichia coli isolates. Vet. Microbiol. 40:231-238. Olde Riekerink, R. G. M., H. W. Barkema, D. F. Kelton, and D. T. Scholl. 2008. Incidence rate of clinical mastitis on Canadian dairy farms. J. Dairy Sci. 91:1366-1377.

7

Chapter 1 Owens, W. E., C. H. Ray, J. L. Watts, and R. J. Yancey. 1997. Comparison of success of antibiotic therapy during lactation and results of antimicrobial susceptibility tests for bovine mastitis. J. Dairy Sci. 80:313-317. Peeler, E. J., M. J. Green, J. L. Fitzpatrick, and L. E. Green. 2003. The association between quarter somatic-cell counts and clinical mastitis in three British dairy herds. Prev. Vet. Med. 59:169180. Pullinger, G. D., M. López-Benavides, T. J. Coffey, J. H. Williamson, R. T. Cursons, E. Summers, J. Lacy-Hulbert, M. C. Maiden, and J. A. Leigh. 2006. Application of Streptococcus uberis multilocus sequence typing: analysis of the population structure detected among environmental and bovine isolates from New Zealand and the United Kingdom. Appl. Environ. Microbiol. 72:1429-1436. Pyörälä, S. 2003. Indicators of inflammation in the diagnosis of mastitis. Vet. Res. 34:565-578. Pyörälä, S. and S. Taponen. 2009. Coagulase-negative staphylococci-Emerging mastitis pathogens. Vet. Microbiol. 134:3-8. Reksen, O., L. Sølverød, A. J. Branscum, and O. Østerås. 2006. Relationships between milk culture results and treatment for clinical mastitis or culling in Norwegian dairy cattle. J. Dairy Sci. 89:2928-2937. Reksen, O., L. Sølverød, and O. Østerås. 2008. Relationships between milk culture results and composite milk somatic cell counts in Norwegian dairy cattle. J. Dairy Sci. 91:3102-3113. Salat, O., F. Sérieys, B. Poutrel, L. Durel, and L. Goby. 2008. Systemic treatment of subclinical mastitis in lactating cows with penethamate hydriodide. J. Dairy Sci. 91:632-640. Sampimon, O., H. W. Barkema, I. Berends, J. Sol, and T. Lam. 2009. Prevalence of intramammary infection in Dutch dairy herds. J. Dairy Res. 76:129-136. Sampimon, O. C., R. G. M. Olde Riekerink, and T. J. G. M. Lam. 2008. Prevalence of subclinical mastitis pathogens and adoption of udder health management practices on Dutch dairy farms: preliminary results. In: Mastitis control - From science to practice. Wageningen Academic Publishers, Wageningen, the Netherlands. Sandgren, C. H., K. Persson Waller, and U. Emanuelson. 2008. Therapeutic effects of systematic or intramammary antimicrobial treatment of bovine subclinical mastitis during lactation. Vet. J. 175:108-177. Sarikaya, H., G. Schlamberger, H. H. D. Meyer, and R. M. Bruckmaier. 2006. Leukocyte Populations and mRNA Expression of Inflammatory Factors in Quarter Milk Fractions at Different Somatic Cell Score Levels in Dairy Cows. J. Dairy Sci. 89:2479-2486. Schukken, Y. H., J. A. Hertl, D. Bar, G. J. Bennett, R. N. González, B. J. Rauch, C. Santisteban, H. F. Schulte, L. Tauer, F. L. Welcome, and Y. T. Gröhn. 2009. Effects of repeated gram-positive and gram-negative clinical mastitis episodes on milk yield loss in Hostein dairy cows. J. Dairy Sci. 92:3091-3105. Schukken, Y. H., D. J. Wilson, F. Welcome, L. Garrison-Tikofsky, and R. N. González. 2003. Monitoring udder health and milk quality using somatic cell counts. Vet. Res. 34:579-596. Smith, B. S., L. E. Green, G. F. Medley, H. E. Bird, and C. G. Dowson. 2005. Multilocus sequence typing of Staphylococcus aureus isolated from high-somatic-cell-count cows and the environment of an organic dairy farm in the United Kingdom. J. Clin. Microbiol. 43:4731-4736. Sol, J., O. C. Sampimon, H. W. Barkema, and Y. H. Schukken. 2000. Factors associated with cure after therapy of clinical mastitis caused by Staphylococcus aureus. J. Dairy Sci. 83:278-284. Sol, J., O. C. Sampimon, J. J. Snoep, and Y. H. Schukken. 1994. Factors associated with bacteriological cure after dry cow treatment of subclinical staphylococcal mastitis with antibiotics. J. Dairy Sci. 77:75-79. Sol, J., O. C. Sampimon, J. J. Snoep, and Y. H. Schukken. 1997. Factors associated with bacteriological cure during lactation after therapy for subclinical mastitis caused by Staphylococcus aureus. J. Dairy Sci. 80:2803-2808.

8

General introduction Sordillo, L. M., K. Shafer-Weaver, and D. DeRosa. 1997. Immunobiology of the mammary gland. J. Dairy Sci. 80:1851-1865. Steeneveld, W., J. M. Swinkels, and H. Hogeveen. 2007. Stochastic modelling to assess economic effects of treatment of chronic subclinical mastitis caused by Streptococcus uberis. J. Dairy Res. 74:459-467. Swinkels, J. M., H. Hogeveen, and R. N. Zadoks. 2005a. A partial budget model to estimate economic benefits of lactational treatment of subclinical Staphylococcus aureus mastitis. J. Dairy Sci. 88:4273-4287. Swinkels, J. M., J. G. A. Rooijendijk, R. N. Zadoks, and H. Hogeveen. 2005b. Use of partial budgeting to determine the economic benefits of antibiotic treatment of chronic subclinical mastitis caused by Streptococcus uberis or Streptococcus dysgalactiae. J. Dairy Res. 72:75-85. Taponen, S., L. Björkroth, and S. Pyörälä. 2008. Coagulase-negative staphylococci isolated from bovine extramammary sites and intramammary infections in a single dairy herd. J. Dairy Res. 75:422429. Taponen, S. and S. Pyörälä. 2009. Coagulase-negative staphylococci as cause of bovine mastitis—Not so different from Staphylococcus aureus?. Vet. Microbiol. 134:29-36. Taponen, S., H. Simojoki, M. Haveri, H. D. Larsen, and S. Pyörälä. 2006. Clinical characteristics and persistence of bovine mastitis caused by different species of coagulase-negative staphylococci identified with API or AFLP. Vet. Microbiol. 115:199-207. van der Zwaag, H., G. Van Schaik, and T. J. G. M. Lam. 2005. Mastitis control program in the Netherlands: Goal, tools and conditions. In: The 4th IDF International Mastitis Conference 'Mastitis in Dairy Production: Current Knowledge and Future Solutions'. H. Hogeveen, ed. Wageningen Academic Publishers, Wageningen, the Netherlands. Whist, A. C. and O. Østerås. 2007. Associations between somatic cell counts at calving or prior to dryingoff and clinical mastitis in the remaining or subsequent lactation. J. Dairy Res. 74:66-73. Whist, A. C., O. Østerås, and L. Sølverød. 2009. Association between isolation of Staphylococcus aureus one week after calving and milk yield, somatic cell count, clinical mastitis, and culling through the remaining lactation. J. Dairy Res. 76:24-35. Willeberg, P. 1994. An international perspective on bovine somatotropin and clinical mastitis. J. Am. Vet. Med. Assoc. 205:538-541. Wilson, D. J., R. N. González, K. L. Case, L. L. Garrison, and Y. T. Gröhn. 1999. Comparison of seven antibiotic treatments with no treatment for bacteriological efficacy against bovine mastitis pathogens. J. Dairy Sci. 82:1664-1670. Yamagata, M., W. J. Goodger, L. Weaver, and C. Franti. 1987. The economic benefit of treating subclinical Streptococcus agalactiae mastitis in lactating cows. J. Am. Vet. Med. Assoc. 191:1556-1561. Zadoks, R., W. van Leeuwen, H. Barkema, O. Sampimon, H. Verbrugh, Y. H. Schukken, and A. van Belkum. 2000. Application of pulsed-field gel electrophoresis and binary typing as tools in veterinary clinical microbiology and molecular epidemiologic analysis of bovine and human Staphylococcus aureus isolates. J. Clin. Microbiol. 38:1931-1939. Zadoks, R. N., H. G. Allore, H. W. Barkema, O. C. Sampimon, Y. T. Gröhn, and Y. H. Schukken. 2001. Analysis of an outbreak of Streptococcus uberis mastitis. J. Dairy Sci. 84:590-599. Zadoks, R. N., H. G. Allore, T. J. Hagenaars, H. W. Barkema, and Y. H. Schukken. 2002. A mathematical model of Staphylococcus aureus control in dairy herds. Epidemiol. Infect. 129:397-416. Zadoks, R. N., B. E. Gillespie, H. W. Barkema, O. C. Sampimon, S. P. Oliver, and Y. H. Schukken. 2003. Clinical, epidemiological and molecular characteristics of Streptococcus uberis infections in dairy herds. Epidemiol. Infect. 130:335-349. Zadoks, R. N. and J. L. Watts. 2009. Species identification of coagulase-negative staphylococci: genotyping is superior to phenotyping. Vet. Microbiol. 134:20-28.

9

V{tÑàxÜ

E

Variation in herd level mastitis indicators between primi- and multiparae in Dutch dairy herds

Bart H. P. van den Borne Gerdien van Schaik Theo J. G. M. Lam Mirjam Nielen

Preventive Veterinary Medicine (2010) 96:49-55

Chapter 2

ABSTRACT Composite somatic cell count data from the national test day recording and reported cases of farmer diagnosed clinical mastitis were used to estimate the occurrence of mastitis from July 2004 to June 2005 in primi- and multiparae in the Netherlands. Herds had to participate in the test day recording and had to have at least 50 cows. A random selection of 396 of these dairy herds provided composite somatic cell count data, while 205 dairy herds additionally reported on clinical mastitis cases. Prevalence of subclinical mastitis was calculated per herd as the proportion of cows with somatic cell count > 200,000 cells/mL. The incidence rate of clinical mastitis was calculated as the number of clinical mastitis cases divided by the number of cow days at risk per herd. Negative binomial models were used to correct for overdispersion. Mean herd level subclinical mastitis prevalence was 12.8% (95% CI: 12.2 – 13.5%) in primi- and 27.1% (95% CI: 26.2 – 28.1%) in multiparae. Mean herd level clinical mastitis cases were observed 20.2 (95% CI: 18.3 – 22.4) and 39.6 (95% CI: 37.1 – 42.3) times per 100 cow-years at risk, respectively. Some herds had a high mastitis occurrence in one parity group, while it was low in the other. Parity-specific monitoring is needed to identify such herds. Keywords: clinical mastitis, somatic cell count, primiparae, multiparae, monitoring

