Mastitis is a common multifactorial disease and is the most costly production issue in dairy cows. The major factor influencing BMSCC or composite weighted mean cow SCC (CMSCC) (Valde et al. Reference Valde, Osteras and Simensen2005) is the prevalence of intramammary infections within the herd. A lot of epidemiological studies reported herd-level factors influencing BMSCC: first, milking procedures and general cleanliness (Barnouin et al. Reference Barnouin, Fayet, Jay, Brochart and Faye1986; Chassagne et al. Reference Chassagne, Barnouin and Le Guenic2005; Rodrigues & Ruegg, Reference Rodrigues and Ruegg2005); second, effects related to the lactation, such as milk production, lactation stage and parity (Berry et al. Reference Berry, O'Brien, O'Callaghan, Sullivan and Meaney2006; Wenz et al. Reference Wenz, Jensen, Lombard, Wagner and Dinsmore2007; Green et al. Reference Green, Bradley, Medley and Browne2008; Madouasse et al. Reference Madouasse, Huxley, Browne, Bradley and Green2010b); third, some structural farm characteristics (herd size) and replacement policy (De Vliegher et al. Reference De Vliegher, Laevens, Barkema, Dohoo, Stryhn, Opsomer and de Kruif2004; Rodrigues et al. Reference Rodrigues, Caraviello and Ruegg2005; Wenz et al. Reference Wenz, Jensen, Lombard, Wagner and Dinsmore2007; Elmoslemany et al. Reference Elmoslemany, Keefe, Dohoo, Wichtel, Stryhn and Dingwell2010). The influence of these factors on BMSCC are probably indirect, because they impact the prevalence of (sub)clinical mastitis.
In addition to these well known effects, BMSCC or CMSCC also seem to depend on other factors. First, few studies evaluated the direct effect of nutrition or metabolic diseases on BMSCC (Nyman et al. Reference Nyman, Emanuelson, Holtenius, Ingvartsen, Larsen and Waller2008, Reference Nyman, Emanuelson, Gustafsson and Persson Waller2009). Because the effects of metabolic diseases on immunity are extensively reported (Scalia et al. Reference Scalia, Lacetera, Bernabucci, Demeyere, Duchateau and Burvenich2006), an association between udder health and negative energy balance or subacute ruminal acidosis can be suspected. Second, farmers’ characteristics and competencies were reported to highly influence clinical mastitis (Jansen et al. Reference Jansen, van den Borne, Renes, van Schaik, Lam and Leeuwis2009) and BMSCC (Barkema et al. Reference Barkema, Van der Ploeg, Schukken, Lam, Benedictus and Brand1999; Rougoor et al. Reference Rougoor, Hanekamp, Dijkhuizen, Nielen and Wilmink1999; Khaitsa et al. Reference Khaitsa, Wittum, Smith, Henderson and Hoblet2000; Valeeva et al. Reference Valeeva, Lam and Hogeveen2007). The farmer's perception of BMSCC as a problem has an effect on SCC management, and BMSCC is associated with precise or fast workers (Valeeva et al. Reference Valeeva, Lam and Hogeveen2007). In addition, various combinations of farm activities (dairy, beef or fattening cattle, goats, sheep, poultry and pigs) within one farm usually occur in France (Renting et al. Reference Renting, Rossing, Groot, Ploeg, Laurent, Perraud, Stobbelaar and Ittersum2009); the behaviour of the farmer towards the dairy herd could differ between specialized and non-specialized farmers. Third, some studies showed that BMSCC and CMSCC depend on the geographical areas or on the territorial characteristics of the dairy production (Allore et al. Reference Allore, Oltenacu and Erb1997; Norman et al. Reference Norman, Miller, Wright and Wiggans2000), but such results are rare. This effect is potentially important in France, because of its large livestock diversity. Dairy cattle are mainly in the north, beef, sheep and goat in the centre and the south and Protected Designation of Origin milk productions in the mountains (a fifth of the agricultural land) (Rouquette & Pflimlin, Reference Rouquette and Pflimlin1995; Sarzeaud et al. Reference Sarzeaud, Bécherel, Perrot, Sarzeaud, Dimitriadou and Zjalic2008).
