effect of repeated episodes of generic clinical mastitis on milk yield in dairy cows

J. Dairy Sci. 90:4643–4653doi:10.3168/jds.2007-0145© American Dairy Science Association, 2007.

Effect of Repeated Episodes of Generic Clinical Mastitison Milk Yield in Dairy Cows

D. Bar,*1 Y. T. Grohn,* G. Bennett,† R. N. Gonzalez,† J. A. Hertl,* H. F. Schulte,† L. W. Tauer,‡F. L. Welcome,† and Y. H. Schukken†*Section of Epidemiology,†Quality Milk Production Services, Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, and‡Department of Applied Economics and Management, College of Agriculture and Life Sciences, Cornell University, Ithaca, NY 14853

ABSTRACT

Our objective was to estimate the milk losses associ-ated with multiple occurrences of generic bovine clinicalmastitis (CM) within and across lactations. We studied10,380 lactations from 5 large, high-producing dairyherds that used automatic recording of daily milkyields. Mixed models, with a random herd effect andan autoregressive covariance structure to account forrepeated measurements, were used to quantify the ef-fect of CM and other control variables (parity, week oflactation, other diseases) on milk yield. Many cows thatdeveloped CM were higher producers than their non-mastitic herdmates before CM occurred. Milk yield be-gan to drop after diagnosis; the greatest loss occurredin the first weeks (up to 126 kg) and then graduallytapered to a constant value approximately 2 mo afterCM. Mastitic cows often never recovered their potentialyield. First-lactation cows lost 164 kg of milk for thefirst episode and 198 kg for the second in the 2 moafter CM diagnosis, compared with their potential yield.Among older cows, this estimate was 253 kg for thefirst, 238 kg for the second, and 216 kg for the thirdCM case. A cow that had 1 or more CM episodes in herprevious lactation produced 1.2 kg/d less milk over thewhole current lactation (95% confidence interval: 0.6,1.7) than a cow without CM in her previous lactation.These findings provide dairy producers with informa-tion on the average milk loss associated with CM caseswithout considering the causative agent, and can beused for economic analysis.Key words: clinical mastitis, recurrent, dairy cow,mixed model

INTRODUCTION

Mastitis is a common disease in dairy herds in manycountries (Barkema et al., 1998; Rajala-Schultz et al.,

Received February 24, 2007.Accepted June 21, 2007.1Corresponding author: [email protected]

4643

1999; Sviland and Waage, 2002). It can be very detri-mental for dairy farm profitability because of lost pro-duction and treatment costs (Houben et al., 1993; See-gers et al., 2003; Wilson et al., 2004). A dairy producermay decide it is more economical to cull a mastitic cowthan to treat her, if her expected future revenue is lessthan that from a replacement heifer. Clinical mastitis(CM) can be caused by different pathogens, differingin their effects (Grohn et al., 2004) and treatment poten-tial. Still, the current situation is that the farmer hasto weigh options regarding treatment, culling, and pre-ventive measures, because CM milk is not culturedwithout knowledge about the CM-causing agent.

We applied the technique of mixed linear models tostudy the effect of CM without specific pathogen identi-fication on milk yield in both Finnish (Rajala-Schultzet al., 1999) and 2 New York State (Wilson et al., 2004)dairy herds. In the Finnish study, milk losses in thefirst 2 wk after diagnosis ranged from 1.0 to 2.5 kg/d,and the total loss over the entire lactation ranged from110 to 552 kg. In the New York study, milk losses causedby CM in first-parity cows were 5 to 7 kg/d in the first 2wk after diagnosis, and 690 kg over the entire lactation.Among older cows, milk losses caused by CM in thefirst 2 wk following diagnosis ranged from 6 to 9 kg/d,and 570 kg over the entire lactation. Yet among theseolder cows, many mastitic cows were higher producersbefore disease onset than their nonmastitic herdmates,having a potential daily advantage of 2.6 kg. Therefore,the total lactational loss among cows in parity 2+ (Wil-son et al., 2004) was more accurately estimated as 1,155kg. Thus, when studying the effect of a disease on milkyield, it is important to look at repeated measures ofmilk yield (daily, weekly, monthly), rather than a singlesummary measure for the 305-d lactational milk yield(Grohn et al., 1999), because milk yield may be higheramong mastitic than nonmastitic cows before the CMepisode(s).