12

Occurrence of mastitis in the Netherlands

INTRODUCTION Primiparae are the future production potential of a dairy herd. After 2 years of investment, primiparae are supposed to earn the investment back through milk production. A healthy start of the lactation cycle is therefore essential. However, mastitis is often seen in freshly calved primiparae. Although the incidence rate of clinical mastitis (IRCM) is generally higher in multiparae, a higher IRCM has been observed in primiparae in early lactation (Miltenburg et al., 1996; Barkema et al., 1998; Valde et al., 2004; Nyman et al., 2007; Olde Riekerink et al., 2008; Nyman et al., 2009). Subclinical mastitis (SCM) in primiparae in early lactation has also been associated with higher somatic cell counts (SCC) during the rest of the lactation (De Vliegher et al., 2004), with production losses (De Vliegher et al., 2005a), and with higher probabilities of clinical mastitis (CM) (Whist et al., 2007; Steeneveld et al., 2008) and culling (De Vliegher et al., 2005b; Whist et al., 2007). Therefore, mastitis in primiparae can be economically detrimental for a farmer’s business (Huijps et al., 2009). Mastitis control programs often focus on herds with a high occurrence of mastitis because the greatest gain is to be expected in those herds. Bulk milk somatic cell count (BMSCC) or the overall (regardless of parity) IRCM are commonly used to identify these herds. A high IRCM in primiparae in early lactation was associated with a high IRCM in multiparae in two Swedish studies (Svensson et al., 2006; Nyman et al., 2007), suggesting that herds with a high occurrence of mastitis in their early lactating primiparae and a low occurrence of mastitis in multiparae (‘heifer mastitis problem herds’) are not common. This, however, is in contrast to field observations. Quantification of the relationship between mastitis in primi- and multiparae may be beneficial for mastitis control programs due to parity-specific interventions in these herds, but seems scarcely investigated in countries other than Sweden. The most recent data available in the Netherlands were collected in 1993-1994 from a non-random population of herds (Barkema et al., 1998) and were deemed both outdated and not representative for the general Dutch dairy population. The aims of this study were to quantify the occurrence of mastitis in Dutch primi- and multiparae and to determine the relationship between mastitis in primiparae and multiparae at herd level.

MATERIALS AND METHODS Herds Herds were randomly selected to collect data on SCM and CM occurrence during a 1-year observational period from July 1, 2004 to June 30, 2005. To focus on herds that would stay in business for the next 5 year, herds had to participate in the national test day recording, the herd size of the farms had to be over 50 cows and farmers had to be younger

13

Chapter 2

than 57. To exclude regional differences, the number of randomly selected herds was equally distributed over three regions of the Netherlands: North, East and South and West, each representing an equal number of herds at the start of the study period (CRV, 2004). Per region, 181 herds were contacted to participate in the study. In total, 404 herds agreed to participate, giving a response rate of 74%. Composite SCC from all cows of the participating herds that were ≥4 days in lactation (Barkema et al., 1999), provided by CRV (Arnhem, the Netherlands), were used to gather information on SCM. Additionally, a subset of participating farmers voluntarily reported at each test day which cows had developed CM in the preceding period. A CM case was diagnosed by the farmer as a cow with visual abnormalities in the milk and/or quarter. Those cows were uniquely identified in the test day recording and were digitally reported. Subsequently, farmers were regularly contacted by mail during the data collection to report the affected quarter and the date of occurrence of all the reported clinical mastitis cases in their herd. Follow-up calls were made if the requested information had not been returned. Bimonthly BMSCC measurements were obtained from Qlip (Zutphen, the Netherlands). Mastitis indicators A cow was assumed to have SCM in this study when it had an elevated composite SCC. SCM prevalence was calculated at the herd level as the proportion of cows at a test day with a composite SCC above a certain threshold. The threshold of 200,000 cells/mL was chosen as the default because this threshold is commonly used internationally (Schukken et al., 2003). Because sensitivity and specificity differ for different thresholds (Dohoo and Leslie, 1991; Schepers et al., 1997) and different thresholds are used worldwide, the thresholds of 250,000 cells/mL (Dohoo and Meek, 1982), 150,000 cells/mL and 100,000 cells/mL (Pyörälä, 2003) were also evaluated. Incidence rate of CM was calculated as the number of quarter cases of CM divided by the number of cow days at risk, both assessed at herd level. Every CM case diagnosed by the farmer was assigned to be a new case of CM with the exception of CM events occurring within 14 days in the same quarter. Those were considered to be the same case and were not included. Recurrent cases in the same quarter occurring ≥14 days later were included as different cases in the numerator. The number of days at risk was assessed as the number of days a cow was present at the farm during the study period, with the exception for primiparae, who became at risk at their first calving. Statistical analyses Herd level SCM prevalence was estimated using a negative binomial model with an autoregressive correlation structure to correct for repeated measures. The number of cows in a herd above the composite SCC threshold at a certain test day was used as the dependent variable, while the natural logarithm of the total number of cows at a certain

14


test day was used as the offset. A combination of sine-cosine components modelling the monthly changes in SCM prevalence was added to the model to correct for seasonal influences. SCM prevalence was calculated as eintercept x 100% and represents the average proportion of cows within a herd with an elevated SCC throughout the study period. Herd level IRCM was also estimated using a negative binomial model. The number of quarter cases in the 1-year observational period was used as the dependent variable, while the natural logarithm of the number of cow days at risk in each herd was used as the offset. IRCM was expressed as the number of quarter cases per 100 cow-years at risk and was calculated as eintercept x 365 x 100. Separate statistical models were fitted for primi- and multiparae to observe differences in the occurrence of mastitis in both groups. Standard errors, as estimated by the negative binomial models, were used to calculate 95% confidence intervals (CI) to provide information about precision of the estimated population means. Model fit was evaluated for all models by observation of normality of the residuals and all showed a good fit. Some alternative indicators were defined to determine the relationship between herd level mastitis indicators in primi- and multiparae. Annual herd level proportion of SCC measurements above 200,000 cells/mL (Prop200) was used as an annual herd level indicator for SCM status. It was calculated as the sum of all test day cow records with an elevated SCC per herd throughout the year of study divided by the sum of all test day cow records per herd performed in the year of study. Annual mean herd level BMSCC was calculated as the geometric mean of all bimonthly BMSCC measurements during the year of observation. Spearman rank correlation coefficients were calculated between all available herd level mastitis indicators to determine their relationships. Calculations were performed for the full lactation. IRCM was also calculated for the first 4 weeks of lactation. All calculations were performed in SAS 9.1 (SAS Institute, Cary, USA) using PROC CORR and PROC GENMOD.

RESULTS Herds with less than five test day recordings (n=8) during the 1-year observation period were deleted from the study, leaving 396 herds available for statistical analyses of SCM. Data collection on CM cases was completed by 205 herds and data from those herds was used for IRCM calculations. CM data were incomplete in the other herds and were not included in the analysis. BMSCC data was available from 387 and 201 herds for the total and CM dataset, respectively. The dataset for SCM compromised 396 dairy herds with 16,571 primiparae and 28,751 multiparae, while records on 205 herds with 9,850 primiparae and 16,029 multiparae were available for analysis of IRCM. In the total dataset (n=396) the average

15

Chapter 2

herd size was 77.4 (SD=25.0) cows and the geometric BMSCC was 176,000 cells/mL. Because cows calved and were culled throughout the study period, the median time primiand multiparae were in the 1-year study period was 183 (min=1; max=365) and 346 (min=1; max=365) days, respectively. The majority of herds had a yearly average test day interval of 4 weeks (73.5%), whereas a five (10.1%) or six (10.9%) weekly interval was less common. The remaining herds had three (1.0%), seven (2.5%), eight (1.8%) or nine (0.3%) weekly intervals. Descriptive observations did not numerically differ between the total dataset and the CM subset of 205 herds. Prevalence of subclinical mastitis Mean herd level SCM prevalence was estimated to be 12.8% (95% CI: 12.2-13.5) in primiparae and 27.1% (95% CI: 26.2-28.1) in multiparae for the default value (Table 1).

Table 1. Estimated within herd level subclinical mastitis prevalence with 95% confidence interval in primi- and multiparae according to 4 different thresholds in 396 Dutch dairy herds from July 1, 2004 to June 30, 2005, based on a negative binomial model with a repeated herd effect and corrected for seasonal variations. Subclinical mastitis prevalence (%) Threshold (cells/mL)

Primiparae

Multiparae

250,000

9.6 (9.1-10.2)

21.5 (20.7-22.4)

200,000

12.8 (12.2-13.5)

27.1 (26.2-28.1)

150,000

18.5 (17.7-19.4)

35.7 (34.5-37.0)

100,000

29.9 (28.8-31.0)

49.5 (48.0-51.1)

Overall (regardless of parity) herd level SCM prevalence was 23.0% (95% CI: 22.2-23.9). Higher herd level SCM prevalences were observed with lower SCC thresholds in both primi- and multiparae (Table 1). One third and half of the primi- and multiparae, respectively, were associated with SCM at a threshold of 100,000 cells/mL (Table 1). Cow level SCM prevalence was the highest in July/August and the lowest in January/February (Figure 1). For the default threshold, cow level SCM prevalence was 34% and 31% in the first week after calving for primi- and multiparae, respectively (Figure 2). SCM prevalence declined steeply in the second week of lactation and increased again slowly in the remaining part of lactation. Almost half of the multiparae had a SCC > 200,000 cells/mL by the end of lactation, while only 15-20% of the primiparae had a SCC > 200,000 cells/mL in that period of lactation (Figure 2). Annual mean herd level Prop200 was

16


12.7% (SD=6.2; range 1.6-44.9) and 26.3% (SD=8.4; range 5.7-59.3) in primi- and multiparae, respectively.

35

SCM prevalence (%)

30 25 20 15 10 5 0 jul-04

sep-04

nov-04

jan-05

mar-04

may-05

Month

Figure 1. Mean cow level subclinical mastitis (SCM) prevalence per calendar month in primiparae (●) and multiparae (▲) for the threshold of 200,000 cells/mL in 396 Dutch dairy herds from July 1, 2004 to June 30, 2005.

SCM prevalence (%)

60 50 40 30 20 10 0 1

6

11

16

21

26

31

36

41

46

51

56

Weeks after calving

Figure 2. Mean cow level subclinical mastitis (SCM) prevalence per week postpartum in primiparae (●) and multiparae (▲) having a lactation stage ≥4 days for the threshold of 200,000 cells/mL in 396 Dutch dairy herds from July 1, 2004 to June 30, 2005.

17

Chapter 2

IRCM per 100 cow-weeks

5 4 3 2 1 0 1

6

11

16

21

26

31

36

41

46

51

56

Weeks after calving

Figure 3. Mean cow level clinical mastitis incidence rate (IRCM; quarter cases per 100 cow-weeks at risk) per week postpartum in primiparae (●) and multiparae (▲) in 205 Dutch dairy herds from July 1, 2004 to June 30, 2005.

