Mastitis, an intramammary inflammation, is a common and costly disease in many dairy herds around the world (e.g. Sears & Wilson Reference Sears and Wilson2003; Halasa et al. Reference Halasa, Huijps, Østerås and Hogeveen2007). Mastitic dairy cows experience milk loss (Houben et al. Reference Houben, Dijkhuizen, van Arendonk and Huirne1993; Gröhn et al. Reference Gröhn, Wilson, González, Hertl, Schulte, Bennett and Schukken2004) and poor quality of milk (Kitchen, Reference Kitchen1981; Coulon et al. Reference Coulon, Gasqui, Barnouin, Ollier, Pradel and Promiès2002). Milk loss is a large component of the cost of mastitis. Poor quality of milk, as exemplified by high somatic cell count (SCC) and compositional change in milk, can be detrimental to profitability because it may decrease milk price. It is critical for the dairy industry, especially for cheese production and quality (Leitner et al. Reference Leitner, Krifucks, Merin, Lavi and Silanikove2006; Blum et al. Reference Blum, Heller and Leitner2014).
Somatic cell count in milk is a parameter of great economic importance in dairy herds. A high SCC is usually indicative of a response to an intramammary infection (Harmon, Reference Harmon1994; Schukken et al. Reference Schukken, Günther, Fitzpatrick, Fontaine, Goetze, Holst, Leigh, Petzl, Schuberth, Sipka, Smith, Quesnell, Watts, Yancey, Zerbe, Gurjar, Zadoks and Seyfert2011). Many studies have reported high SCC for mastitic cows (e.g. Coulon et al. Reference Coulon, Gasqui, Barnouin, Ollier, Pradel and Promiès2002; Leitner et al. Reference Leitner, Krifucks, Merin, Lavi and Silanikove2006; Jamrozik & Schaeffer, Reference Jamrozik and Schaeffer2012). Other parameters for milk quality include lactose, fat and protein contents in milk. These may affect milk pricing and are important in the dairy industry. Mastitis seems to reduce lactose content (Coulon et al. Reference Coulon, Gasqui, Barnouin, Ollier, Pradel and Promiès2002; Bezman et al. Reference Bezman, Lemberskiy-Kuzin, Katz, Merin and Leitner2015) but has little or no effect on protein and fat content (Botaro et al. Reference Botaro, Cortinhas, Dibbern, Silva, Benites and dos Santos2015; Kester et al. Reference Kester, Sorter and Hogan2015; Tomazi et al. Reference Tomazi, Gonçalves, Barreiro, Arcari and dos Santos2015). However, it is not prudent to consider these changes in milk without pathogen information and precise cow information including parity and days in milk (DIM), since the change in milk due to mastitis might be pathogen-specific (e.g. Coulon et al. Reference Coulon, Gasqui, Barnouin, Ollier, Pradel and Promiès2002; Bezman et al. Reference Bezman, Lemberskiy-Kuzin, Katz, Merin and Leitner2015), and parity- and DIM-dependent (Stanton et al. Reference Stanton, Jones, Everett and Kachman1992; de Haas et al. Reference de Haas, Barkema and Veerkamp2002, Reference de Haas, Veerkamp, Barkema, Gröhn and Schukken2004; Jensen et al. Reference Jensen, Hogeveen and De Vries2016).
In this study, our objective was to estimate the parity- and DIM-dependent associations of the first occurrence of pathogen-specific clinical mastitis (CM) with milk yield and milk composition in dairy cows. Five pathogen groups were studied: Streptococcus spp.; Staphylococcus aureus (S. aureus); coagulase-negative staphylococci (CNS); coliforms; and fungi (mainly yeast like). Use of mixed models enabled us to study (1) how much milk yield and milk composition change due to pathogen-specific CM, and (2) when the milk losses and compositional changes start and end in the lactation. The models provide estimates that can be used to calculate precise costs of CM, and provide better indicators of mastitis that could be useful for earlier and automatic detection of infection and more timely treatment of infected cows.