Mastitis is often a recurrent event (Houben et al.,1993; Dopfer et al., 1999; Zadoks et al., 2001). In ourprevious studies, because of the limited size of the data

BAR ET AL.4644

set and the length of the observation period, only thefirst case of CM could be modeled, masking the possibleeffect of repeated CM cases. Repeated CM cases mighthave different effects on milk losses than the first case.In addition, repeated cases can cause an additive effect(i.e., if 2 cases are closer in time, the resultant milkloss is higher than if they are far apart).

In this study, the objective was to estimate the effectsof multiple occurrences of CM on milk production indairy farms with high milk production and with a lowincidence of contagious mastitis-causing pathogens.

MATERIALS AND METHODS

Herd Descriptions

The data were from 5 dairy farms located in NewYork State. These farms milked an average of 1,200,1,100, 750, 650, and 600 Holstein milking cows andwere followed for approximately 18 mo. The rolling herdaverage was close to 11,000 kg per cow/yr on a 305-dbasis (range 10,700 to 11,500); monthly SCC was225,000 cells/mL (range 180,000 to 355,000) and variedlittle among the 5 farms. Cows were housed in coveredbarns with concrete floors and free stalls and were clas-sified by lactation, production, and reproductive statusinto milking groups. All groups of cows were fed a bal-anced TMR via feed alleys with headlocks that allowedrestraint of cows for examination and treatments. Cowswere milked 3 times daily. Each milking unit had milkmeters capable of automatically recording milk produc-tion and (on 2 farms) milk conductivity. Most cases ofCM were identified by milkers (warm, swollen udderor changes in the milk consistency), whereas otherswere detected by the herdsperson examining cowswhose milk electrical conductivity had increased 30%above their last 10-d mean (Afimilk, SAE Afikim, Israel)or had a concurrent drop in milk production (on mostfarms, this was set to 30% below their average milkyield). Sick cows were treated according to well-definedprotocols that were similar, but not completely thesame, on all farms and throughout the study (one farmalso treated gram-negative CM cases with antibioticsin the first months of the study). Farm personnel usedDairyComp305 herd management software (Valley Ag-ricultural Software, Tulare, CA) to record lactation, re-productive, and medical data for each cow. Informationon parity, diseases, drying off, calving, and culling wasreadily available.

Case Definition

All lactating cows in the 5 study herds were eligiblefor inclusion as cases of CM. Training and standardiza-tion concerning CM detection was provided at the begin-

Journal of Dairy Science Vol. 90 No. 10, 2007

ning of the study. Although we were specifically inter-ested in the milk loss associated with CM withoutknowledge of the causative agent, farm personnel sam-pled milk for microbiological culture from quarters withsigns of CM. The samples were collected daily and werecultured by the Quality Milk Production Services labo-ratories. The bacteriological culture procedures are de-scribed in detail in Grohn et al. (2004).

Some cows had 2 clinical episodes in the same quarterwithin several days of each other. Any such episodethat occurred within 5 d, or that occurred within 14 dwith the same etiologic agent isolated from both occur-rences, was considered the same case of mastitis. Anyepisode that occurred more than 14 d after the previousepisode was considered a new CM case.

Other Diseases

While focusing on CM, we chose 6 other diseases forinclusion in the models as potential confounders. Thesediseases are among the most common clinical condi-tions that are universally a problem in dairy cows, andreliable information about their occurrence was presentin the data set for all farms. The rationale for choosingthem is that they may cause milk loss, in addition tothe effects of CM.

The additional 6 recorded diseases were milk fever,retained placenta, metritis, ketosis, displaced aboma-sum (DA), and pneumonia. They were defined as follows(Wilson et al., 2004): 1) milk fever occurred if a cowwas unable to rise or had cool extremities and sluggishrumen motility near the time of calving, but was treatedsuccessfully with calcium; 2) retained placenta was re-tention of fetal membranes for at least 24 h postcalving;3) metritis involved a febrile state accompanying a pu-rulent or fetid vaginal discharge, or a diagnosis of anenlarged uterus by veterinary palpation; 4) ketosis wasdiagnosed by detection of ketones in milk or urine, andresponse to treatment; 5) DA occurred when the aboma-sum was enlarged with fluid, gas, or both, and wasmechanically trapped in either the left or right side ofthe abdominal cavity (nearly every DA was confirmedby surgery); 6) pneumonia was diagnosed by the pres-ence of pathological breathing sounds (using a stetho-scope). Every effort was taken to ensure that diseasedefinition and diagnostic criteria were the same in allherds. Written disease definitions were provided to thedairy producers and veterinarians involved.

