Report Research Internship W. Dohmen

FACULTEIT DIERGENEESKUNDE / FACULTY OF VETERINARY MEDICINE DEPARTMENT FARM ANIMAL HEALTH, EPIDEMIOLOGY AND ECONOMICS WIETSKE DOHMEN RESEARCH INTERNSHIP MAY-OCTOBER 2009 SUPERVISOR: HENK HOGEVEEEN IN COLLABORATION WITH WAGENINGEN UR, LIVESTOCK RESEARCH RELATIONSHIP BETWEEN UDDER HEALTH AND HYGIENE ON FARMS WITH AN AUTOMATIC MILKING SYSTEM Table of contents Abstract Introduction Materials and Methods Results Discussion 12 Conclusion 14 Acknowledgement 14 References 15 Appendices 17 Abstract Poor hygiene is an important risk factor for reduced udder health On farms with an automatic milking system, this might even be more important because of the automatic cleaning of the udder The aim of this study is to identify the relationship between hygiene and udder health on farms with an AMS on farm as well as the cow level Information on hygiene and udder health was collected on 151 Dutch dairy farms with an AMS Data collection consisted of a partially open-ended questionnaire, a scoring protocol and data from the Dutch national milk recording system Stepwise general linear models were used to analyze the relation between hygiene and udder health on farm level Dependent variables were average SCC, the average percentage of new cows with a high SCC and the percentage of clinical mastitis, all in the year previous on the visiting date The year average SCC was positively related to the fraction of cows with dirty teats before milking and the fraction of cows with dirty thighs The year average percentage new cows with a high SCC was positively related to the fraction of cows with dirty teats before milking and the fraction of milkings where teats were not sprayed by the AMS The year percentage clinical mastitis was positively related to the frequency of replacing the milking filters At the cow level, hygiene scores of udder, thighs and legs (range to 4, where is clean and is very dirty) were related with the cow SCC from the test day nearest to the visiting date using general linear mixed model There was a positive relationship between cow SCC and the hygiene score of the udder The relationship between cow hygiene and udder health is confirmed, also on AM farms Introduction The first automatic milking systems (AMS) on commercial farms were introduced in The Netherlands in 1992 (de Koning and Rodenburg, 2004) In 2008, it was reported that about 8,000 automatic milking units on approximately 5,500 farms were in use worldwide, and the numbers are still growing (Svennersten-Sjaunja and Pettersson, 2008) Mastitis is a frequent and costly problem in many dairy herds (e.g., Halasa et al., 2007) In a recent study the average clinical mastitis incidence rate in a year was 30.3 (Jansen et al., 2009) Udder health is at risk on AM farms Several studies have been published on the consequences of automatic milking (AM) on bulk milk somatic cell count (BMSCC) (e.g., Rasmussen et al., 2002; Vorst and Hogeveen, 2000) These studies showed an increase in BMSCC by the transition of the herd from conventional milking to AM Koning et al (2004) also found an increase in BMSCC shortly after the transition but a recovery of BMSCC to the same level previous to the introduction was seen in about months In another study BMSCC did not increase after introducing AM but was already higher before the change of system compared to other conventional herds (Klungel et al., 2000) Not only BMSCC increased, somatic cell count (SCC) of individual cows also increased after introduction of AMS (Kruip et al., 2002; Rasmussen et al., 2001) Poelarends et al (2004) investigated possible relationships between cow factors and SCC and found that a significant increase of SCC was seen on second and third parity cows Contrary to BMSCC and SCC, quarter SCC decreased in an experimental study where AM was compared to conventional milking (Berglund et al., 2002) It is hard to draw any conclusions about the factors that cause the results and their differences found Besides that many more aspects than just milking technique changes by the transition of the herd from conventional milking to AM (Poelarends et al., 2004) Previously it has been shown that on farms with conventional milking a lower BMSCC is found when there is paid more attention to hygiene management (Barkema et al., 1998) Schreiner and Ruegg (2003) found that udder hygiene scores and leg hygiene scores were significantly associated with cow SCC on 1,250 lactating dairy cows from eight farms Another observational study on 1,093 lactating dairy cows from eight farms showed significant relationships between cow SCC and hygiene scores of the