The hypothesis of this study was that French CMSCC depends on several herd-level factors, including metabolic diseases and farmers’ competencies, as well as some territorial risk factors, regarding the livestock farming system. This study aimed to describe CMSCC in France and, secondly, it quantified the weight of selected herd-level and territorial-level risk factors on CMSCC.
Materials and Methods
Datasets
Several datasets reporting animal or farm data were used to calculate dairy herd-level and farm environmental indicators. All data were geo-located at the municipal level. There are 3600 municipalities in France, with a mean area of 15 km2.
The detailed characteristics of the National Bovine Identification Database (NBID) was previously described (Raboisson et al. Reference Raboisson, Cahuzac, Sans and Allaire2011). Briefly, NBID contains routine records on individual data of farms and animals. The number of dairy cows identified for at least one day was 5·3 and 5·1 million and the number of cow-years was 3·8 and 3·7 million for 2005 and 2006, respectively. Animals were sorted and associated to a dairy, beef (suckler cows) and fattening (bulls, steers or veal calves) herd within each farm. The records from French herds in the Milk Control Programme (MCP) during 2005 and 2006 included lactational and test-day records for the 4294 million cows with at least one control in 2005 or 2006. The other datasets used were previously described (Raboisson et al. Reference Raboisson, Cahuzac, Sans and Allaire2011). Briefly, they included the farms registered as following the charter of Good Breeding Practices member, the municipal agricultural land as ‘always with grass’ (named grassland) and the municipal overall agricultural land (named overall land). Protected Designation of Origin areas were provided from the National Institute for Quality (www.inao.gouv.fr). Data related to weather conditions were provided by Météo-France (www.meteofrance.fr).
Variables
The variables were calculated at the dairy herd level, for each month or year (2005, 2006) from the databases. Calculations made from NBID (breed, typology, primiparous proportion, culling rate, age at first calving, calving interval) were previously described in detail (Raboisson et al. Reference Raboisson, Cahuzac, Sans and Allaire2011). The variables calculated from the MCP data included all the lactating animals present in the farm on the days of the controls.
CMSCC is the herd arithmetical mean of all individual SCC weighted by each cow's 24-h milk production for one test-day, as previously suggested (Valde et al. Reference Valde, Osteras and Simensen2005).
Three ‘structural’ variables were included: herd size (average number of cows for all test-days of the year), predominant breed and farm typology (Dairy; Dairy and Beef; Dairy and Fattening). The lactation characteristics were defined using the test-day mean milk production of the herd and the test-day average days in milk. The average length of the dry-off period was calculated yearly. Several variables accounted for the farm replacement policy: monthly proportion of primiparous cows, monthly average age at first calving, yearly culling rate (irrespectively of the reasons for the removal) and yearly purchase of cows from other farms (No purchase, Low purchase or High purchase). The thresholds used to distinguish Low and High classes was the 75% quartile of the purchase percentage (27 and 19%, for 2005 and 2006, respectively).
For each test-day, cows were considered to have a negative energy balance (NEB) when their milk protein-to-fat ratio was ⩽0·66 (Duffield et al. Reference Duffield, Kelton, Leslie, Lissemore and Lumsden1997). They were considered to have subacute ruminal acidosis (SARA) when their fat percentage minus protein percentage was ⩽0·30% (including fat percentage <protein percentage). When the predominant breed was not Holstein, the thresholds were corrected proportionally to the mean milk and fat protein percentages of each breed (rule of three; Holstein with 3·2% and 4·1% of protein and fat). Three classes were built for monthly SARA and NEB: Low risk (percentage=0% at one test-day), Moderate risk and High risk. The threshold used to distinguish Moderate and High classes was 10 and 25%, for NEB and SARA, respectively, as suggested by the distribution of the values.