Materials and methods
Herd descriptions
Thirty-one Holstein herds, from Obihiro City in Hokkaido, Japan, participated in the study. Average herd size was 54 milking cows (range: 32 to 107 cows) with an average 305-d milk yield of 9187 kg/cow (range: 7107 to 11 956 kg/cow), and monthly mean SCC across the population of the tested cows of 197 200 (range: 96 991 to 299 053) cells/mL. Data on milk production, milk composition, parity, reproductive performance, diseases, calving, drying-off, and culling were collected from June 2011 until February 2014. Most cows (21 herds) were housed in tie stalls, and the others (10 herds) were housed in free stalls in covered barns. They were fed either total mixed ration (TMR) or by separate feeding. Most of them were milked twice a day.
Data collection
The test-day data of milk yield and milk composition, and medical records, were obtained from the Obihiro Husbandry Center and Tokachi Agricultural Insurance Association, respectively. The test-day milk data were collected monthly in each farm. The test-day milk yield was estimated from single milkings (alternate (AT) method, Smith & Pearson, Reference Smith and Pearson1981) in 25 herds or directly measured from evening-morning samples in 6 herds. Milk sampled by staff in the Obihiro Husbandry Center was carried in a box containing refrigerant and artificial preservatives to the Milk Testing Laboratory in Obihiro City in Hokkaido, Japan. FOSS MilkoScan™ FT+ was used to measure milk composition including SCC, concentration of milk fat, protein, lactose, solids-not-fat (SNF) and milk urea nitrogen (MUN), using milk samples heated at 40 °C (range: 38–42 °C). Solids-not-fat was excluded from the analysis, since it included protein content.
Case definitions
Fourteen veterinarians identified the CM cases, according to their prescription for diagnosis, characterised by changes in milk consistency or hard swollen udder. All CM cases were diagnosed by veterinarians’ clinical examination and microbial culture based on requests from dairy farmers. In order to identify pathogens, milking samples were first collected by farmers according to the prescription: (1) scrub teat ends with alcohol swab, (2) discard a few streams of milk from the teat and (3) collect milk sample in a sterile tube. The milk samples were then refrigerated at 4 °C, sent to a laboratory for analysis, plated on 5% sheep blood agar and incubated aerobically overnight at 37 °C. Identification of isolates was according to Gram staining. If it was impossible to identify isolates, additional biochemical tests, such as the catalase test and cytochrome-oxidase test, were performed. Antibiotic susceptibility tests were performed to select the most appropriate antimicrobial agent as necessary. If 2 different pathogens (e.g. Streptococcus spp. and coliforms) were isolated on the same day, the veterinarian determined which pathogen would contribute to the analysis, based on bacterial counts. This study focused on CM cases that occurred on DIM ≤ 365. Repeated CM cases of any pathogen (e.g. Strep. spp. then some other pathogen), which were incorporated into the mixed model in order to exclude their effects on the changes in milk after the first CM cases, were assumed to occur ≥14 d after the previous CM (Barkema et al. Reference Barkema, Schukken, Lam, Beiboer, Wilmink, Benedictus and Brand1998; Hertl et al. Reference Hertl, Schukken, Bar, Bennett, González, Rauch, Welcome, Tauer and Gröhn2011). Any CM case occurring <14 d after the previous case (of the same pathogen) was considered to be the same case.
The other 4 diseases (displaced abomasum (DA), milk fever, puerperal fever and ketosis), which were diagnosed and treated by veterinarians, were included in the model as potential confounders. This study focused on the cases of DA, milk fever, puerperal fever and ketosis that occurred on DIM ≤ 365.
Statistical analysis
The 43 149 test-day records (primipara: 13 054, multipara: 30 095) of milk yield and composition, from 3149 cows (primipara: 1763 and multipara: 2422; a cow can belong in both classes) with 5996 lactations (primipara: 1763 and multipara: 4233), were analysed. All records for cows that experienced mastitis before the first calving were excluded from the dataset. Also, all records for 8th and higher lactations were excluded from the dataset.