udder and lower legs and the udder-lower leg composite score (Reneau et al., 2005) According to these results it is expected that hygiene aspects are also of importance in relationship to SCC on AM farms, and maybe even more because of the automatic cleaning of the udder Some research had been done on the influence of poor hygiene on udder health on AM farms, but still knowledge on this subject is very poor An observational study on 28 AM farms in the Netherlands to identify risk factors affecting milk quality showed an increased BMSCC on farms with a poor overall hygiene (Koning et al., 2003) Knappstein et al (2004) determined significant differences in teat cleaning efficiency of different brands of AMS by measuring total bacterial count It also showed that the initial contamination of teats is of more influence on the efficiency of teat cleaning than the cleaning itself Several management factors associated with high teat contamination were found However no relationship was made with SCC and only 18 farms were included The aim of the present study is to identify the relationship between hygiene and udder health on AM farms Materials and Methods Survey Design A survey was developed consisting of a partially open-ended questionnaire and a scoring protocol The questionnaire consisted of five parts, namely general information, AMS, udder health, housing, and cow hygiene The contents of the general information part were number of dairy cows, milk quota, way of business (conventional or organic), and average 305-day milk yield The AMS part consisted of brand, frequency of cleaning of the AMS in general, and cleaning of the different parts of the AMS (camera or laser, milking tubes, teat cups, air inlets, robot arm, feeding crib, floor, and pre-milking teat cleaning system), pre-milking teat cleaning system (disinfection and replacement of the roller brush system or lining of the cleaning cup, the type disinfectant and control of the stock of disinfectant), treatment after milking (use of teat spray, control of the stock of spray, control of the spraying process and frequency of replacing the milking filters), and control of the functioning of the different AMS parts (pre-milking teat cleaning system, spraying process, and cleaning of the AMS) Udder health was determined by the number or percentage of clinical mastitis cases in the past year Housing was questioned by information on the cubicles (ratio cow/cubicles, bedding material, and the storage of bedding material), flooring type, manure removal (by automatic scraper and by hand), and calving area (facilities, cleaning, and disinfection) Cow hygiene was covered by the frequencies of shearing/burning the udders and tails, the percentage of cows which are leaking milk, and the percentage of cows lying on alleys The scoring protocol was used to gain specific information about the cleanliness of the AMS parts, the functioning of the AMS, and cow hygiene by visually inspection Cleanliness of the AMS parts (the camera or laser, milking tubes, teat cups, air inlets, robot arm, feeding crib, floor, and premilking teat cleaning system) were scored for each robot in a range from to 4, where score was clean, score was slightly dirty, score was dirty, and score was very dirty During 10 milkings, the following information about the functioning of the AMS was collected Cleanliness of the teats before milking and after teat cleaning was scored in a range from to in the same way as the cleanliness of the different AMS parts Teat cleaning itself was measured by frequency and the amount of teats cleaned Cleaning of the roller brush system was scored in a range from to 4, where score was good, score was tolerable, score was moderate, and score was no cleaning at all Teat spraying process was scored in a range from to according to the amount of surface of each teat sprayed, where score was > 60 % covered, score was 30-60 % covered, score was 030 % covered, and score was no spraying at all Cleaning of the milking cluster was scored in a range from to 3, where score was good, score was moderate, and score was no cleaning at all Steam cleaning was determined present or absent Cow hygiene was scored for at least 10 lactating cows Hygiene of the udder, thighs and legs were scored in a range from to 4, where score was completely free of or has very little dirt, score was slightly dirty, score was mostly covered in dirt, and score was completely covered, caked-on dirt (Schreiner and Ruegg, 2003) The Dutch national milk recording system (CRV, Arnhem, the Netherlands) provided information from the to 6-weekly milk production registration system, including