Farmers’ competencies were included through variables related to the general farming management, as suggested by a previous study (Valeeva et al. Reference Valeeva, Lam and Hogeveen2007). The yearly average calving interval was used as an indicator of the reproduction management efficiency on the farm. The presence of an ‘autumn calving peak’ was defined when at least 35% of the annual calvings occurred in a 3-month period that was between July and November. This variable aims to detect the sensitivity of farmers to produce milk during the best paid period of the year. Because one-third of farms had no mortality each year, the annual mortality was a categorical variable defined by ‘No dairy cow death’ or ‘Having at least one dairy cow death’. Being a Good Breeding Practices member was defined once for the 2-year period.
Two territorial factors used in the present study were previously described [(Raboisson et al. Reference Raboisson, Cahuzac, Sans and Allaire2011): cattle density (expressed in livestock unit/km2 (Sarzeaud et al. Reference Sarzeaud, Bécherel, Perrot, Sarzeaud, Dimitriadou and Zjalic2008)] and the grass on land (grassland on overall land) ratio. Moreover, the municipal monthly mean temperature, wind, relative humidity, quantity of rain and number of days with rain and frost per month were included. The 11 dairy production areas (Fig. 1) used to characterize the French territories (Raboisson et al., Reference Raboisson, Cahuzac, Sans and Allaire2011) overlap approximately the French breeding systems (Rouquette & Pflimlin, Reference Rouquette and Pflimlin1995; Sarzeaud et al. Reference Sarzeaud, Bécherel, Perrot, Sarzeaud, Dimitriadou and Zjalic2008).
Fig. 1. Definition of the dairy production areas (DPA). Numbers refer to dairy production areas (see key in Table 5).
Statistical analysis
Data were analysed using R (version 2.10.1). A linear mixed effect model was used to explain monthly CMSCC, with the farm as random variable (package nlme). CMSCC was ln-transformed, because it suited a non-normal distribution. Estimates reported are expressed in % change of CMSCC (UCLA, 2012).
All variables including month were included in the model, except the following ones. The variables ‘parity’ and ‘cow age’ were excluded, because of their high correlation with primiparous proportion and age at first calving (r>0·7). Because the grass on land ratio was strongly correlated with the dairy production areas and altitude (r=0·65), it was not included. Moreover, the effects of month or weather were included alternatively, because of the correlation among these variables. The altitude was neither correlated with temperature (r=−0·18), nor with the number of rainy days (r=−0·30), and all 3 remained in the models. Only the mean temperature and the number of days with rain were included in the model, because of the high correlations with the other weather variables. Because of the large variations of altitude within the areas, both remained in the model. All the possible two-factor interactions were included (one by one) in the model with all the main effects. Depending on the coefficient of the interaction and on the AIC of the model, the interaction can be removed from the model, even if significant: it was interpreted as a significant interaction without any biological importance.
Results
Descriptive analysis of CMSCC
Monthly CMSCC and percentage of herds above 250000, 300000 and 400000 cells/ml were very close between the 2 years (Table 1). The correlations of CMSCC between 2 successive months were 0·55–0·57 for months from January to July and 0·49–0·52 for months from July to December. The within-farm correlations of CMSCC between the same months of 2005 and 2006 were 0·33–0·35, but raw data show the same monthly distribution among areas (Fig. 2). Percentages of cows and herds in MCP and average CMSCC varied widely among dairy production areas (Table 2). No significant difference (P>0·05, t-test and χ2 tests) was observed for continuous and categorical variables between 2005 and 2006 (Tables 3 & 4). The correlations between proportion of cows with SARA or NEB for the whole lactation and the first 4 months were 0·84 for SARA and 0·86 for NEB.
Fig. 2. Monthly composite milk weighted mean cow somatic cell count (CMSCC) per dairy production area (DPA) and per month. The values of each month, indicated by the month number, is the 2005 and 2006 mean CMSCC value.