Incidence rate of clinical mastitis Mean herd level IRCM was 20.2 (95% CI: 18.3-22.4) and 39.6 (95% CI: 37.1-42.3) (per 100 cow-years at risk) for primi- and multiparae, respectively. Overall (regardless of parity) herd level IRCM was 33.8 (95% CI: 31.7-36.1) (per 100 cow-years at risk). Cow level IRCM in both parity groups was the highest in the first week after calving, decreased considerably in the second week and gradually declined towards the end of lactation (Figure 3). Also, IRCM in primiparae was higher in the first week of lactation compared to IRCM in multiparae, while it was lower for the rest of lactation (Figure 3). After approximately 30 weeks in lactation, IRCM in primi- and multiparae were comparable. IRCM in multiparae seemed to vary more at the herd level than IRCM in primiparae (Figure 4). Mastitis in primiparae compared to other herd level mastitis indicators Increased occurrence of mastitis during the full lactation in primiparae was associated with a higher occurrence of mastitis in multiparae (Table 2). Spearman rank correlation coefficients were 0.450 (P < 0.0001) and 0.429 (P < 0.0001) for Prop200 and IRCM during the full lactation, respectively. However, the association was less strong for IRCM in the first 4 weeks of lactation: Spearman rank correlation coefficient was 0.215 (P = 0.0022). A plot of IRCM in the first 4 weeks of lactation in multiparae versus primiparae (Figure 5) identified that some herds had a high IRCM in primiparae but a low IRCM in multiparae, indicating the potential existence of ‘heifer mastitis problem herds’.

18


40 35

Frequency

30 25 20 15 10 5

0

0

0

0

0 14

13

12

11

90

10

80

70

60

50

40

30

20

10

0

0

IRCM in primiparae 40 35

Frequency

30 25 20 15 10 5

0

0

0

0

0 14

13

12

11

10

90

80

70

60

50

40

30

20

10

0

0

IRCM in multiparae Figure 4. Histograms of the within herd level clinical mastitis incidence rate (IRCM; quarter cases per 100 cow-years at risk) in primiparae (top) and multiparae (bottom) in 205 Dutch dairy herds, from July 1, 2004 to June 30, 2005.

19

0.450**

Prop200_primi7

**

0.020

0.226*

0.092 0.009

1

0.215

0.649 *

0.284

**

**

-0.047

0.073

0.292**

0.694**

0.831**

0.516**

0.429**

-0.018

-0.030

BMSCC6

-0.026

0.111

IRCM 4_multi5

0.084

0.031

IRCM full_primi2 IRCM full_multi3 IRCM 4_primi4

P 200,000 cells/mL in primiparae.

*

IRCM 4_multi

IRCM 4_primi

IRCM full_multi

IRCM full_primi

Prop200_multi

Prop200_multi1

Mastitis indicator

Table 2. Spearman rank correlation coefficients between 7 herd level mastitis indicators in primi- and multiparae in 201 Dutch dairy herds from July 1, 2004 to June 30, 2005.


IRCM - 4 in multiparae

30 25 20 15 10 5 0 0

5

10

15

20

25

30

35

40

45

50

IRCM - 4 in primiparae

Figure 5. Scatter plot of the within herd level clinical mastitis incidence rate in the first 4 weeks of lactation (IRCM-4; quarter cases per 100 cow-first 4 weeks of lactation at risk) in multiparae vs. primiparae.

No CM in the first 4 weeks of lactation in primiparae was observed in 77 (37.6%) herds, while this was 10.7% (n=22) for multiparae (Figure 5). A high Prop200 in primiparae was positively associated with a high BMSCC, but Prop200 in multiparae was more strongly associated with BMSCC (Table 2). Neither IRCM in primiparae nor IRCM in multiparae was associated with BMSCC.

DISCUSSION To estimate SCM prevalence and IRCM in primi- and multiparae in a random sample of dairy herds in the Netherlands, criteria were set for herds to participate in the study. The sample taken was therefore not a true representation of the whole population of dairy herds in the Netherlands. Average BMSCC and herd size were 210,000 cells/mL (MCS Nederland, 2004) and 60.4 (CRV, 2004), respectively, in the Netherlands in 2004, while in this study these were 190,000 cells/mL and 77.4. In addition, farms had to participate in the national test day recording, while nationally 21% of the herds do not participate (CRV, 2004). The estimates on the occurrence of mastitis from this study are nevertheless deemed to be valid for herds that are expected to stay in business for the next 5 yrs, because a random sample of herds was taken from that population. A cow was assumed to have SCM when it had an elevated composite SCC. Although not perfectly, SCC do reflect an inflammatory response (Dohoo and Leslie, 1991; Schukken et al., 2003). The estimates of SCM prevalence are believed to be useful 21

Chapter 2

because different thresholds were evaluated and within herd comparisons were made. Although this obviously affected the estimates, conclusions were not influenced by the threshold chosen. The most commonly used threshold of 200,000 cells/mL (Schukken et al., 2003) was used in further analyses. Additionally, composite SCC are cheap and easily available through the test day recording for research and monitoring purposes, making it possible to analyse large datasets with very precise estimates due to high statistical power. Finally, the proportion of cows with an elevated SCC is commonly used in herd health programmes, making the results from this study easily applicable in practice. Detection bias could have been introduced because farmers diagnose CM differently in their herds. This partly could explain the variation found between herds, although farmers received instructions before the start of the study on how to diagnose CM during the study period. No relationship was previously found between the visual properties of submitted milk samples of CM quarters and the overall IRCM at herd level (Lam et al., 1993; Miltenburg et al., 1996), indicating that differences in IRCM at the herd level were not related to the diagnostic capability of the farmer. The estimates in this study are therefore assumed to be valid. The occurrence of both CM and SCM was generally lower in primiparae compared to multiparae, but IRCM was higher in early lactation in our study, similar to other studies (Miltenburg et al., 1996; Barkema et al., 1998; Valde et al., 2004; Nyman et al., 2007; Olde Riekerink et al., 2008; Nyman et al., 2009). Because primiparae have not been exposed to contagious pathogens during milking before calving, transmission routes of pathogens are different from multiparae. In lactating cows, spreading of pathogens during milking is considered an important transmission route (Lam et al., 1996). Other factors, like the environment (Pullinger et al., 2006) and flies (Nickerson et al., 1995) can be the origin of intramammary infections in all age groups. Although the same pathogens are observed in both parity groups (Fox, 2009), mastitis in primiparae has a different pathological and epidemiological context than mastitis in multiparae. Overall (regardless of parity) IRCM found in our study lay in between estimates from other studies. Of 22 identified studies reporting on overall IRCM since 1998, only 8 studies were performed using a random selection of dairy herds (Bartlett et al., 2001; Beaudeau et al., 2002; Gianneechini et al., 2002; Sviland and Waage, 2002; Valde et al., 2005; Karimuribo et al., 2006; Bradley et al., 2007; Kivaria et al., 2007), while the other studies were performed using non-randomly selected herds. IRCM estimates from nonrandomly selected herds may be inaccurate. This was indicated by a larger variation in overall IRCM in non-randomly selected dairy herds compared to IRCM estimated from randomly selected herds. In Europe, IRCM ranged from 9 (Hamilton et al., 2006) to 94 (Wolfová et al., 2006) (per 100 cow-years at risk) in non-randomly selected herds, while IRCM varied from 36.0 (Valde et al., 2005) to 48.5 (Sviland and Waage, 2002) per 100 cow-years at risk in randomly selected herds. Parity-specific IRCM estimates from our study also lay in between estimates from 11 other studies performed since 1998 reporting

22


on parity-specific estimates (Barkema et al., 1998; Sargeant et al., 1998; Shpigel et al., 1998; Bartlett et al., 2001; Valde et al., 2004; Kalmus et al, 2006; Svensson et al., 2006; Wolfová et al., 2006; Parker et al., 2007; Nyman et al., 2009; Persson Waller et al., 2009). The lowest IRCM in both primi- and multiparae was found in Sweden in 2000 (Svensson et al., 2006), while the highest IRCM in both primi- and multiparae was found in the Czech Republic in 1996-2003 (Wolfová et al., 2006). Besides regional and seasonal differences, different scales, designs and ways of presenting results of studies make it difficult to compare estimates from different studies. Scale of the reported studies (n=22) ranged from 5 (Wolfová et al., 2006) to 23,000 herds (Sviland and Waage, 2002) and indicators for reporting on occurrence of CM varied from incidence rates (number of cases per 100 cow-years at risk) to percentage of cows having one or more CM cases during lactation, and percentage of cows having CM during a 6 weeks observational period. Additionally, only four studies reported 95% confidence intervals (Berry, 1998; McDougall, 1999; Hamilton et al., 2006; Parker et al., 2007) and thus gave information on the precision of the estimates. Most studies were designed to investigate relationships of certain factors with the occurrence of CM, partly explaining the variation in study designs and ways of presenting results. Just a few studies were identified that solely reported on IRCM, while most other studies investigated factors related to IRCM, where IRCM itself was of secondary interest. Some herds had up to twice the population average IRCM in one or both parity groups. Mastitis control programs should focus on these herds, because possible gain is largest in these herds. Herds with a high occurrence of mastitis are generally identified by a high overall (regardless of parity) IRCM or a high BMSCC. However, high BMSCC was not correlated with IRCM in primiparae in our study and IRCM in multiparae did not show a very strong relationship with IRCM in primiparae either, especially in the first 4 weeks of lactation when primiparae are at the highest risk for developing CM. Therefore, for herd and national level monitoring purposes, it is important to report different mastitis indicators. BMSCC, IRCM, and Prop200, should be determined, with the latter two, preferably, reported separately for primi- and multiparae. In particular, some herds showed relatively more IRCM in early lactation in primiparae than in multiparae, suggesting that ‘heifer mastitis problem herds’ can be identified by a parity-specific and lactation stagedependent monitoring of IRCM.

CONCLUSIONS In a random sample of Dutch dairy herds followed for 1 year, herd level SCM prevalence and IRCM were estimated in primi- and multiparae. SCM prevalence and IRCM were approximately two times higher in multiparae, although IRCM was higher in primiparae in the first week postpartum. Some herds showed different mastitis occurrences in their

23

Chapter 2

primi- and multiparae, suggesting that parity-specific monitoring is useful to identify ‘heifer mastitis problem herds’.

ACKNOWLEDGEMENTS This study is part of the 5-year mastitis control program of the Dutch Udder Health Centre and was financially supported by the Dutch Dairy Board. The authors want to thank the farmers for participating in the study and CRV and Qlip for providing composite and bulk milk somatic cell count data.