cow identification, date of milk recording, test-day milk yields, and SCC (cells/mL) for all cows All the information on the milk recordings in the year previous on the visit and the available data after the visit of each farm was selected Percentage clinical mastitis (CM) in the past year was selected from the questionnaire On farm level the milk recording information dataset consisted of 120,700 milk recordings in the year previous before visiting date for 144 farms Evaluations with a SCC of and/or kilograms of milk were deleted Also the database was checked for less than 10 milkings per evaluation, they were deleted as well Finally the dataset consisted of 114,258 milk recordings On cow level the milk recording information dataset consisted of 2,294 milk recordings Data Collection and Preparation The Dutch dairy cooperative FrieslandCampina approached 400 AM farms in The Netherlands with the request to participate in the survey, eventually 151 farms were visited between May 2008 and November 2008 Data was collected, using the survey design as described in the previous paragraph Each farm was visited by a team of two trained students from the Faculty of Veterinary Science in Utrecht From the total of 151 farms four farms were excluded because they were milking with an AMS shorter then one year, one farm was excluded because there was no milk recording information available, another farm was excluded because there were also cows milked conventionally, and one farm was left out because some unlikely values were reported, such as % of clinical mastitis cases in the past year On farm level (dataset 1), some transformations of the data were made The scores on the cleanliness of the AMS parts were averaged by farm and classified into clean (score 1) or not clean (score ≥ 1) The scores on functioning of the AMS and cow hygiene were calculated into percentages of the observations above a specific score by farm The choice of the score used as threshold to transform a categorical score variable into a fraction varied between variables and was based on the significance of the relationship with the dependant variables shown by univariate analysis To investigate the associations between the independent variables the correlation between the continuous variables was studied using Pearson’s correlation Relations between the categorical variables were analyzed using chi-square analysis The only high correlation seen was between the four variables on teat spraying process (R > 0.95) Therefore they were averaged into one variable on teat spraying process for all the teats Furthermore, there were no variables highly correlated Four dependant variables on udder health for analysis on farm level were determined Year average SCC was calculated on each farm as follows: average SCC of the cows was calculated for each test day in the year previous on the visiting day Then year average SCC for that year was calculated High SCC (HSCC) was defined as an individual cell count > 150,000 cells/mL for primiparous cows and > 250,000 cells/ml for multiparous cows, based on the currently used cut off levels for sub-clinical mastitis in The Netherlands Year average percentage cows with HSCC was calculated on each farm as follows: percentage cows with HSCC was determined by the number of cows with a high SCC, divided by the total number of cows for each test day in the year previous on the visiting day Then year average percentage cows with HSCC for that year was calculated The year average percentage new cows with HSCC (NHSCC) was calculated on each farm as follows: percentage new cows with HSCC was determined by the number of cows with HSCC in the current milk recording and low SCC in the previous milk recording, divided by the total number of cows for each test day in the year previous on the visiting day Then year average percentage new cows with HSCC was calculated for the year previous on the visiting date (NHSCC) Year percentage CM is determined as the number of clinical mastitis cases per year, divided by the average number of dairy cows present on the farm that year The correlation between these four dependant variables was studied using Pearson’s correlation Year average SCC and year average percentage cows with HSCC were highly correlated (R = 0.89), so the year average percentage cows with HSCC was excluded for further analysis The natural logarithm of the udder health parameters year average SCC, year average percentage NHSCC, and year percentage CM was used for analysis Finally, dataset consisted of 62 independent variables and dependent variables measured on 144 farms The independent variables