Table 1. Characteristics of French monthly composite milk weighted mean cow somatic cell count (Mo-CMSCC)† in 2005 and 2006
† CMSCC was the herd arithmetic means of all individual SCC weighted by each cow's 24-h milk production for each test-day
Table 2. Characteristics of dairy production areas
† Percentages are expressed relative to the overall number of units and to the number of cow-years within dairy production areas (MCP, milk control programme; DPA, dairy production area)
‡ CMSCC (composite milk weighted mean cow somatic cell count) were the herd arithmetic means of all individual SCC weighted by each cow's 24-h milk production, for all test-days of the year
§ PDO: protected designation of origin
¶ Livestock units
Table 3. Descriptive statistics of continuous variables in 2005 (because 2005 and 2006 results are very close, only 2005 results are reported)
Table 4. Descriptive statistics of categorical variables in 2005 (because 2005 and 2006 results are very close, only 2005 results are reported
† Results refers to the number of farm-controls
Regression analysis
The results of the model are expressed in CMSCC change (Table 5). For instance, a 10 cow increase is associated with an 3·3% CMSCC increase: if the initial value of CMSCC is 300000 cells/ml, the expected CMSCC after an 10 cow increase is 309900 cells/ml (300000×1·033) (UCLA, 2012).
Table 5. Variables associated with a monthly composite milk weighted mean cow somatic cell count (CMSCC) change. Results are expressed in CMSCC change (%). For instance, a 10-cow herd-size increase is associated with an 3·3% CMSCC increase: if the initial value of CMSCC is 300000 cells/ml, the expected CMSCC after a 10-cow increase is 309900 cells/ml (300000×1·033)
† For an increase in herd size of 10 cows
‡ For an increase of 1000 l
§ for an increase of 10 d
¶ for an increase of 10%
*P<0·05; **P<0·01; ***P<0·001; NA, not available
Among the effects reported, a 10% increase of the cow proportion at risk for NEB and SARA is associated with a 1–2% CMSCC increase (SARA) and a 3–6% CMSCC increase (NEB) (Table 5). CMSCC was also positively associated with the average calving interval and having at least one dairy cow death this year, although association was negative for having a calving peak in autumn and being a Good Breeding member. For the territorial variables, the cattle density and the altitude were negatively associated with CMSCC, but association was positive for temperature. All area but area 11 were positively associated with CMSCC, compared with area 1 (Table 5).
The AIC value of the model was lower when the weather variables replaced the variable month, even if difference was low (AIC=1536000 and 1548000, respectively). Removing the variables altitude and weather from the model induced important changes for the coefficient of areas 4, 7 and 10 (CMSCC change were −2, −2, −1·5%, respectively), but not for the coefficients of other variables.
All significant interactions were considered as non-biologically relevant, according to the coefficients and the AIC.
Discussion
Datasets and CMSCC
French MCP represented 61, 57 and 85% of the herds, cows and milk produced, respectively. MCP is reported to represent 40–90% of the dairy cows and herds, depending on countries (Lukas et al. Reference Lukas, Hawkins, Kinsel and Reneau2005). In spite of differences among dairy production areas (Table 2), present results seem to represent well the French situation and probably also the udder health in other countries.
The use of CMSCC in this study allowed investigation of the udder health situation of the herds and taking into account the issues relative to the BMSCC. A high correlation (0·78) among prevalence of high SCC and CMSCC is reported (Lievaart et al. Reference Lievaart, Kremer and Barkema2007). In small herds (mean herd size=15 cows), CMSCC is highly correlated with both BMSCC (r=0·77) and with the percentage of individual cow milk samples >200000 cells/ml (r=0·72) (Valde et al. Reference Valde, Osteras and Simensen2005). Nevertheless, differences between CMSCC and BMSCC are reported to vary from 10000 to 182000 cells/ml in larger herds, depending on at least the milk withheld from the bulk tank (Lievaart et al. Reference Lievaart, Barkema, Hogeveen and Kremer2009).