REFERENCES Barkema, H. W., H. A. Deluyker, Y. H. Schukken, and T. J. G. M. Lam. 1999. Quarter-milk somatic cell count at calving and at the first six milkings after calving. Prev. Vet. Med. 38:1-9. Barkema, H. W., Y. H. Schukken, T. J. G. M. Lam, M. L. Beiboer, H. Wilmink, G. Benedictus, and A. Brand. 1998. Incidence of clinical mastitis in dairy herds grouped in three categories by bulk milk somatic cell counts. J. Dairy Sci. 81:411-419. Bartlett, P. C., J. F. Agger, H. Houe, and L. G. Lawson. 2001. Incidence of clinical mastitis in Danish dairy cattle and screening for non-reporting in a passively collected national surveillance system. Prev. Vet. Med. 48:73-83. Beaudeau, F., C. Fourichon, H. Seegers, and N. Bareille. 2002. Risk of clinical mastitis in dairy herds with a high proportion of low individual milk somatic-cell counts. Prev. Vet. Med. 53:43-54. Berry, E. A. 1998. Mastitis incidence in straw yards and cubicles. Vet. Rec 142:517-518. Bradley, A. J., K. A. Leach, J. E. Breen, L. E. Green, and M. J. Green. 2007. Survey of the incidence and aetiology of mastitis on dairy farms in England and Wales. Vet. Rec. 160:253-258. CRV. 2004. Jaarstatistieken 2004. CRV, Arnhem (in Dutch). De Vliegher, S., H. W. Barkema, H. Stryhn, G. Opsomer, and A. de Kruif. 2004. Impact of early lactation somatic cell count in heifers on somatic cell counts over the first lactation. J. Dairy Sci. 87:3672-3682. De Vliegher, S., H. W. Barkema, H. Stryhn, G. Opsomer, and A. de Kruif. 2005a. Impact of early lactation somatic cell count in heifers on milk yield over the first lactation. J. Dairy Sci. 88:938947. De Vliegher, S., H. W. Barkema, G. Opsomer, A. de Kruif, and L. Duchateau. 2005b. Association between somatic cell count in early lactation and culling of dairy heifers using Cox frailty models. J. Dairy Sci. 88:560-568. Dohoo, I. R., and A. H. Meek. 1982. Somatic cell counts in bovine milk. Can. Vet. J. 23:119-125. Dohoo, I. R., and K. E. Leslie. 1991. Evaluation of changes in somatic cell counts as indicators of new intramammary infections. Prev. Vet. Med. 10:225-237. Fox, L. K. 2009. Prevalence, incidence and risk factors of heifer mastitis. Vet. Microbiol. 134:82-88. Gianneechini, R., C. Concha, R. Rivero, I. Delucci, and J. Moreno López. 2002. Occurence of clinical and sub-clinical mastitis in dairy herds in the West Littoral Region in Uruguay. Acta Vet. Scand. 43:221-230. Hamilton, C., U. Emanuelson, K. Forslund, I. Hansson, and T. Ekman. 2006. Mastitis and related management factors in certified organic dairy herds in Sweden. Acta Vet. Scand. 48:11.

24

Occurrence of mastitis in the Netherlands Huijps, K., S. De Vliegher, T. J. G. M. Lam, and H. Hogeveen. 2009. Cost estimation of heifer mastitis in early lactation by stochastic modelling. Vet. Microbiol. 134:121-127. Kalmus, P., A. Viltrop, B. Aasmäe, and K. Kask. 2006. Occurrence of clinical mastitis in primiparous Estonian dairy cows in different housing conditions. Acta Vet. Scand. 48:21. Karimuribo, E. D., J. L. Fitzpatrick, C. E. Bell, E. S. Swai, D. M. Kambarage, N. H. Ogden, M. J. Bryant, and N. P. French. 2006. Clinical and subclinical mastitis in smallholder dairy farms in Tanzania: Risk, intervention and knowledge transfer. Prev. Vet. Med. 74:84-98. Kivaria, F. M., J. P. T. M. Noordhuizen, and H. M. Msami. 2007. Risk factors associated with the incidence rate of clinical mastitis in smallholder dairy farms in the Dar es Salaam region of Tanzania. Vet. J. 173:623-629. Lam, T. J. G. M., Y. H. Schukken, F. J. Grommers, J. A. H. Smit, and A. Brand. 1993. Within-herd and between-herd variation in diagnosis of clinical mastitis in cattle. J. Am. Vet. Med. Assoc. 202:938-942. Lam, T. J. G. M., M. C. M. de Jong, Y. H. Schukken, and A. Brand. 1996. Mathematical modeling to estimate efficacy of postmilking teat disinfection in split-udder trials of dairy cows. J. Dairy Sci. 79:62-70. McDougall, S. 1999. Prevalence of clinical mastitis in 38 Waikato dairy herds in early lactation. New Zeal. Vet. J. 47:143-149. MCS Nederland. 2004. Jaarverslag 2004. Stichting Melkcontrolestation Nederland, Zutphen (in Dutch). Miltenburg, J. D., D. de Lange, A. P. P. Crauwels, J. H. Bongers, M. J. M. Tielen, Y. H. Schukken, and A. R. W. Elbers. 1996. Incidence of clinical mastitis in a random sample of dairy herds in the southern Netherlands. Vet. Rec. 139:204-207. Nickerson, S. C., W. E. Owens, and R. L. Boddie. 1995. Mastitis in dairy heifers: initial studies on prevalence and control. J. Dairy Sci. 78:1607-1618. Nyman, A. K., T. Ekman, U. Emanuelson, A. H. Gustafsson, K. Holtenius, K. Persson Waller, and C. Hallén Sandgren. 2007. Risk factors associated with the incidence of veterinary-treated clinical mastitis in Swedish dairy herds with a high milk yield and a low prevalence of subclinical mastitis. Prev. Vet. Med. 78:142-160. Nyman, A. K., U. Emanuelson, A. H. Gustafsson, and K. Persson Waller. 2009. Management practices associated with udder health of first-parity dairy cows in early lactation. Prev. Vet. Med. 88:138-149. Olde Riekerink, R. G. M., H. W. Barkema, D. F. Kelton, and D. T. Scholl. 2008. Incidence rate of clinical mastitis on Canadian dairy farms. J. Dairy Sci. 91:1366-1377. Parker, K. I., C. W. R. Compton, F. M. Anniss, A. M. Weir, and S. McDougall. 2007. Management of dairy heifers and its relationship with the incidence of clinical mastitis. New Zeal. Vet. J. 55:208-216. Persson Waller, K., B. Bengtsson, A. Lindberg, A. Nyman, and H. Ericsson Unnerstad. 2009. Incidence of mastitis and bacterial findings at clinical mastitis in Swedish primiparous cows - Influence of breed and stage of lactation. Vet. Microbiol. 134:89-94. Pullinger, G. D., M. López-Benavides, T. J. Coffey, J. H. Williamson, R. T. Cursons, E. Summers, J. Lacy-Hulbert, M. C. Maiden, and J. A. Leigh. 2006. Application of Streptococcus uberis multilocus sequence typing: analysis of the population structure detected among environmental and bovine isolates from New Zealand and the United Kingdom. Appl. Environ. Microbiol. 72:1429-1436. Pyörälä, S. 2003. Indicators of inflammation in the diagnosis of mastitis. Vet. Res. 34:565-578. Sargeant, J. M., H. M. Scott, K. E. Leslie, M. J. Ireland, and A. Bashiri. 1998. Clinical mastitis in dairy cattle in Ontario: Frequency of occurrence and bacteriological isolates. Can. Vet. J. 39:33-38. Schepers, A.J., T. J. G. M. Lam, Y. H. Schukken, J. B. M. Wilmink, and W. J. A. Hanekamp. 1997. Estimation of variance components for somatic cell counts to determine thresholds for uninfected quarters. J. Dairy Sci. 80:1833-1840.

25

Chapter 2 Schukken, Y. H., D. J. Wilson, F. Welcome, L. Garrison-Tikofsky, and R. N. Gonzalez. 2003. Monitoring udder health and milk quality using somatic cell counts. Vet. Res. 34:579-596. Shpigel, N. Y., M. Winkler, G. Ziv, and A. Saran. 1998. Clinical, bacteriological and epidemiological aspects of clinical mastitis in Israeli dairy herds. Prev. Vet. Med. 35:1-9. Steeneveld, W., H. Hogeveen, H. W. Barkema, J. van den Broek, and R. B. M. Huirne. 2008. The influence of cow factors on the incidence of clinical mastitis in dairy cows. J. Dairy Sci. 91:1391-1402. Svensson, C., A. K. Nyman, K. Persson Waller, and U. Emanuelson. 2006. Effects of housing, management, and health of dairy heifers on first-lactation udder health in Southwest Sweden. J. Dairy Sci. 89:1990-1999. Sviland, S., and S. Waage. 2002. Clinical bovine mastitis in Norway. Prev. Vet. Med. 54:65-78. Valde, J. P., L. G. Lawson, A. Lindberg, J. F. Agger, H. Saloniemi, and O. Østerås. 2004. Cumulative risk of bovine mastitis treatments in Denmark, Finland, Norway and Sweden. Acta Vet. Scand. 45:201-210. Valde, J. P., O. Østerås, and E. Simensen. 2005. Description of herd level criteria for good and poor udder health in Norwegian dairy cows. J. Dairy Sci. 88:86-92. Whist, A. C., O. Østerås, and L. Sølverød. 2007. Streptococcus dysgalactiae isolates at calving and lactation performance within the same lactation. J. Dairy Sci. 90:766-778. Wolfová, M., M. Štípková, and J. Wolf. 2006. Incidence and economics of clinical mastitis in five Holstein herds in the Czech Republic. Prev. Vet. Med. 77:48-64.

26

V{tÑàxÜ

F

Quantification of the relationship between high somatic cell counts and subsequent clinical mastitis in dairy cows with two analytical methods

Bart H. P. van den Borne Hans C.M. Vernooij Athumani M. Lupindu Gerdien van Schaik Klaas Frankena Theo J. G. M. Lam Mirjam Nielen

Submitted for publication.

Chapter 3

ABSTRACT The objectives of this study were 1) to quantify the relationship between high composite somatic cell counts (CSCC) in dairy cows and the first subsequent case of clinical mastitis (CM) and 2) to compare the effect estimates for CSCC of a cohort study design with the outcomes of a matched case-control study design. Farmer-diagnosed cases of CM and test day CSCC measurements during 1 year of 13,917 cows in 196 randomly selected Dutch dairy herds were available for analysis. Cows were followed in 1 lactation from the first test day postpartum until CM, drying off, culling or end of study. A Cox proportional hazards model with time-varying CSCC levels was used to estimate the effect of high CSCC (≥200,000 cells/mL) on the time until the first case of CM. A frailty effect was included to adjust for clustering of cows within herds. A matched case-control study design was also applied to the data. A cow with CM was defined a case cow and was matched based on its lactation stage to 4 control cows without CM. CSCC status (high or low) at the test day before CM occurrence was compared to CSCC status from the matched test day in control cows using conditional logistic regression. High CSCC primiparae had a 4 fold higher risk for subsequent CM than low CSCC primiparae; high CSCC multiparae had a 2 fold higher risk. Additionally, multiparae with a low CSCC had a 2 times higher risk for CM occurrence than primiparae with a low CSCC. The risk for CM also increased with increased milk production and in the housing period. Increasing the threshold for high CSCC, showed that the risk for CM increased. If the last CSCC was low, CSCC information of 2 consecutive test days was more predictive than CSCC information from only the last test day, but this was not so when the last CSCC was high. The effect estimates of CSCC from the Cox proportional hazards model were considered less biased than the effect estimates from the conditional logistic regression model and thus are more suitable for use in further analyses. This study identified that approximately 25% of first subsequent CM cases after a CSCC measurement can potentially be prevented when high CSCC cows are prevented or removed from the population. This corresponds with an additional proportion of 4% of cows that do not develop CM after a CSCC measurement. Keywords: clinical mastitis, somatic cell count, survival analysis, conditional logistic regression

28

High somatic cell counts and clinical mastitis

INTRODUCTION Somatic cell counts (SCC) can be high in dairy cows before clinical mastitis (CM) occurs (de Haas et al., 2002). High SCC during lactation or around the dry period are therefore considered to be predictive for the development of CM (Beaudeau et al., 1998; Rupp et al., 2000; Rupp and Boichard, 2000; Peeler et al., 2003; Green et al., 2004a, Whist and Østerås, 2007). Several statistical methods have been used to study the relationship between SCC and CM. Some authors used logistic regression to study the effect of high SCC at a given time point on the occurrence of CM in the same (Beaudeau et al., 1998; Steeneveld et al., 2008) or in the next lactation (Rupp et al., 2000). Others used a lactational mean of SCC (Deluyker et al., 1993; Green et al., 2004b; Steeneveld et al., 2008). A few studies investigated the relationship of SCC around calving with the occurrence of CM later in lactation and accounted for the length of the time period up to CM occurrence using survival analysis (Rupp and Boichard, 2000; Whist and Østerås, 2007). However, these studies only observed SCC around calving and neglected the dynamic nature of SCC. New intramammary infections can occur during lactation resulting in new SCC elevations (Dohoo and Leslie, 1991; Schukken et al., 2003). Peeler et al. (2003) and Green et al. (2004a) used a multilevel study design that incorporated the dynamics of quarter SCC during lactation using time-varying predictors (i.e., quarter SCC measured on a monthly basis) to investigate the relationship between quarter SCC and CM. However, composite SCC (CSCC) measurements are often used for intervention decisions in dairy practice (e.g., antimicrobial treatment, segregation or culling). Also, no corrections were made for the timing of CM occurrence in both aforementioned studies because logistic regression analysis was applied. In spite of quarter level SCC information, this can result in lower statistical power and less accurate and precise estimates compared with survival analysis (Green and Symons, 1983; Annesi et al., 1989; van der Net et al., 2008). The goals of this study were 1) to quantify the relationship between high CSCC and the first subsequent case of CM and 2) to compare the effect estimates from an unmatched cohort study design, using a Cox proportional hazards model with time-varying predictors, with a matched case-control study design using conditional logistic regression.