and their unit or categories or used threshold values and the dependent variables are listed in Appendix A On cow level (dataset 2), there was available milk recording information within a range from weeks before till weeks after the visiting date for all the cows included Double or unlikely scores were excluded Besides the hygiene scores the amount of days between the visiting date and the milk recording test day was also determined Relations between the categorical variables were analyzed using chi-square analysis No variables needed to be transformed or excluded for this reason One dependant variable on udder health for analysis on cow level was used Cow SCC (CSCC) was determined for the cows involved in hygiene scoring by taking the individual CSCC from the nearest milk recording date, within the previously described range of weeks before and weeks after the visiting date The natural logarithm of CSCC was used for analysis Finally, dataset consisted of independent variables and dependent variable measured on 2,294 cows from 108 farms The independent variables and their unit or categories or used cut off values are listed in Appendix B Statistical analyses On farm level, all the variables measured by the questionnaire and the scoring protocol were analyzed univariately on their relationship with the udder health parameters year average SCC, year average percentage NHSCC and year percentage CM using general linear models Also on farm level a selection of independent hygiene specific variables from dataset was made, based on their expected importance (biological relevance) and the amount of information known on that variable (maximum of five missing values) This selection was used to create three general linear models with the above described udder health parameters Correlation coefficients of the models were calculated The udder health parameters were determined by the following equation: Ln (UHP) = α + β1X1 + β2X2 + … + βiXi + group1 + group2 + + groupj, + ε Where UHP = Udder health parameter (year average SCC or year average percentage NHSCC or year percentage CM), α = intercept, βi = regressive coefficient of continuous hygiene specific variable Xi, groupj = effect of categorical hygiene specific variable, and ε = residual random error On cow level, an analysis was made using a general linear mixed model CSCC was used as dependent variable and hygiene score of the udder, thighs and legs as independent variables The number of days between the visiting day and the test day was taken into analysis Also the interaction of the number of days with the hygiene scores was tested The hygiene scores, the amount of days, and their products were fixed effects with herd as a random effect To determine CSCC by hygiene scores the following equation was used: Ln (CSCC) = α + HU + HT + HL + Days + (HU*DAYS) + (HT*DAYS) + (HL*DAYS) + μherd + ε Where CSCC = Cow somatic cell count, α = intercept, HU = effect of hygiene score of the udder, HT = effect of hygiene score of the thighs, HL = hygiene score of the legs, DAYS = effect of number of days between visiting day and test day, (HU*DAYS)/ (HT*DAYS)/ (HL*DAYS) = interaction terms, μherd = random herd effect, ε = residual random error For both the farm level analysis and the cow level analysis, model selection was performed by a backward stepwise procedure The final models retained only the variables that were significant at p ≤ 0.05 using an F-test Differences between groups in a significantly related categorical variable were tested for significance by pair wise comparison in the Student’s t-test Significance levels were multiplied by the number of comparisons made according to the Bonferroni correction (Abdi, 2007) The normality was assessed by residual analysis The residuals were plotted in a normal probability plot and checked for peculiarities (Dohoo et al., 2003) All data preparation and statistical analyses were carried out using SAS version 9.1 (SAS Institute Inc., Cary, NC) Results Descriptive statistics Descriptive statistics of general farm data and the udder health parameters of the farms and cows used in analysis are listed in Table For the 144 farms used in analysis on farm level the number of dairy cows on a farm ranged from 30 up to 420 with 85 as an average The annual milk production quota held between 154,000 kg and 5,000,000 kg with an average of 796,000 kg Average 305-day milk yield varied from 5,500 kg to 11,000 kg with 9,008 kg as an average There were to robots on a farm with an average of 1.6 This means an average of 54 cows per robot within a range from 30 to 85 Year average SCC was 267,000 cells/mL within a