Monthly CMSCC mean and median were close to the lowest penalties threshold of BMSCC (250000 cells/ml) and important standard deviations were reported. The limits of comparison with other studies come from the variations among the calculation methods (in particular between CMSCC and BMSCC), among the thresholds for penalties and premiums and among the data collection dates. Reported mean herd SCC varied from 140000 to 340000 cells/ml in studies carried out in the 1990s in USA, Ontario, New Zealand or Norway (Schukken et al. Reference Schukken, Leslie, Weersink and Martin1992; Sargeant et al. Reference Sargeant, Schukken and Leslie1998; Norman et al. Reference Norman, Miller, Wright and Wiggans2000; McDougall, Reference McDougall2003; Valde et al. Reference Valde, Osteras and Simensen2005). Taken altogether, this suggests an average CMSCC of the French dairy cows that is within the range reported for other countries.
The mean CMSCC values of a country must be interpreted in relation with the local thresholds for penalties and exclusion, which vary widely among countries. It is 750000 cells/ml for USA, 500000 cells/ml for Canada and 400000 cells/ml for most European countries (2 consecutive trimestrial geometric mean BMSCC >400000 cells/ml; Directive 92/46/EEC), Australia and New Zealand (Norman et al. Reference Norman, Miller, Wright and Wiggans2000). Moreover, SCC tended to decrease across time: for instance, BMSCC was 345000 and 250000 cells/ml in Ontario in 1986–1987 and 1994–1995, respectively (Sargeant et al. Reference Sargeant, Schukken and Leslie1998).
Effects of farm structures, lactation characteristics and replacement policy on CMSCC
The positive association of herd size and CMSCC was in accordance with previous results (Skrzypek et al. Reference Skrzypek, Wojtowski and Fahr2004; Valde et al. Reference Valde, Osteras and Simensen2005), but in disagreement with others (Fenlon et al. Reference Fenlon, Logue, Gunn and Wilson1995; Norman et al. Reference Norman, Miller, Wright and Wiggans2000). To our knowledge, the effects of the predominant breed and of specialization into dairy production on CMSCC or BMSCC were not previously reported. Among the hypotheses proposed, the specialized farmers could give more attention for the dairy cows and had better management acumen for the dairy production and better biosecurity measures. From the authors’ personal observations, in France, dairy and beef or fattening herds of the same farm are often not strictly separated.
The negative association of CMSCC and the milk production was in accordance with previous results and probably results from a dilution effect (Emanuelson & Funke, Reference Emanuelson and Funke1991; Fenlon et al. Reference Fenlon, Logue, Gunn and Wilson1995; Wenz et al. Reference Wenz, Jensen, Lombard, Wagner and Dinsmore2007).
The 5·6% decrease of CMSCC for each 10% increase of primiparous cow percentage is in agreement with the decreased odds ratio of having high SCC in early lactation for primiparous cows compared with cows in parity >5 (Green et al. Reference Green, Bradley, Medley and Browne2008). In the present study, the culling rate and CMSCC were positively associated, although no significant association between culling-whatever the reason for the culling-and BMSCC was found in previous studies (Fenlon et al. Reference Fenlon, Logue, Gunn and Wilson1995; Wenz et al. Reference Wenz, Jensen, Lombard, Wagner and Dinsmore2007). The present study cannot distinguish culling as a consequence of high CMSCC or to decrease CMSCC in the future.