MATERIALS AND METHODS Data A previously described dataset was available for analysis (van den Borne et al., 2010). In brief, CSCC measurements of the regular test day recording and farmer-diagnosed CM occurrence were collected in an observational study from July 1, 2004 to June 30, 2005 on 29

Chapter 3

205 randomly selected Dutch dairy herds participating in the 4-weekly test day recording (van den Borne et al., 2010). Test day recordings and CM cases were selected from cows that calved within the study period to evaluate the relationship between CSCC and the first subsequent case of CM. Recurrent CM cases were not evaluated. Cows without a test day record preceding a CM event (n=892) and test day records occurring within 4 days after calving were removed from the dataset (Barkema et al., 1999). The first lactation was analyzed when cows had 2 calvings within the study period (n=48 cows). The dataset contained cow level information on stage of lactation, parity, CSCC and production parameters at each test day. Management related factors were extracted from a questionnaire on farmers’ attitude and behaviour sent to the farmers before the start of the study (Jansen et al., 2009) to adjust for the influence of these herd level factors on the association between CSCC and CM. Nine herds were removed from the data because of missing herd level data, resulting in a dataset of 13,917 cows in 196 herds, of which 1,560 cows (11.2%) had a first case of CM in lactation, preceded by a CSCC measurement. This was the initial dataset for both analytical procedures. Survival analysis The initial dataset was used as the basis for the Cox proportional hazards model to evaluate how the hazards of CM differed for cows with low (200 days

30


in lactation to determine the change in the effect estimate of CSCC. Non-significant Pvalues of the proportional hazards test, a straight line in the plot with the scaled Schoenfeld residuals and a similar CSCC effect estimate after censoring of lactations after 200 days indicate proportional hazards. Model fit was determined by plotting the CoxSnell residuals against the cumulative hazard. All model evaluations were performed without the herd level frailty effect included in the Cox proportional hazards models. All survival models were performed in R, version 2.9.0 (R Development Core Team, 2009).

Table 1. Description of variables tested in univariable analyses for association with occurrence of clinical mastitis in 196 Dutch dairy herds. Variable Test day interval level1 Composite somatic cell count (CSCC) ≥200,000 cells/mL Number of test day since calving2 Milk production (kg/day) Protein production (%) Fat production (%) Season

Categories Yes, No 1, 2, 3, 4, 5, 6, 7 and ≥8 Continuous Continuous Continuous Pasture period (May-Oct.) or housing period (Nov.-April)

Cow level Parity 1, ≥2 Herd level Average yearly interval between test days (days) 31 Quartiles of average milking herd size (n cows) 85 Farmer having a antimicrobial treatment protocol Yes, No Cleaning teats before milking Yes, No Wearing gloves during milking Yes, No Fore-stripping cows before milking Yes, No Post-milking teat disinfection Yes, No Preventing cows from laying down after milking Yes, No Milking clinical and subclinical mastitis cows last Yes, No Performing a static milking test >1 each year Yes, No Performed a dynamic milking test in last year Yes, No Frequency of cubicle cleaning per day (n times) 0, 1, 2, 3, 4 and ≥5 Use of blanket dry cow therapy Yes, No Yes, No Antimicrobial treatment of cows with a 1st high CSCC Performing bacteriological culture on high CSCC cows Yes, No Farm type Regular, Organic Bulk milk SCC at the last measurement (·103 cells/mL) ≤150, 151-250 and >250 1 These variables were used as time-varying predictors in the Cox proportional hazards models. 2 Used as a proxy for lactation stage.

31

Chapter 3

Conditional logistic regression In addition to the Cox proportional hazards model, a matched case-control study design was also applied to the initial dataset in which a case was defined to be a cow with a first CM case after at least 1 CSCC measurement. CSCC status of these CM cases was determined at the last test day preceding the first CM observation (td = -1). The preceding test day was evaluated if CM occurred on a test day. A cow was considered to have high CSCC when CSCC was ≥200,000 cells/mL, similar to the definition used in the Cox proportional hazards model. Test day recordings from control cows in the same stage of lactation, without CM in the study, were matched to td = -1 from case cows. Therefore, eligible control cows had to have their calving dates 7 days around the calving date of a case cow. Additionally, the test day recordings from control cows needed to be within 7 days of td = -1 of their case cow and were subsequently also defined td = -1. CSCC status at td = -1 for control cows was determined similar to case cows. The matching procedure corrected for differences in CSCC with different lactation stages. Four control cows from any herd were randomly selected to each case cow to increase precision (Dohoo et al., 2003). Control cows were matched to only 1 case cow. Conditional logistic regression was used with CM occurrence as the binary response variable. All available risk factors (Table 1) were also evaluated in the conditional logistic regression. Lactation stage was included in the analysis to verify if matching did not result in confounding with the risk factor exposure (i.e., CSCC status; Rothman, 1986). Analyses started with univariable and bivariable conditional logistic regressions for all eligible variables (Table 1). CSCC status was forced into all bivariable models to observe the effect of the variable of interest on the parameter estimate of CSCC status. All variables with a P < 0.25 qualified for multivariable analyses, based on type 3 Wald tests. Thereafter, a stepwise backwards multivariable conditional logistic regression was performed based on the reduction in deviance for all observations without any missing values until all variables in the model had a P < 0.05. Observations with missing values for some herd level variables were added to the dataset again when these variables were not included in the model anymore. All models were checked for confounding, which was assumed to occur when estimates changed >25%. Biologically relevant interactions between cow level variables were also tested. Odds ratios (OR) were calculated from the final model by exponentiation of the regression coefficients. No random herd effect could be added to the model due to the matched case-control study design and the accompanying analytical procedure (i.e., conditional logistic regression). Standard errors of herd level variables in the final model may therefore be underestimated resulting in too low P values. Consequently, the herd level variables in the final model were not interpreted or further tested in interaction terms, but were assumed to adjust the effect estimates of CSCC status for a potential herd effect. Conditional logistic regression analysis was performed using proc logistic in SAS 9.2 (SAS Institute, Cary, USA). Goodness of fit of the final conditional logistic regression

32


model was evaluated using the linktest command in STATA version 11 (Statacorp LP, College Station, USA), which produces a predicted value and a predictive value squared for the model. The predicted value should be significant while the predictive value squared should be insignificant for a properly specified model. Sensitivity analyses Different CSCC thresholds were evaluated with both final models. The effect estimates for CSCC status using the thresholds of 100,000 cells/mL; 150,000 cells/mL for primiparae and 250,000 cells/mL for multiparae (the Dutch CSCC thresholds); and 250,000 cells/mL were compared to the effect estimate of the default threshold of 200,000 cells/mL. The deviance and the Akaike Information Criterion (AIC) for the Cox proportional hazards model and the conditional logistic regression model, respectively, were used to determine which threshold gave statistically the best fit. The relationship between different CSCC profiles, based on 2 consecutive test day recordings (td = -1 and td = -2), and CM occurrence was evaluated for both models. Case and control cows without a recording at td = -2 were deleted for this purpose. The deviance and the AIC, based on the same number of observations, were used to determine whether incorporating more test day recordings in the final models gave statistically a better fit in the Cox proportional hazards model and conditional logistic regression model, respectively. The models evaluating different CSCC thresholds and CSCC profiles were run with the same main effects, but without interaction terms for ease of comparison and interpretation. Population attributable fraction The population attributable fraction (PAF) is the proportion of disease events that potentially can be prevented from the total population when a perfect intervention is applied to a certain risk factor (Uter and Pfahlberg, 2001; Dohoo et al., 2003). Because the PAF is a useful measure to identify the possible reduction in CM events in the population, by preventing or removing high CSCC using some intervention (e.g., culling or antimicrobial treatment), it was calculated for both analytical approaches. The PAF according to Eide and Gefeller (1995) was calculated for high CSCC using a readily available SAS macro (Rückinger et al., 2009). Only the cow level variables and the interaction between CSCC status and parity were included in this unconditional logistic regression model due to computational capacity, but the change in PAF estimate for CSCC status was determined when each herd variable was added one at the time to this model. The PAF was calculated differently for the final Cox proportional hazards model. The PAF becomes dynamic with censored time-to-event data as the event rate accumulates over time and exposed subjects are more rapidly depleted from the population than nonexposed subjects (Chen et al., 2006; Samuelsen and Eide, 2008; Cox et al., 2009). For the Cox frailty model, the PAF was calculated as follows (Cox et al., 2009): 33

Chapter 3

PAF (t) =

S * (t ) − S (t ) 1 − S (t )

[1]

with S(t) = ∑j ρjSj(t) the survival function of the total population (both exposed and unexposed individuals) for j strata, S*(t) = ∑j ρj*jSj(t) the resulting survival distribution for j strata when a perfect intervention of the exposure variable (i.e., high CSCC) is assumed to result in an alternative distribution of ρj* (absence of high CSCC), and ρj the proportion of exposed individuals in stratum j. The PAF estimates the reduction in diseased individuals, but interest is mainly in survival of individuals when conducting survival analysis. Cox et al. (2009) therefore introduced a new measure of association to assess the impact of interventions with survival data: the attributable survival (hereafter called the population attributable survival (PAS), in agreement with the PAF). The PAS represents the additional proportion of individuals in the population who survive to a given time, if a fully effective intervention to exposed individuals has been administered at t = 0. For our study, the PAS represents the additional proportion of cows without subsequent CM after a CSCC measurement when preventing or removing high CSCC from the population. The PAS is calculated as follows (Cox et al., 2009):

PAS(t) =

S * (t ) − S (t ) S * (t )

[2]

Both PAF(t) and PAS(t) provide information about the timing of intervention because the survival function is dependent on t (Cox et al., 2009). PAF(t) en PAS(t) calculations were based on the final Cox proportional hazards model without the herd level frailty effect included.