The associations of the first occurrence of pathogen–specific CM with weekly milk yield and milk composition for that same week (week in milk, WIM) were studied using monthly test-day milk data. Estimation of weekly milk production and composition from the many monthly test-day milk data was possible, since each cow had different WIM on the test day. Other control variables in the model were herd, parity, calving season, WIM, and the other four diseases. The R procedure, lme (R version 3.1·2, 2014), was used to fit mixed models. Herd was modelled as a random (intercept) effect. The other variables were modelled as fixed effects. A first-order autoregressive correlation structure among the repeated measurements of milk yield and composition within a cow's lactation was incorporated. These models were adopted from earlier research on the effects of pathogen-unidentified or -specific CM on milk yield (Gröhn et al. Reference Gröhn, Wilson, González, Hertl, Schulte, Bennett and Schukken2004; Wilson et al. Reference Wilson, González, Hertl, Schulte, Bennett, Schukken and Gröhn2004; Bar et al. Reference Bar, Gröhn, Bennett, González, Hertl, Schulte, Tauer, Welcome and Schukken2007; Hertl et al. Reference Hertl, Schukken, Welcome, Tauer and Gröhn2014). Wilson et al. (Reference Wilson, González, Hertl, Schulte, Bennett, Schukken and Gröhn2004) showed a first-order autoregressive correlation structure among the repeated measurements of milk yield was the best structure in their analyses.
Parity groups (first (‘parity 1’), and second and greater (‘parity ≥2’)) were analysed separately. For parity ≥2 cows, the following mixed model with a correlated error term, represented by e, was used for study of pathogen-specific CM:
$$\! \! \! \eqalign{Y =\,& {\rm parity }\;({\rm 6}\;{\rm index}\;{\rm variables}) \cr &+ {\rm calving}\;{\rm season }\; ({\rm 4}\;{\rm index}\;{\rm variables}) \cr & + \,{\rm WIM }\;({\rm 44}\;{\rm index}\;{\rm variables}) + {\rm DA }\;({\rm 6}\;{\rm index}\;{\rm variables}) \cr & + \,{\rm milk fever }\;({\rm 6}\;{\rm index}\;{\rm variables}) \cr & + {\rm puerperal}\;{\rm fever (6}\;{\rm index}\;{\rm variables)} \cr & + \,{\rm ketosis}\;\;({\rm 6}\;{\rm index}\;{\rm variables}) \cr &+ Streptococcus\,{\rm spp}. {\rm }\;({\rm 8}\;{\rm index}\;{\rm variables}) \cr & + \,S.{\rm }aureus\;({\rm 8}\;{\rm index}\;{\rm variables}) + {\rm CNS}\;({\rm 8}\;{\rm index}\;{\rm variables}) \cr & + \,{\rm Coliforms}\;({\rm 8}\;{\rm index}\;{\rm variables}) + {\rm Fungi}\;({\rm 8}\;{\rm index}\;{\rm variables}) \cr & + \,{\rm others}\;({\rm 8}\;{\rm index}\;{\rm variables}) + {\rm herd }({\rm random}) + {\rm e},} $$where Y is test-day milk yield or composition. Separate models were fit for each Y (i.e. milk yield or each composition (SCC, lactose, fat, protein content in milk, MUN)). Within the older group of parity ≥2 cows, parity was subdivided into 6 categories of parity 2, 3, 4, 5, 6, and 7. Calving season had 4 categories: March through May, June through August, September through November, and December through February. WIM was modelled as an index variable, representing 44 levels for the first 44 week of lactation, which covers 305 d of lactation. For each disease, an index variable was created to classify the milk yield and composition according to when they were measured in relation to disease occurrence. For each mastitis pathogen, an index variable was created with 7 levels: ≥29, 15–28, 1–14 d before diagnosis, the same day (0)–13, 14–27, ≥28 d after diagnosis without any more cases of CM and ≥28 d after diagnosis and experienced second and more cases of CM. This index enabled us to precisely determine when mastitis affected milk yield and composition, even before diagnosis. Each of the other diseases (DA, milk fever, puerperal fever and ketosis) was modelled with 6 levels: before diagnosis, same day (0)–14, 15–28, 29–42, 43–56, and ≥57 d after diagnosis. These other diseases occurred very early in lactation. For primipara (parity1), the above model excluding the parity term was fit.