range of 90,000 cells/mL and 509,000 cells/mL The year average percentage NHSCC ranged from % up to 18 % with an average of 10 % Year percentage CM varied from % to 135 % with 26 % as an average The 108 farms used for the cow level analysis were almost equal to the 144 farms used for farm analysis Average CSCC for the involved 2,294 cows in cow analysis was 291,000 cells/mL within a range from 4,000 cells/mL up to 9,999,0001 cells/mL Table 1: General farm data and udder health parameters Number of observations (N), mean value (Mean), minimum value (Min), and maximum value (Max) are given for each variable Variable Dairy cows (n) Milk production quota (kg) 305-day milk yield (kg) Robot (n) Cow/robot Year SCC (cells/mL) Year NHSCC (%) Year CM (%) CSCC (cells/mL) Farm level N Mean 144 85 142 796,00 144 9,008 144 1.6 144 54 144 267,00 144 10 144 26 NR1 NR Min 30 154,00 5,500 30 90,000 Max 420 5,000,00 11,000 85 509,000 NR 18 135 NR Cow level N Mean 108 84 106 798,00 108 9,041 108 1.6 108 55 108 269,00 108 10 106 28 2,29 291,00 Min 38 165,00 6,200 33 90,000 Max 420 5,000,000 4,000 14 135 9,999,000 11,000 80 453,000 NR = Not relevant of the 144 farms used in analysis on farm level were organic farms; the other 140 were conventional farms 86 of the farms milked with al Lely, 42 milked with a DeLaval, the other 16 farms milked with SAC (8), RMS (5) or a Fullwood (3) of the 108 farms used in analysis on cow level were organic farms; the other 106 were conventional farms 67 of the farms milked with a Lely, 39 milked with a DeLaval, the remaining farms milked with a Fullwood Detection system does dot measure levels above 9,999,000 cells/mL Average hygiene scores of the different body parts and the distribution of the different scores are listed in Table The average hygiene scores of the udder, thighs and legs of the cows involved in analysis on cow level were 2.76, 2.54 and 2.45 respectively Hygiene score and were mostly observed for as well the udder, thighs and legs Table 2: Frequencies and averages of hygiene score of the udder, thighs and legs of dairy cows Mean Frequency (%) = clean = slightly dirty = dirty = very dirty Udder 2.76 Thighs 2.54 Legs 2.45 97 (4.23) 762 (33.22) 1034 (45.07) 401 (17.48) 128 (5.58) 1052 (45.86) 861 (37.53) 253 (11.03) 96 (4.18) 1182 (51.53) 912 (39.76) 104 (4.53) Univariate analysis Results of the univariate analysis on farm level are given in Table Only variables that were significantly related (p ≤ 0.05) to one or more dependent variable(s) are listed Year average SCC was positively related to the cleaning frequency of the feeding trough, the fraction of cows with dirty teats before milking, the fraction of cows with a dirty udder, the fraction of cows with dirty thighs, and the fraction of cows with dirty legs Year average SCC was negatively related to the cleaning frequency of the laser or camera Positive relationships were seen between year average percentage NHSCC and the fraction of cows with dirty teats before milking, the fraction of milkings where teats are not sprayed, the fraction of cows with a dirty udder, and the fraction of cows with dirty thighs Year average percentage NHSCC was also related to the cleanliness of the milking tubes, it showed that year average percentage NHSCC was higher when the milking tubes were determined not clean Year percentage CM was positively related to the frequency of replacing the milking filters and the fraction of cows with dirty udders The type of bedding material was also related with year percentage CM; it was significantly higher when the bedding material was straw instead of sawdust Year percentage CM turned out to be significantly higher when the cleanliness of the feeding crib was determined not clean Table 3: Results of the univariate analysis on farm level Variable Cleaning frequency of the laser or camera Cleaning frequency of the feeding trough Replacing the milking filters ≤1/day >1≤ 2/day 3/day Year average SCC -1* Year average percentage NHSCC NS +2* NS3 NS NS Year percentage CM NS NS -** -* Ref.4 Bedding material: NS straw other none sawdust Cleanliness of the milking tubes: NS clean not clean Cleanliness of the feeding crib: NS clean not clean Fraction of cows with dirty teats before +** milking Fraction of milkings teats not sprayed NS Fraction of cows with a dirty udder +** Fraction of cows with dirty thighs +** Fraction of cows with dirty legs +** + = Positive significant relationship with dependent variable - = Negative significant relationship with dependent variable NS = Not significant Ref = Reference category NS +* NS NS Ref NS Ref +* NS +* Ref +* NS +** +** +* NS NS +* NS NS * = p