RESULTS In the initial dataset, the median number of cows per herd with a first CM case after a CSCC measurement was 7 (average 8, minimum 1, maximum 28). The median lactation stage at which the first CM case after a CSCC measurement occurred was 86 (average 104, minimum 7, maximum 362) days postpartum. Survival analysis The number of test day intervals per cow varied from 1 to 10, with few (2.6%) cows having more than 10 test day intervals (Figure 1). The majority of the test day intervals 34


(83.5%) were between 26 and 34 days (Figure 2). Survival without CM was the highest in primi- and multiparae with a low CSCC, while survival without CM was the lowest in multiparae with a high CSCC (Figure 3).

1600 1400

Frequency

1200 1000 800 600 400 200 0 1

2

3

4

5

6

7

8

9

10

11

12

13

14

Number of test day intervals

Figure 1. Distribution of the number of test day intervals per cow (n=11,876) in the dataset for the Cox proportional hazards model.

45

Frequency (x 1,000)

40 35 30 25 20 15 10 5 0 10

20

30

40

50

60

Length of test day interval (days)

Figure 2. Distribution of the length of test day intervals without clinical mastitis in the dataset for the Cox proportional hazards model (n = 61,641).

35

Chapter 3

Figure 3. Observed survival without clinical mastitis of 4 groups of cows since their first test day postpartum with different composite somatic cell count (CSCC) status and parity (From top to bottom: primiparae with a low CSCC (< 200,000 cells/mL); multiparae with a low CSCC; primiparae with a high CSCC (≥ 200,000 cells/mL); and multiparae with a high CSCC).

The results from the final Cox proportional hazards model are shown in Table 2 and are based on 62,742 test day intervals, of which 1,280 (2.0%) had a first case of CM. A high CSCC was associated with a higher hazard for a first subsequent CM and was the highest for primiparae compared with multiparae (HR = 4 and HR = 2 for primi- and multiparae, respectively). Multiparae with a low CSCC had a 2 fold higher hazard for CM than primiparae with a low CSCC. Higher milk production levels were associated with increasing CM hazards, especially in cows with a high CSCC. CM occurrence was higher in the housing period than in the pasture period. No herd level variables were significant in the final Cox frailty model. Variance of the herd level frailty effect was 0.37 (P < 0.0001). The test for the overall proportional hazards, based on the scaled Schoenfeld residuals, was significant (P = 0.04) indicating non-proportional hazards, although the individual tests for each variable in the final model were not significant (P > 0.05). The hazards were proportional for the first 200 days in lactation but the hazards became non-proportional after 200 days as observed for CSCC status and milk production by plotting the scaled Schoenfeld residuals against the survival time. Censoring of cows at 200 days in lactation resulted in a non-significant proportion hazards test (P = 0.18).

36

Category No Yes Yes 1 ≥2 ≥2 Housing Pasture Continuous

Milk production (kg/day) In low CSCC cows In high CSCC cows Herd frailty effect Variance of the herd frailty effect was 0.37 (P < 0.0001) 1 Composite somatic cell count at the first test day within a test day interval.

Variable High CSCC1 (≥200,000 cells/mL) In primiparae In multiparae Parity For low CSCC cows For high CSCC cows Season 1.2 2.4

18,710 42,970 23,455 2.5 39,287 1.8 mean=31.1, SD=8.0

% CM 1.5 5.5

Frequency 54,053 8,689

1.02 1.04

Hazard ratio 1.0 4.0 2.1 1.0 1.9 1.0 1.2 1.0 1.00 1.03

1.6 0.8 1.0

2.5 1.3

1.03 1.05

2.4 1.3 1.3

6.5 3.6

95% confidence interval Lower Upper

Table 2. Factors associated with the first case of clinical mastitis (CM; n=1,280) after a composite somatic cell count measurement in the final multivariable Cox proportional hazards model with time-varying predictors.

Chapter 3

Figure 4. Estimated population attributable fraction (top) and population attributable survival (below) for clinical mastitis in cows with a high composite somatic cell count.

38


However, the final Cox proportional hazards model was assumed to be representative for the whole lactation because only a 15% increase in the effect estimate for CSCC was observed when lactations were censored at 200 days. Plotting the Cox-Snell residuals against the cumulative hazards gave a straight line with an intercept of 0 and a slope of 1, indicating a proper model fit (Dohoo et al., 2003). The PAF(t) and PAS(t) for the final Cox proportional hazards model are plotted in Figure 4. The PAF starts at 0.29 at the beginning of the lactation and is monotone decreasing to 0.25 towards the end of the lactation (300 days). The PAS starts at 0 directly after calving and is increasing towards 0.04 at the end of lactation. Hence, an additional 4% of cows do not develop a first subsequent CM case after a CSCC measurement during their lactation, resulting in a reduction of cows with a first case of CM during their lactation from 11% to 7%. Conditional logistic regression The data analyzed in the conditional logistic regression consisted of 1,560 CM cows and 6,240 selected control cows. There were no herd mates within 1,479 matched case and control sets while in 81 matched sets there was 1 herd mate among the controls for the conditional logistic regression model, indicating that only a very small proportion (5.2%) in matched sets came from the same herd. High CSCC at td = -1 was observed in 41.3% of the CM cows while high CSCC were observed in 15.3% of the control cows. Results from the final conditional logistic regression model are presented in Table 3 and are based on 7,684 observations (including 1,538 cases) due to missing values for some herd level variables. All cow level variables significant in the final Cox proportional hazards model were also significantly associated with CM in the conditional logistic regression. Primi- and multiparae with a high CSCC at td = -1 had a higher odds (OR = 6 and OR = 4 for primi- and multiparae, respectively) for CM occurrence than cows with a low CSCC. CM was found more frequently in multiparae than in primiparae when they had a low CSCC (OR = 2). Crude proportions of CM occurrence were 35.5% and 42.7% in primi- and multiparae with a high CSCC, respectively, and were 8.5% and 18.3% for primi- and multiparae with a low CSCC. A linear increase in milk production was associated with a log linear increase in CM occurrence. In contrast to the Cox proportional hazards model, CM occurred slightly more frequent in the pasture period than in the housing period. Four herd level variables were significant in the final conditional logistic regression model. The number of test day recordings since calving, as a proxy for stage of lactation, was not associated with CM occurrence, indicating proper matching of case and control cows. The AIC of the final model was 4,328. Rerunning the final model in STATA with the linktest procedure gave a significant (P < 0.001) value for the predicted value, while the predicted value squared was insignificant (P = 0.59), indicating a proper fit of the model. The PAF for high CSCC was 0.221 and 0.166 in primi- and multiparae, respectively. 39

Category No Yes Yes 1 ≥2 ≥2 Pasture Housing Continuous

11.9 23.2

2,169 5,515 2,588 20.6 5,096 19.7 mean=32.2, SD=8.2

% CM 15.3 41.3

Frequency 6,299 1,385

Odds ratio 1.0 6.1 3.6 1.0 1.9 1.1 1.0 0.6 1.03 NI2 NI NI NI 1.5 0.8

4.5 3.1

0.9 1.04

2.3 1.5

8.3 4.2

95% confidence interval Lower Upper

0.5 Milk production (kg) 1.02 Average milking herd size Post-milking teat disinfection Treatment of cows with a 1st high CSCC Performing a dynamic milking test 1 Composite somatic cell count. 2 Not Interpreted. Herd level variables were not interpreted because they were assumed to correct cow level estimates for clustering within herds.

Variable High CSCC1 (≥200,000 cells/mL) In primiparae In multiparae Parity For low CSCC cows For high CSCC cows Season

Table 3. Factors associated with the first case of clinical mastitis (CM; n=1,538) after a composite somatic cell count measurement in the final multivariable conditional logistic regression model.

100,000 cells/mL 200,000 cells/mL Variable Class % CM HR 95% CI % CM HR 95% CI Cox proportional hazards model CSCC status High 3.7 3.2 2.8 3.5 5.5 4.1 3.7 4.6 Low 1.4 1.0 1.5 1.0 Deviance 21,810 21,711 % CM OR 95% CI % CM OR 95% CI Conditional logistic regression model CSCC status High 32.3 3.0 2.7 3.4 41.3 4.0 3.5 4.6 Low 14.3 1.0 15.4 1.0 AIC2 4,418 4,336 1 150,000 cells/mL for primiparae and 250,000 cells/mL for multiparae (the Dutch thresholds). 2 Akaike Information Criteria. 41.7 15.7

% CM

5.6 1.5

4.2 3.7 1.0 4,325

4.9

4.4 3.9 4.9 1.0 21,695 OR 95% CI

150,000 / 250,0001 % CM HR 95% CI

45.0 15.7

% CM

6.3 1.5

4.5 3.9 1.0 4,317

5.2

4.6 4.1 5.2 1.0 21,682 OR 95% CI

250,000 cells/mL % CM HR 95% CI

Table 4. Hazard ratios (HR), odds ratios (OR) and 95%-confidence intervals (95%-CI) for clinical mastitis (CM) in dairy cows after a composite somatic cell count (CSCC) measurement according to 4 CSCC thresholds in the final Cox proportional hazards model with time-varying predictors (top) and the final conditional logistic regression model (below). The interaction term between CSCC status and parity and milk production were excluded from these models. Other variables in the final models (Tables 2 and 3) were not reported.

Chapter 3

The maximum change in PAF for CSCC status was 5.3% when herd level variables were added to the model one at the time, indicating a robust PAF estimate. Sensitivity analysis According to both final models, higher probabilities of CM were observed with increasing CSCC thresholds in cows with a high CSCC, while this was relatively stable in low CSCC cows (Table 4). Consequently, the risk for CM increased with increasing CSCC thresholds. The estimates of the other variables in the model did not change. The lowest deviance in the Cox proportional hazards models and the lowest AIC in the conditional logistic regression models were observed for the threshold of 250,000 cells/mL, indicating that this CSCC threshold gave statistically the best fit. Subsets of data were created to evaluate CSCC information from 2 subsequent measurements and consisted of 1,121 case and 4,450 control cows for the conditional logistic regression analysis and 939 and 53,464 test day intervals with and without CM, respectively, for the Cox proportional hazards model. The risk for CM occurrence was the highest if cows had a high CSCC at td = -1, independent from CSCC at td = -2 (Table 5). When CSCC was low at td = -1, cows with an elevated CSCC at td = -2 had a higher risk for CM occurrence than cows with 2 consecutive low CSCC. The models incorporating CSCC status at td = -2 had statistically a better fit than the models only evaluating CSCC status at td = -1. The deviance and AIC were 15,619 and 3,017 for the Cox proportional hazards model and the conditional logistic regression model, respectively, when CSCC was also evaluated at td = -2, while they were 15,641 and 3,031 when CSCC was only evaluated at td = -1.