Purchasing cows induced a +7% change of CMSCC, compared with the herds without purchasing. The estimators were of the same order of magnitude between Low-Purchase and High-Purchase, suggesting an effect related to ‘at least one cow bought’/‘no cow bought’. The effect of ‘any cattle brought onto unit’ was not associated with BMSCC in a study based on 1031 US dairies (Wenz et al. Reference Wenz, Jensen, Lombard, Wagner and Dinsmore2007). Newly introduced cows must face modifications of their environment and of the milking practices, increasing their risk for mastitis. Purchasing cows also increases the risk of introducing infection from another herd. It can concern both pathogens directly related to udder health or other pathogens. The involvement of other pathogens was suggested by the positive association between purchasing replacement animals (before calving) and BMSCC (Fenlon et al. Reference Fenlon, Logue, Gunn and Wilson1995). Purchasing could also be the consequence of a high culling because of high CMSCC, but a low yearly association between culling and purchasing was described (r<0·1).
Effects of metabolic diseases on CMSCC
Studies dealing with the relationship between energy balance and milk composition suggest an association of NEB and fat percentage increase, protein percentage decrease, and protein-to-fat ratio decrease (Duffield et al. Reference Duffield, Kelton, Leslie, Lissemore and Lumsden1997; de Vries & Veerkamp, Reference de Vries and Veerkamp2000; Heuer et al. Reference Heuer, Van Straalen, Schukken, Dirkzwager and Noordhuizen2001). Differences among the conclusions of these studies did not allow the defining of a consensual best tool among these three indicators of NEB.
Lactation stage and season were reported to influence protein-to-fat ratio, in particular when large datasets were used (Madouasse et al. Reference Madouasse, Huxley, Browne, Bradley, Dryden and Green2010a). The potential effects of breed, lactation stage, season (through month or weather) and diets (through month, weather and dairy production areas) were taken into account in the models. Moreover, a high correlation is reported among the indicators calculated for the whole lactation or the first 4 months. Taken altogether, this suggests using protein-to-fat ratio as NEB indicator in the present large database study, even if its accuracy remains difficult to define.
The sensitivity and specificity to detect subclinical ketosis were 58 and 69% with a test-day protein-to-fat ratio threshold of 0·75 (Duffield et al. Reference Duffield, Kelton, Leslie, Lissemore and Lumsden1997). Because a high specificity was looked for to analyse the relationship between NEB and CMSCC, the threshold used was 0·66 (sensitivity and specificity reported were 20 and 85%, respectively).
SARA is known to induce a milk fat depression, with no or low milk protein depression (Kleen et al. Reference Kleen, Hooijer, Rehage and Noordhuizen2003; Oetzel, Reference Oetzel2004; Enjalbert et al. Reference Enjalbert, Videau, Nicot and Troegeler-Meynadier2008), suggesting its association with the test-day protein and fat percentages difference. Nevertheless, the sensitivity and specificity of milk composition tools to test SARA were not known (Raboisson & Schelcher, Reference Raboisson and Schelcher2008).
This study clearly shows the positive association between the SARA and NEB indicators and CMSCC, with an increased change (twice higher) from the Moderate risk class to the High risk class. SCC at first test-day of the lactation was positively associated with the non-esterified fatty acid (NEFA) concentration before calving and with the difference of the NEFA concentration before and after calving, whereas the association with the beta-hydroxybutyric acid (BHBA) concentration before calving was negative (Nyman et al. Reference Nyman, Emanuelson, Holtenius, Ingvartsen, Larsen and Waller2008). This is in agreement with the present results when considering the direct relationship of NEFA on immune cells (Lacetera et al. Reference Lacetera, Scalia, Franci, Bernabucci, Ronchi and Nardone2004; Scalia et al. Reference Scalia, Lacetera, Bernabucci, Demeyere, Duchateau and Burvenich2006) and the potential origin of BHBA from the rumen (Oetzel, Reference Oetzel2004). To our knowledge, the relationship between SARA and CMSCC has not been reported previously.