Results
Descriptive findings
Table 1 presents the lactational incidence risk, number, median and mean DIM (and range) of clinical cases in the 31 Hokkaido dairy herds in Japan by parity group (1, and ≥2) and for each pathogen studied. Streptococcus spp. and coliforms were two of the most commonly isolated pathogens, followed by CNS. For most pathogens except fungi, the incidence of CM was higher among older cows than those in parity 1. Cases of CM occurred throughout lactation. The median day of lactation for diagnosis of most cases occurred in midlactation. Cases of CM by fungi tended to occur later than the other pathogens. Most of these findings, except for CNS and fungi, are in line with previous studies (e.g. Gröhn et al. Reference Gröhn, Wilson, González, Hertl, Schulte, Bennett and Schukken2004). Descriptive findings for repeated CM cases of any pathogen are shown in the online Supplementary file, and Supplementary Table S1 shows the lactational incidence risk of the other diseases controlled for in the models estimating milk loss and compositional change in milk.
Table 1. Lactational incidence risk (number of clinical cases) and median and mean days in milk (and range) of the first occurrence of pathogen-specific clinical mastitis by parity group in 31 Hokkaido dairy herds in Japan
† Parity 1 (n = 1763), Parity ≥2 (n = 4233)
‡ ‘Others’ group includes the cases in which (1) pathogens were not identified (main cases in this group), (2) other minor pathogens such as Pseudomonas aeruginosa and Trueperella pyogenes, which caused only a few cases of CM in this study, were identified, and (3) 3 or more different pathogens were isolated on the same day
Milk losses and compositional changes
Figures 1–6 present estimated milk losses and compositional changes associated with the first occurrence of each CM pathogen in each parity group (1 and ≥2), assumed calved in mid spring and the random effect for herds was averaged. The results of each component (milk yield or composition) in each parity group were estimated by a model in this study that included all pathogens simultaneously, and by using monthly milk yield and milk composition of non-mastitic and mastitic cows. For the mastitic cow, diagnosis of CM was assumed to occur on the median DIM of diagnosis of all cows with that pathogen-specific CM (indicated by an arrow). The difference at each point of the curves in each panel corresponds to the estimates in online Supplementary Tables S2 and S3.
Fig. 1. Estimated lactation curves from 1763 parity 1 and 2422 parity ≥2 cows, respectively. Each curve is for a mastitic cow (- -Δ- -) infected with each pathogen (in separate panels) and for a non-mastitic cow ( ____ ). The significant change is represented by (- -▲- -, black triangle). The arrow indicates median DIM of diagnosis of the mastitic cow.
Fig. 2. Estimated somatic cell count (SCC) curves from 1763 parity 1 and 2422 parity ≥2 cows, respectively. Each curve is for a mastitic cow (- -Δ- -) infected with each pathogen (in separate panels) and for a non-mastitic cow ( ___ ). The significant change is represented by (- -▲- -, black triangle). The arrow indicates median DIM of diagnosis of the mastitic cow.
Fig. 3. Estimated lactose curves from 1763 parity 1 and 2422 parity ≥2 cows, respectively. Each curve is for a mastitic cow (- -Δ- -) infected with each pathogen (in separate panels) and for a non-mastitic cow ( ___ ). The significant change is represented by (- -▲- -, black triangle). The arrow indicates median DIM of diagnosis of the mastitic cow.
Fig. 4. Estimated fat curves from 1763 parity 1 and 2422 parity ≥2 cows, respectively. Each curve is for a mastitic cow (- -Δ- -) infected with each pathogen (in separate panels) and for a non-mastitic cow ( ___ ). The significant change is represented by (- -▲- -, black triangle). The arrow indicates median DIM of diagnosis of the mastitic cow.
Fig. 5. Estimated protein curves from 1763 parity 1 and 2422 parity ≥2 cows, respectively. Each curve is for a mastitic cow (- -Δ- -) infected with each pathogen (in separate panels) and for a non-mastitic cow ( ___ ). The significant change is represented by (- -▲- -, black triangle). The arrow indicates median DIM of diagnosis of the mastitic cow.