Table 5. Hazard ratios (HR), odds ratios (OR) and the 95%-confidence intervals (95%-CI) for clinical mastitis (CM) for 4 composite somatic cell count (CSCC) profiles at the 2 test days before CM occurrence (td = -1 and td = -2) in the final Cox proportional hazards model and the conditional logistic regression model. The threshold to define high CSCC was 200,000 cells/mL. The models did not include the interaction term between CSCC status and parity. Other variables significant in the final models (Tables 2 and 3) were not reported. CSCC status at td = -2 High Low High Low

42

CSCC status at td = -1 High High Low Low

Cox proportional hazards model with time-varying predictors % CM HR 95% CI 4.8 4.2 3.6 5.0 4.9 4.0 3.3 4.9 2.6 1.9 1.5 2.5 1.3 1.0

Conditional logistic regression model % CM OR 95% CI 41.7 4.0 3.3 4.9 41.3 3.8 3.0 4.8 25.0 2.0 1.5 2.7 15.0 1.0


DISCUSSION This study aimed to quantify the relationship between high CSCC cows and the first CM case later in lactation using Cox proportional hazards models and conditional logistic regression models. Both statistical approaches identified high CSCC cows to have a higher risk to develop CM during lactation, similar to other studies (Deluyker et al., 1993; Beaudeau et al., 1998; Rupp and Boichard, 2000; Peeler et al., 2003; Green et al., 2004a; Green et al., 2004b; Steeneveld et al., 2008). The effect estimates from the current study were consistent with the quarter level estimates from 2 British studies also incorporating time-varying predictors (Peeler et al., 2003; Green et al., 2004a). Also, the identified risk factors behaved as determined by others (Rupp et al., 2000; Rupp and Boichard, 2000; Peeler et al., 2003; Green et al., 2004a; Whist and Østerås, 2007; Steeneveld et al., 2008). The opposite effect of season in the conditional logistic regression compared with the Cox proportional hazards model was probably due to the matching procedure. Case and control cows were in the same season in 85% of the matched groups, indicating limited variation within matching groups due to overmatching (case and control cows calved within 7 days of each other). Several study designs and statistical approaches have been applied in other studies to quantify the relationship between high CSCC and subsequent CM. This makes it complicated to select the most accurate and precise estimates for e.g. economical evaluations of the effect of risk factors on disease occurrence. We therefore compared the estimates from a Cox proportional hazards model with the estimates from a conditional logistical regression model. It was shown earlier in human medical literature, using simulation studies, that Cox proportional hazards models resulted in more accurate and precise effect estimates than logistic regression models (Green and Symons, 1983; Abbott, 1985; Annesi et al., 1989). In addition, statistical power is higher for Cox proportional hazards models than for logistic regression models (Annesi et al., 1989; van der Net et al., 2008) because the former take the time until event into account, where logistic regression do not, thereby giving equal weight to ‘early’ and ‘late’ events (Green and Symons, 1983). Nevertheless, the effect estimates and the accompanying standard errors from Cox proportional hazards models became similar to the model outcomes from logistic regression models when the follow-up period was short and disease occurrence was low (Green and Symons, 1983; Abbott, 1985; Annesi et al., 1989). In our study, similar effect estimates and standard errors were expected between both statistical approaches because the follow-up period and disease occurrence were relatively short (5% (Beaudeau and Fourichon, 1998), where Cox proportional hazards models are known to give unbiased estimates (Cox, 1972). In our study, 11.2% of the cows developed CM after a CSCC measurement and, hence, an overestimation of the true effect can be expected in the logistic regression model. We therefore suggest that the effect estimates from the Cox proportional hazards model are to be used for further analyses (e.g. economic evaluations). This study only evaluated the time from a first CSCC measurement to the first subsequent CM case in the same lactation. Hence, CM cases that occurred directly after calving and were not preceded by a CSCC measurement were not investigated in the current study. This reduced the proportion of CM cows to 11.2%, while the proportion was 16.6% when all cows with a calving date in the study period were included (892 CM cows were deleted because CM was not preceded by a CSCC measurement), indicating that approximately one third of the CM cows were not covered by the current investigation. Management measures other than intervention of high test day CSCC are needed to improve udder health in early lactating cows and may additionally affect subclinical mastitis developing into CM. Additionally, recurrent CM cases were not evaluated and might have affected the magnitude of the effect estimate of CSCC. Furthermore, recurrent CM cases will probably also not appear when the first CM case is prevented because they most likely originate from the same intramammary infection, resulting in an underestimation of the current PAF estimation. We used cow level data to quantify the relationship between high SCC and subsequent CM. This may have resulted in other estimates compared to quarter level data. Cows can have a high SCC in 1 quarter, but low SCC in the other quarters, resulting in a low SCC at cow level. These cows are probably at a higher risk to develop CM in that quarter but they are not identified as high CSCC at cow level, resulting in an underestimation of the high SCC effect. However, the effect estimates at quarter level approximate the effect estimates at cow level due to the log linear nature of SCC. Because the cow level effect estimates from our study were comparable to the quarter level estimates from Peeler et al. (2003) and Green et al. (2004a) and because generally only CSCC data are available in dairy

44


practice, we conclude that both may be used in dairy practice and for economic evaluations. As expected, the risk for CM occurrence increased with higher CSCC thresholds. A higher CSCC is generally assumed to be more closely related to an intramammary infection, thereby resulting in a higher probability to develop a CM event. Interestingly, the interaction term between CSCC status and parity was neither significant at the thresholds of 100,000 cells/mL nor at the thresholds used in the Netherlands (150,000 cells/mL for primiparae and 250,000 cells/mL for multiparae; results not shown). This may indicate that CM risk does not differ between parity groups at the CSCC threshold of 100,000 cells/mL and that the parity effect is corrected away at the parity-specific CSCC thresholds of 150,000 cells/mL and 250,000 cells/mL. Different CSCC profiles were evaluated in the sensitivity analysis. It was identified that the hazard and odds for CM occurrence were similar for cows with a high CSCC at td =-1, regardless of their CSCC status at td =-2. When cows had a low CSCC at td =-1, there was a risk difference between cows with a different CSCC status at td =-2. Similar conclusions were drawn from additional analyses that also incorporated the third (td = -3) CSCC measurement before CM occurrence (results not shown). Hence, it seems sufficient to only evaluate the most recent CSCC measurement in dairy practice when predicting the risk of CM occurrence, except when the latest CSCC is low. Then, CSCC must also be evaluated at td =-2, if available. Although it is widely known that high CSCC are predictive for CM occurrence in dairy cows, CM cases also occur in low CSCC cows. Differences between CM in high and low CSCC cows are likely to be pathogen specific (Peeler et al., 2003; de Haas et al., 2004), indicating that eliminating high CSCC will not remove all CM cases from the dairy population. Therefore, the PAF and PAS were calculated to quantify the proportion of CM cases prevented. It was assumed for PAF and PAS estimation that a perfectly effective intervention occurred at td = -1, but interventions on high CSCC are often postponed in dairy practice (the results of the sampling are not directly available at td = -1) or imperfect in dairy practice. Nevertheless, given these assumptions, PAF estimations identified that a substantial proportion (25%) of cows up to 300 days in lactation can be prevented from subsequent CM when high CSCC are prevented or eliminated at an early stage. This corresponds with an additional gain of 4% of cows at 300 days in lactation that do not develop a first subsequent CM case, when preceded by a CSCC measurement. The magnitude of PAF and PAS stress the importance of prevention and early intervention of high CSCC cows to reduce the impact of CM in dairy herds. This population effect is even larger if it is considered that many cows with a first case of CM also develop one or more repeated CM cases.

45

Chapter 3

CONCLUSIONS Primi- and multiparae with a high CSCC had a higher risk for subsequent CM than low CSCC cows. Multiparae with a low CSCC had a higher risk for CM than primiparae with a low CSCC. The effect estimates for CSCC according to the Cox proportional hazards model differed from the effect estimates from the conditional logistical regression model. Estimates from the survival analysis are to be used for further analyses because those are considered less biased compared with the estimates from the conditional logistic regression. Calculation of the PAF and the PAS identified that a substantial reduction in CM occurrence during lactation can be achieved in dairy herds when high CSCC are prevented or early removed from the population.

ACKNOWLEDGEMENTS This study is part of the 5-year mastitis control program of the Dutch Udder Health Centre and was financially supported by the Dutch Dairy Board. The preliminary work of Shafqat Mehmood and Giske van Es was highly appreciated by the authors.

REFERENCES Abbott, R.D. 1985. Logistic regression in survival analysis. J. Epidemiol. 121:465-471. Annesi, I., T. Moreau, and J. Lellouch. 1989. Efficiency of the logistic regression and Cox proportional hazards models in longitudinal studies. Stat. Med. 8:1515-1521. Barkema, H. W., H. A. Deluyker, Y. H. Schukken, and T. J. G. M. Lam. 1999. Quarter-milk somatic cell count at calving and at the first six milkings after calving. Prev. Vet. Med. 38:1-9. Beaudeau, F., and C. Fourichon. 1998. Estimating relative risk of disease from outputs of logistic regression when the disease is not rare. Prev. Vet. Med. 36:243-256. Beaudeau, F., H. Seegers, C. Fourichon, and P. Hortet. 1998. Association between milk somatic cell counts up to 400,000 cells/mL and clinical mastitis in French Holstein cows. Vet. Rec. 143:685687. Chen, Y. Q., C. Hu, and Y. Wang. 2006. Attributable risk function in the proportional hazards model for censored time-to-event. Biostatistics 7:515-529. Cox, C., H. Chu, and A. Muñoz. 2009. Survival attributable to an exposure. Stat. Med. 28:3276-3293. Cox, D. R. 1972. Regression models and life tables (with discussion). J. R. Stat. Soc. Series B 34:187220. de Haas, Y., H. W. Barkema, and R. F. Veerkamp. 2002. The effect of pathogen-specific clinical mastitis on the lactation curve for somatic cell count. J. Dairy Sci. 85:1314-1323. de Haas, Y., R. F. Veerkamp, H. W. Barkema, Y. T. Gröhn, and Y. H. Schukken. 2004. Associations between pathogen-specific cases of clinical mastitis and somatic cell count patterns. J. Dairy Sci. 87:95-105. Deluyker, H. A., J. M. Gay, and L. D. Weaver. 1993. Interrelationships of somatic cell count, mastitis, and milk-yield in a low somatic cell count herd. J. Dairy Sci. 76:3445-3452.