Statistical Methods

The SAS PROC MIXED (SAS Institute, 2006) wasused to study the effects of CM and the control variables[herd, parity, week in milk (WIM), and other diseases]

REPEATED BOVINE MASTITIS CASES AND MILK YIELD 4645

on weekly averaged milk yield in 10,380 lactations. Be-cause we were not interested in individual farms, butrather the farms in general with these characteristics(i.e., large, high-milk-producing dairy farms with a lowincidence of contagious mastitis), herd was modeled asa random (intercept) effect. The other covariates weremodeled as fixed effects.

The outcome variable, weekly averaged milk yield,was calculated by adding the milk weights of the 3 dailymilkings. Within each week of lactation, the 7 dailyvalues were then summed and divided by 7 to givethe mean daily milk yield for the particular week inlactation. Randomly occurring zero-values for a particu-lar milking were filled by using the weekly averagevalue for the particular milking. We used weekly aver-age milk yields over daily measurements because thelatter made the size of the data set unsolvable by usingavailable hardware and did not deliver statistical effi-cacy because daily milk yields had a larger variance.

The data set contained repeated measurements ofmilk yield within a cow over a lactation, and thesewere correlated with one another. This correlation wascorrected for in the regression model by specifying acorrelation structure among the repeated measure-ments (R matrix). In previous work, the first-order auto-regressive correlation structure was found optimal forthis purpose (Wilson et al., 2004); therefore, it was usedin the current analysis.

Parity was divided into 2 groups, which were ana-lyzed separately: first, and second and higher. Withinthe older group, parity was further subdivided into pari-ties 2, 3, 4, 5, 6, and 7. Older cows (>parity 7) were notanalyzed because of the low number of observations.Milk yields were modeled for the first 50 wk of lactation.

The first 3 episodes of CM during the current lacta-tion and a carryover effect of CM in the previous lacta-tion were studied; the other diseases controlled for inthe models were retained placenta, milk fever, metritis,DA, ketosis, and pneumonia.

An index variable for each CM episode was createdto classify the milk weights according to when theywere measured in relation to disease occurrence. Thisenabled precise determination of when a disease hadan effect on milk yield. After the initial data analysis,these indices were collapsed as follows: before diagno-sis; same week as disease diagnosis; and 1, 2, 3, 4, 5,6, 7, and ≥8 wk after diagnosis. The same index schemewas used for the other 6 diseases.

Several carryover effects across lactations were stud-ied. In the final model the simple definition of havingany CM occurrence in the previous lactation was pre-ferred over more complex definitions.

Herd was modeled as a random effect. It was chosenover other possible random effects after the initial data


analysis, based on Akaike’s information criterion (AIC)as a measure of goodness of fit (Wilson et al., 2004).

For parity 2+ cows, the following linear mixed modelwas used:

Y = parity (6 index variables)

+ WIM (50 index variables)

+ milk fever (9 index variables)

+ retained placenta (9 index variables)

+ metritis (10 index variables)

+ DA (10 index variables)

+ ketosis (10 index variables) [1]

+ pneumonia (10 index variables)

+ first CM in the current lactation (10 index variables)

+ second CM in the current lactation (9 index variables)

+ third CM in the current lactation (9 index variables)

+ CM in the previous lactation (0/1)

+ herd (random) + e,

where Y is the mean milk yield per day in a particularweek of lactation, the independent variables are as de-fined above, and e is a complex error term representingthe within-cow correlation of milk weights and residualerror. For parity 1 cows, the same model was used,except that the terms for parity, CM in the previouslactation, and milk fever were omitted because theywere not applicable. The reference cow was always acow free of that disease at the time of milk measure-ment. This parameterization (3 sets of covariates: 1 foreach CM occurrence) assumes an additive carryovereffect of the previous CM case on the current CM milkloss. For example, if a cow had her second CM case 4wk after the first case, in that week, her milk yield wasassumed to be the sum of the effect the first case caused(i.e., 4 wk after first CM) and the milk loss associatedwith the same week of the second CM case.

Primipara and multipara were analyzed separatelybecause of the greatly differing shapes of their lactationcurves and a possibly different CM effect. After re-stricting the lactation follow-up period to the first 50WIM in the mixed model analysis, there were 3,681parity 1 cows and 6,699 parity 2+ cows. In the analysisof parity 1 cows, 112,475 weekly milk weights wereused. In the analysis of parity 2+ cows, 192,691 weeklyobservations were used.