Effects of dairy farmers’ competencies on CMSCC
Farmers’ motivations to improve mastitis management referred to monetary factors (premium-penalties oriented motivations and basic economic motivations) and to non-monetary factors as efficient (well-organized) farming (Valeeva et al. Reference Valeeva, Lam and Hogeveen2007). In the present study, the calving interval, the autumnal calving peak and having at least one dairy cow dead this year were included in the models to indicate the farmers’ management skills. A high level of management skill was needed to achieve a low herd-calving interval, to gather calvings to a few months and to have no cow death. A beneficial effect of this level of general management on CMSCC was likely to occur. Previous studies suggested taking into account variables related to the farmers’ attitudes, and not to the farmers’ behaviour, to evaluate farmers’ characteristics (Barkema et al. Reference Barkema, Schukken, Lam, Beiboer, Benedictus and Brand1998; Rougoor et al. Reference Rougoor, Hanekamp, Dijkhuizen, Nielen and Wilmink1999; Khaitsa et al. Reference Khaitsa, Wittum, Smith, Henderson and Hoblet2000; Jansen et al. Reference Jansen, van den Borne, Renes, van Schaik, Lam and Leeuwis2009). For instance, low BMSCC was reported to be associated with ‘clean and accurate’ farmers, compared with ‘quick and dirty’ (Barkema et al. Reference Barkema, Van der Ploeg, Schukken, Lam, Benedictus and Brand1999).
The effect of being a Good Breeding Practices member appears to be high when considered in the light of the limited conditions relative to membership (Dockes et al. Reference Dockes, Frappat and Godefroy2006). Maybe a selection bias occurred, with the most informed and competent farmers that would tend to become Good Breeding Practices member.
Effects of farm environmental factors on CMSCC
Municipal cattle density was an indicator of the local farming system intensification level. Increased cattle density was associated with a decreased CMSCC and was previously associated with decreased mortality (Raboisson et al. Reference Raboisson, Cahuzac, Sans and Allaire2011).
The environment of the farm and its localization among dairy production areas had important and various effects on CMSCC.
The positive association of the mean temperature and CMSCC is in agreement with previous studies reporting higher BMSCC in southern compared with northern regions of USA (Allore et al. Reference Allore, Oltenacu and Erb1997; Norman et al. Reference Norman, Miller, Wright and Wiggans2000; Oleggini et al. Reference Oleggini, Ely and Smith2001; Wenz et al. Reference Wenz, Jensen, Lombard, Wagner and Dinsmore2007). The negative association of altitude and CMSCC is not related to the lower temperature in the mountains, because the temperature was included within the model and because the correlation among altitude and temperature was low. The effect of the number of days with rain remains unexplained.
If altitude and weather variables were excluded from the model, areas 2, 4, 7, 10, 11 were negatively associated with CMSCC. These 5 areas include at least half of the herds in a Protected Designation of Origin (PDO) area (Table 2), whereas the other areas had less than half of the herds in a PDO area. Some, but not all of these 5 areas are in the mountains (Table 2). PDO goods are produced, processed and prepared in a given geographical area. They are related to specific farm structural factors and management practices and they refer to specifications and a recognized know-how. PDO are also collective tools to promote the products in association with high milk price; the specific payment scheme concerns SCC and bacteria count thresholds among other specifications. This context probably impacts SCC management. A direct evaluation of being in a PDO area would need specific analysis and was not done here.
With the altitude and weather variables in the model, the effect of each DPA was still significant. The coefficients of the areas represented specific and homogeneous risk factors relative to the farming system which were not included in the other effects. The effect of area or farming system summarized a global effect that is over the sum of the specific factors of this area (Ploeg & Renting, Reference Ploeg, Renting, Haverkort, van't Hooft and Hiemstra2002). For instance, the very hot weather of area 9 in summer could explain the highest association of this area and CMSCC, compared with other areas.
Conclusions
This study shows the average CMSCC for the French dairy cows and the impact of several herd-level risk factors on CMSCC. An association between metabolic disorders and udder health was established. Farmers’ competencies, general farm management and farmers’ specialization into dairy production also appeared as key factors concerning the udder health. Including territorial considerations in SCC studies also seems of high importance.