46

High somatic cell counts and clinical mastitis Dohoo, I. R., and K. E. Leslie. 1991. Evaluation of changes in somatic cell counts as indicators of new intramammary infections. Prev. Vet. Med. 10:225-237. Dohoo, I. R., S. W. Martin, and H. Stryhn. 2003. Veterinary Epidemiologic Research. Atlantic Veterinary College Inc., Charlottetown, Prince Edward Island, Canada. Eide, G. E., and O. Gefeller. 1995. Sequential and average attributable fractions as aids in the selection of preventive strategies. J. Clin. Epidemiol. 48:645-655. Green, M. J., P. R. Burton, L. E. Green, Y. H. Schukken, A. J. Bradley, E. J. Peeler, and G. F. Medley. 2004a. The use of Markov chain Monte Carlo for analysis of correlated binary data: patterns of somatic cells in milk and the risk of clinical mastitis in dairy cows. Prev. Vet. Med. 64:157-174. Green, M. J., L. E. Green, Y. H. Schukken, A. J. Bradley, E. J. Peeler, H. W. Barkema, Y. de Haas, V. J. Collis, and G. F. Medley. 2004b. Somatic cell count distributions during lactation predict clinical mastitis. J. Dairy Sci. 87:1256-1264. Green, M. S., and M. J. Symons. 1983. A comparison of the logistic risk function and the proportional hazards model in prospective epidemiologic studies. J. Chronic Dis. 36:715-724. Jansen, J., B. H. P. van den Borne, R. J. Renes, G. van Schaik, T. J. G. M. Lam, and C. Leeuwis. 2009. Explaining mastitis incidence in Dutch dairy farming: The influence of farmers' attitudes and behaviour. Prev. Vet. Med. 92:210-223. Peeler, E. J., M. J. Green, J. L. Fitzpatrick, and L. E. Green. 2003. The association between quarter somatic-cell counts and clinical mastitis in three British dairy herds. Prev. Vet. Med. 59:169180. Rothman, K. J. 1986. Modern epidemiology. Little, Brown and Company, Boston/Toronto. Rückinger, S., R. von Kries, and A. M. Toschke. 2009. An illustration of and programs estimating attributable fractions in large scale surveys considering multiple risk factors. BMC Med. Res. Methodol. 9:7. Rupp, R., F. Beaudeau, and D. Boichard. 2000. Relationship between milk somatic-cell counts in the first lactation and clinical mastitis occurrence in the second lactation of French Holstein cows. Prev. Vet. Med. 46:99-111. Rupp, R., and D. Boichard. 2000. Relationship of early first lactation somatic cell count with risk of subsequent first clinical mastitis. Livest. Prod. Sci. 62:169-180. Samuelsen, S. O., and G. E. Eide. 2008. Attributable fractions with survival data. Stat. Med. 27:14471467. Schukken, Y. H., D. J. Wilson, F. Welcome, L. Garrison-Tikofsky, and R. N. Gonzalez. 2003. Monitoring udder health and milk quality using somatic cell counts. Vet. Res. 34:579-596. Steeneveld, W., H. Hogeveen, H. W. Barkema, J. van den Broek, R. B. M. Huirne. 2008. The influence of cow factors on the incidence of clinical mastitis in dairy cows. J. Dairy Sci. 91:1391-1402. Uter, W., and A. Pfahlberg. 2001. The application of methods to quantify attributable risk in medical practice. Stat. Methods Med. Res. 10:231-237. van den Borne, B.H.P., G. van Schaik, T. J. G. M. Lam, and M. Nielen. 2010. Variation in herd level mastitis indicators between primi- and multiparae in Dutch dairy herds. Prev. Vet. Med. 96:4955. van der Net, J. B., A. C. J. W. Janssens, M. J. C. Eijkemans, J. J. P. Kastelein, E. J. G. Sijbrands, and E. W. Steyerberg. 2008. Cox proportional hazards models have more statistical power than logistic regression models in cross-sectional genetic association studies. Eur. J. Hum. Genet. 16:11111116. Whist, A. C., and O. Østerås. 2007. Associations between somatic cell counts at calving or prior to drying-off and clinical mastitis in the remaining or subsequent lactation. J. Dairy Res. 74:66-73.

47

V{tÑàxÜ

G

Therapeutic effects of antimicrobial treatment during lactation of recently acquired bovine subclinical mastitis: Two linked randomized field trials

Bart H. P. van den Borne Gerdien van Schaik Theo J. G. M. Lam Mirjam Nielen

Journal of Dairy Science (2010) 93:218-233

Chapter 4

ABSTRACT Two linked randomized field trials were performed on 39 herds in the Netherlands to 1) determine therapeutic effects of antimicrobial treatment of recently acquired subclinical mastitis (RASCM) during lactation, 2) to evaluate the effect of duration of SCM on therapeutic outcome, and 3) to identify factors related to therapeutic success of RASCM. Cows with a first elevated composite somatic cell count (CSCC) after two consecutive low CSCC measurements were eligible for enrollment in trial 1 (treatment at the first elevated CSCC). Quarter milk samples were collected to determine bacteriological status for major pathogens and coagulase-negative staphylococci. Cows with one or more culture-positive quarters with a quarter somatic cell count (QSCC) ≥ 100,000 cells/mL were defined to have RASCM and were randomly assigned treatment or control (no treatment). Untreated cows from trial 1 that had a second elevated CSCC at the next milk recording were eligible for enrollment in trial 2 (treatment at the second elevated CSCC). In trial 2, staphylococcipositive cows (Staphylococcus aureus and coagulase-negative staphylococci) were randomly assigned to treatment or control. Farmers used their own treatment protocol to treat quarters in both trials. Bacteriological cure was defined as absence of the pathogen identified pre-intervention in two samples post-intervention; QSCC, CSCC, and milk yield were also analyzed. Hierarchical logistic and linear models were used to determine therapeutic effects and to identify factors related to therapy outcome. Treated quarters had a higher bacteriological cure rate than control quarters for all pathogens in both trials. Treatment resulted in lower QSCC and CSCC, whereas milk yield was not affected by treatment. Bacteriological cure of RASCM was better in quarters with a low QSCC preintervention and in coagulase-negative staphylococci-positive quarters. Control quarters with a single culture-positive sample pre-intervention also had a higher bacteriological cure than control quarters with ≥2 culture-positive samples. Time of antimicrobial treatment affected bacteriological cure for penicillin-sensitive Staph. aureus. Bacteriological cure tended to be higher for Staph. aureus after treatment at the first elevated CSCC compared to treatment at the second elevated CSCC. Thus, early treatment of Staph. aureus might be more effective than later treatment. Keywords: subclinical mastitis, antimicrobial treatment, randomized clinical trial

50

Antimicrobial treatment of subclinical mastitis

INTRODUCTION Subclinical mastitis (SCM) leads to decreased milk yield, increased probabilities of culling and clinical mastitis (Reksen et al., 2006, 2007), and an increased somatic cell count. Additionally, SCM can be a source of infection for other cows because pathogens can be transmitted between cows (Lam et al., 1996; Zadoks et al., 2001). Treating SCM with antimicrobials during lactation is one of the options to improve udder health in a dairy herd. Economic calculations, using deterministic (Swinkels et al., 2005) and stochastic (Steeneveld et al., 2007) models, have shown that antimicrobial treatment of SCM during lactation can be beneficial. Several clinical trials have been conducted to estimate efficacy of antibiotic treatment of SCM during lactation (Sol et al., 1997; McDougall, 1998; Wilson et al. 1999; St. Rose et al. 2003; Oliver et al., 2004; Deluyker et al., 2005; Salat et al., 2008; Sandgren et al., 2008). Pathogen and treatment factors, such as duration of treatment, pathogen species, and antimicrobial susceptibility affect bacteriological cure rates of SCM (Barkema et al., 2006). Additionally, several cowlevel factors affected treatment success of chronic SCM. Parity, number of quarters infected, lactation stage, and location of affected quarter were associated with cure of Staphylococcus aureus SCM (Sol et al., 1997; Deluyker et al., 2005; Salat et al., 2008). Early treatment of SCM is generally believed to improve therapeutic success (Barkema et al., 2006). Bacteriological cure seemed to be related to the length of infection in a pilot study on cows in early lactation (Beggs and Wraight, 2006) and in experimentally infused animals (Milner et al., 1997; Owens et al., 1997). Also, consecutive culture-positive samples were associated with reduced bacteriological cure of SCM after treatment in lactation (Sol et al., 1997) and at drying off (Sol et al., 1994; Dingwell et al., 2003). Repeated culturing of a pathogen is advocated to identify SCM pre treatment in field studies (Barkema et al., 2006). However, using consecutive samples to identify SCM is not regularly applied in dairy practice because it is expensive and impractical. In stead, composite somatic cell counts (CSCC) from the milk recording, in combination with a single bacteriological culture, are commonly used to identify SCM in dairy herds worldwide (Schukken et al., 2003). Changes in CSCC reflect new SCM cases and are of diagnostic value in identifying new SCM cases eligible for treatment, but have a low sensitivity (Dohoo and Leslie, 1991). Increasing the duration of elevated CSCC increases the diagnostic ability to identify SCM (Dohoo and Leslie, 1991) but will most likely decrease the probability of cure due to chronicity (Barkema et al., 2006). Thus, there is a need to evaluate the use of CSCC as a diagnostic tool to treat recently acquired subclinical mastitis (RASCM). The objectives of this study were to 1) determine therapeutic effects of RASCM after antimicrobial treatment during lactation, 2) evaluate the effect of duration of SCM on therapeutic outcome, and 3) identify factors related to therapeutic success of RASCM.

51

Chapter 4

MATERIALS AND METHODS Two linked randomized, clustered clinical trials were conducted. Both trials were based on CSCC information from the milk recording to identify RASCM. The first trial randomly allocated antimicrobial treatment to staphylococci and streptococci RASCM identified after a first elevated CSCC. The second trial randomly allocated antimicrobial treatment to staphylococci SCM after the second elevated CSCC. The primary outcome analyzed was bacteriological cure at the quarter level. Quarter somatic cell count (QSCC), CSCC, and milk yield at the cow level were also analyzed in both trials to evaluate practical relevancy. Herds The trials were performed from December 2006 through May 2008 in 40 Dutch dairy herds. Herds with a high number of new SCM cases were targeted. To be included, herds had to 1) participate in the DHI program at 4-weekly intervals, 2) have more than 50 cows, 3) have an average incidence of first elevated CSCC of more than 10% in the milk recording in the year preceding the start of the trials, 4) milk without an automatic milking system, and 5) have reliable record-keeping capabilities. This information was checked in herd records and by contacting the farmers’ veterinarians. Herds were enrolled in December 2006 or January 2007 and data collection ended between October 2007 and May 2008, depending on the farm. In 3 herds, data collection was ceased in October 2007 because of the installation of an automatic milking system (n=1) or an irregular interval of the dairy herd improvement program (n=2). In 19 herds, the last cows were enrolled in January 2008, whereas in 17 herds with a high incidence of Staph. aureus SCM, the last cows were enrolled in March 2008 to obtain the required sample size for this pathogen. Median bulk milk somatic cell count (SCC) of the participating herds was 231,000 cells/mL in the month of enrollment and ranged from 123,000 cells/mL to 365,000 cells/mL. Mean herd size was 87.5 (SD=37.1). Average herd-level 305-days milk production was 8,704 kg (SD=894) and breed of cows was mainly Holstein-Friesian. All cows were milked twice daily and were housed in freestall barns with cubicles. In most herds (31/39), cows were on pasture during summer (April/May to September/October). Udder cleaning was practiced before cluster attaching in all but one herd, and pre-stripping was carried out in 28 herds. Postmilking teat disinfection was carried out in 38 herds, and mastitic cows were milked last in 8 herds. All but 3 herds used blanket dry-cow treatment; in 3 herds an internal teat sealant was used in addition. In 2 herds selective dry-cow treatment was used. One herd farmed under organic conditions and cows were not dried off but milked continuously until the next calving. In this particular herd, some groups of cows were housed on straw bedding and teats of lactating cows were not disinfected after milking.

52

Antimicrobial treatment of subclinical mastitis

Inclusion and exclusion criteria A multiparous cow was assumed to have RASCM during lactation (and thus was eligible for enrollment) when at the 4-weekly milk recording, CSCC was ≥250,000 cells/mL after 2 consecutive milk recordings