To represent a CM effect, regardless of the time ofoccurrence of the previous CM, the parameterizationof the CM was changed so that only one index (M) was

BAR ET AL.4646

used to represent the time of the milk measurement inrelation to the time of CM occurrence. For parity 2+cows, the following linear mixed model was used:

Y = parity (6 index variables)

+ WIM (50 index variables)

+ milk fever (9 index variables)

+ retained placenta (9 index variables)

+ metritis (10 index variables)

+ DA (10 index variables) [2]

+ ketosis (10 index variables)

+ pneumonia (10 index variables)

+ CM in the current lactation (28 index variables)

+ CM in the previous lactation (0/1)

+ herd (random) + e,

where all terms are identical to the previous modelexcept for the representation of the current lactationCM effects. This parameterization (1 covariate with 28index variables expressing when the milk weights weremeasured in relation to the first 3 CM cases) assumedthat the losses observed were associated with this par-ticular CM occurrence without adjusting the losses po-tentially due to the previous CM. For example, if a cowhad her second CM case 4 wk after the first case, inthat week, her milk loss was modeled as the milk lossassociated with the second CM case without separatelymodeling her previous CM history. Table 1 illustratesthe different coding schemes used for model [1] (M1,M2, and M3) vs. model [2] (M). The reference cow wasalways a cow free of that disease at the time of milkmeasurement.

To calculate the milk losses associated with CM, thepotential milk yield of a CM cow was defined as thoughshe would keep her prediseased milk yield levelthroughout the lactation. Therefore, the milk loss asso-ciated with CM was compared with her predisease level(corrected for the other covariates). To calculate thecumulative milk losses associated with CM, the milkyield losses of a CM-diseased cow were summed asthough she would get CM at the median day for eachof the CM occurrences. All first CM cases were takenfor the estimation of the first case; the same was truefor the second CM case. The data were not stratifiedby number of CM occurrences within lactation, becauseat the time of CM the farmer does not know whetherthis case will be followed by another. The incidencesof the diseases modeled were calculated as lactational


incidence risks (LIR; i.e., the probability of a cow havingthe disease in 1 lactation).

RESULTS

Descriptive Findings

Table 2 gives the number of occurrences, LIR, andmedian WIM for the diseases present. The LIR of thefirst CM episode in multipara was twice that in primip-ara (P < 0.0001). Lactational incidence risks for subse-quent CM cases were higher in multipara comparedwith primipara (P < 0.0001). The median WIM (13 WIM)for the first CM episode was the same in both agegroups. Nevertheless, a second CM case occurred soonerfor multipara (20 WIM) and later for primipara (26WIM; P < 0.0001). The same trend was observed for athird CM case.

The LIR of retained placenta was nearly 2 timeshigher in multipara compared with primipara (P <0.0001); the opposite was true for metritis (P < 0.0001).The LIR for other diseases studied were comparable inboth groups of cows. The median WIM for all non-CMdiseases studied, except for pneumonia, was very earlyin lactation (1 to 2 WIM).

Although we did not model the mastitis-causingpathogens separately, the bacteriological results of thesamples taken at the time of CM are presented in Table3. The vast majority of CM cases were environmentalbacteria (P < 0.0001), with Escherichia coli, Streptococ-cus spp., “no growth” (i.e., fewer than 2 colonies perplate), and Klebsiella being the most common findings.

The number of real repeated cases caused by the samepathogen (detailed data not shown) was of interest.Following our definition of recurrent CM cases, 40% ofsecond cases were due to the same pathogen as in thefirst case, and 51% of third cases were due to the samepathogen as in the first or second case.

The estimated lactation curves for the first 305 DIMare graphically presented for primiparous cows in Fig-ure 1 and for multiparous cows in Figure 2. The esti-mated effect of parity was 1.2, 0.3, 0.0, −2.1, and −3.4kg of milk for parity 3, 4, 5, 6, and 7, respectively,compared with a cow in her second lactation. Standarderrors for these estimates were 0.15, 0.18, 0.24, 0.37,and 0.57 kg of milk, respectively.

Estimates of Milk Loss Associated with CM

Estimates for the effects of the first 3 occurrences ofCM are given for primiparous (Table 4) and multipa-rous (Table 5) cows for the parameterization describedin model [1]. The same estimates are given for the alter-native parameterization (model [2]) in Tables 6 and 7,respectively. The estimates of repeated CM cases from


Table 1. Data for an example cow with 2 clinical mastitis (CM) cases illustrating the 2 covariate codingschemes used in the statistical analysis of this study

Week of Index variable Index variable Index variableWeek Week of first second for first CM for second for both CMin milk CM case CM case case (M1)1 CM case (M2)1 cases (M)2