Fig. 6. Estimated milk urea nitrogen (MUN) curves from 1763 parity 1 and 2422 parity ≥2 cows, respectively. Each curve is for a mastitic cow (- -Δ- -) infected with each pathogen (in separate panels) and for a non-mastitic cow ( ___ ). The significant change is represented by (- -▲- -, black triangle). The arrow indicates median DIM of diagnosis of the mastitic cow.
Our findings are summarised below and detailed descriptions of the results are presented in the Supplementary File. The associations of non-CM diseases (DA, milk fever, puerperal fever and ketosis) with milk yield and milk composition are presented in Supplementary Tables S4 and S5.
All pathogens were associated with significant milk losses for at least 2 weeks after the diagnosis, particularly in older cows (Fig. 1). The estimation showed that higher yielding (particularly older) cows were at greater risk for the infection, as Gröhn et al. Reference Gröhn, Wilson, González, Hertl, Schulte, Bennett and Schukken2004 have previously reported. Key findings from our study regarding milk losses, in older cows, are (1) the greatest and longest-term milk losses were caused by S. aureus and fungi and (2) recovery of potential milk production in CNS-infected cows and mild recovery in Streptococcus spp.-infected cows. (3) Only coliforms caused significant long-term milk losses, and S. aureus and CNS infections did not show any significant milk losses, in younger cows.
All pathogens, in particular S. aureus and fungi, significantly increased SCC from at least 2 weeks before the diagnosis until 2 weeks–1 month after the diagnosis in both parity groups (Fig. 2). Staphylococcus aureus was associated with significantly high SCC throughout the lactation even before the diagnosis (from just after calving; this might imply subclinical infection due to S. aureus). In younger cows, the peak SCC level due to S. aureus or fungi infection reached over 500 000–1 000 000 cells/mL. The highest peak SCC level due to coliforms infection reached over 500 000 cells/mL in older cows.
All pathogens, particularly CNS, S. aureus, and also coliforms and fungi, decreased lactose content for 2 weeks–1 month after diagnosis and, sometimes, throughout the lactation. Effects of CNS or S. aureus infections on lactose decline were different between younger and older cows (Fig. 3). All pathogen groups were associated with significant changes in fat (Fig. 4), and all pathogen groups except for fungi were associated with significant changes in protein (Fig. 5). Streptococcus spp., coliforms and S. aureus infections were associated with significantly low MUN (Fig. 6). Some pathogens such as Streptococcus spp. and coliforms seemed to be associated with long-term fat, protein and MUN changes.
Discussion
Classification of the infections (pathogens) based on milk
Staphylococcus aureus and fungi were associated with the most significant milk losses in older cows and high SCC content in both parity groups. These highlight the severity of S. aureus and fungi infections, although they are not the most common pathogens (their lactational incidence risk was 1–2%, Table 1). Also, the high SCC level of older cows that became infected by S. aureus implies that they may be subclinically infected. Interestingly, in younger cows, these pathogens did not cause large milk losses (only temporary losses by fungi infection). This does not imply mildness of the infections, but the immune system of younger cows might overcome the infections due to S. aureus and fungi. Also, S. aureus might not cause significant damage to the mammary gland of younger cows (since lactose contents did not decrease due to the infection) but might cause significant damage to the mammary gland of older cows (since persistent low lactose contents due to the infection were estimated), where the damage to the mammary gland might be functional and/or structural (related to tight junction status). It might be possible to automatically detect S. aureus infections based on (1) higher SCC level even in early lactation in both parity groups, (2) persistent low lactose content, (3) temporary high fat content and (4) low MUN, in older cows. Fungi infections may be detected by high SCC level and low lactose content. A high SCC level, over 500 000–1 000 000 cells/mL in younger cows, might imply S. aureus or fungi infection.
Coliforms caused long-term milk losses, as previous works have shown (Gröhn et al. Reference Gröhn, Wilson, González, Hertl, Schulte, Bennett and Schukken2004; Bezman et al. Reference Bezman, Lemberskiy-Kuzin, Katz, Merin and Leitner2015). In this study, the lactational incidence risks were 4·9% in primipara and 7·5% (the highest) in multipara. Coliforms infections may specifically be detected by (1) high fat content in younger cows and (2) high protein content in both parity groups, while such infections also caused high SCC levels (particularly in older cows), low lactose content, and low MUN.
Streptococcus spp. infections also caused milk losses. Their lactational incidence risk was high (6–7%, Table 1). Specific compositional changes associated with Streptococcus spp. infections were (1) persistent low fat content in older cows, (2) persistent high protein content in younger cows, and (3) persistent low MUN in both parity groups.
CNS did not show persistent milk losses in either parity group, implying mildness of the infection. However, the lactational incidence risk was relatively high (2–3%, Table 1), and it might be very important to distinguish CNS infections from other infections. Specific compositional changes in CNS infections were (1) persistent low lactose in younger cows, (2) low fat content before diagnosis in older cows, and (3) low protein content before diagnosis in both parity groups. The persistent low lactose in younger cows, not in older cows, might reflect persistent damage to their mammary gland due to weakness of the younger cows (such as a defective immune system). The low protein content before diagnosis might reflect insufficient energy intake that may cause vulnerability to CNS infections. These CNS-specific changes in milk might be helpful to distinguish CNS infections from other infections.
Comments on the changes in milk
Milk losses due to CM have been well studied even as pathogen-specific and parity-and DIM-dependent (e.g., Gröhn et al. Reference Gröhn, Wilson, González, Hertl, Schulte, Bennett and Schukken2004). The high SCC content due to mastitis is also well studied, even as pathogen-specific and DIM-dependent, but not parity-specific and not including fungi (e.g. de Haas et al. Reference de Haas, Barkema and Veerkamp2002 and Reference de Haas, Veerkamp, Barkema, Gröhn and Schukken2004), although Green et al. (Reference Green, Green, Schukken, Bradley, Peeler, Barkema and Medley2004) have reported SCC distributions during lactation for the prediction of CM.
Previous studies reported lactose decline due to pathogen-specific CM, but not parity- and DIM-dependent. Coulon et al. (Reference Coulon, Gasqui, Barnouin, Ollier, Pradel and Promiès2002), Leitner et al. (Reference Leitner, Krifucks, Merin, Lavi and Silanikove2006), Bezman et al. (Reference Bezman, Lemberskiy-Kuzin, Katz, Merin and Leitner2015) and Kester et al. (Reference Kester, Sorter and Hogan2015) showed lower lactose contents in milk caused by S. aureus, E. coli, some species of Streptococcus and CNS, although some of these works did not provide parity information, and Botaro et al. (Reference Botaro, Cortinhas, Dibbern, Silva, Benites and dos Santos2015) reported no significant lactose change by S. aureus for subclinical cases.
Clinical mastitis might be associated with fat, protein and MUN changes in milk. Previous works regarding fat and protein changes due to CM, although some of them were not pathogen-specific and DIM- and parity-dependent, showed high fat and low protein contents (Coulon et al. Reference Coulon, Gasqui, Barnouin, Ollier, Pradel and Promiès2002; Jamrozik & Schaeffer, Reference Jamrozik and Schaeffer2012) and, conversely, low fat contents due to Streptococcus spp. infections (Bezman et al. Reference Bezman, Lemberskiy-Kuzin, Katz, Merin and Leitner2015). Other studies noted no significant change in fat and protein content due to mastitis (Botaro et al. Reference Botaro, Cortinhas, Dibbern, Silva, Benites and dos Santos2015; Tomazi et al. Reference Tomazi, Gonçalves, Barreiro, Arcari and dos Santos2015). The association of CM and MUN has not been focused on, but it might be worth further study, particularly for Streptococcus spp. and coliforms infections.
Our findings regarding milk losses and compositional changes can provide parity- and DIM-dependent parameter estimates which can be used to obtain precise cost estimation and better indicators of pathogen-specific CM.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0022029918000456.
This study used the dataset of the herd tests and medical records given by the Obihiro Husbandry Center and Tokachi Agricultural Insurance Association. The authors are grateful to Akihiro Kamikawa for helpful discussion related to the mammary gland and milk composition.