11 13 17 1 111 112 13 17 1 111 113 13 17 2 111 214 13 17 3 111 315 13 17 4 111 416 13 17 5 111 517 13 17 6 2 1218 13 17 7 3 1319 13 17 8 4 1420 13 17 9 5 1521 13 17 10 6 16

1Used in model [1]: Y = parity + week in milk + milk fever + retained placenta + metritis + displacedabomasum + ketosis + pneumonia + first CM in the current lactation (10 index variables) + second CM inthe current lactation (9 index variables) + third CM in the current lactation (9 index variables) + CM inthe previous lactation + herd (random) + e. The index variables indicate when milk weights were measuredin relation to CM occurrence: 1 = before first CM; 2 = week of CM occurrence; 3 = second week after CMoccurrence; 4 = third week after CM occurrence; . . .; 10 = ninth week after CM occurrence; 111 = referencelevel (free of CM at the time of milk measurement).

2Used in model [2]: Y = parity + week in milk + milk fever + retained placenta + metritis + displacedabomasum + ketosis + pneumonia + CM in the current lactation (28 index variables) + CM in the previouslactation + herd (random) + e. The index variables indicate when milk weights were measured in relationto CM occurrence: 1 = before first CM; 2 = week of first CM occurrence; 3 = second week after first CMoccurrence; 4 = third week after first CM occurrence; 5 = fourth week after first CM occurrence; 12 = weekof second CM occurrence; 13 = second week after second CM occurrence; 14 = third week after second CMoccurrence; 15 = fourth week after second CM occurrence; 16 = fifth week after second CM occurrence.

model [2] are generally slightly higher. Model [2] re-sulted in a better model fit (P < 0.0001). The AIC inprimipara decreased to 564,717 from 572,300, and inmultipara the AIC of model [2] was 1,085,058 vs.1,090,422 in model [1]. For incorporation of these re-sults into an economic model, the parameterization ofmodel [2] was simpler (because no memory variablesfor previous CM cases were needed). Therefore, in thefollowing presentation, only the results of model [2]are discussed.

In mastitic primipara, the greatest milk losses oc-curred immediately after diagnosis of CM (Table 6). In

Table 2. Number of cases, lactational incidence risk (LIR), and median week at occurrence (WIM) for thefirst 3 clinical mastitis (CM) cases and the other 6 diseases controlled for in the model analyzing the effectof CM on milk yield

Primipara Multipara

Item n LIR, % WIM n LIR, % WIM

Cows 3,681 6,699First CM 448 12 13 1,639 24 13Second CM 87 2 26 551 8 20Third CM 24 1 35 216 3 27Milk fever NA1 NA NA 137 2 1Retained placenta 216 6 1 745 11 1Metritis 287 8 2 267 4 1Ketosis 166 5 2 488 7 2Displaced abomasum 101 3 2 265 4 2Pneumonia 106 3 8 165 2 14

1NA = not applicable.


the first 7 wk after diagnosis of the first and secondCM cases in primipara, milk yield in mastitic cows re-mained well below (P < 0.01) that of non-CM cows [see95% confidence intervals (CI)]. For a third episode, incontrast, milk loss (P < 0.001) occurred only in the sameweek of diagnosis (this estimate was based on 24 cases);therefore, the effect of the third CM case in primiparawas not considered in the following discussion. Aftertheir first CM case, primiparous cows were producingless milk (P < 0.0001) than their non-CM herdmates,even after more than 8 wk after CM diagnosis. In addi-tion, before the first CM episode, mastitic primipara

BAR ET AL.4648

Table 3. Pathogens isolated from milk of the first, second, and third case of clinical mastitis (CM)

Primipara Multipara

First Second Third First Second ThirdItem CM CM CM CM CM CM

Cases, n, 448 87 24 1,639 551 216Escherichia coli, % 27 23 29 29 21 18Klebsiella, % 5 6 4 11 13 18No growth,1 % 10 14 21 18 21 25Streptococcus spp., % 18 24 13 18 19 17Staphylococcus aureus, % 6 0 4 4 5 6CNS, % 4 2 4 2 2 2Arcanobacterium pyogenes, % 2 2 4 1 2 1Yeast, % 4 6 0 1 1 0Other,2 % 6 10 4 4 5 3Undefined,3 % 18 13 17 11 12 9

1Fewer than 2 colonies per plate.2Any organism not specified above.3More than 3 different organisms isolated (i.e., contaminated sample).

actually outproduced (P < 0.001) their non-CM herd-mates by 0.7 kg/d.

Figure 1 graphically displays the information ob-tained by the statistical model (model [2]). Althoughmastitic primipara had a slight production advantagebefore CM diagnosis (0.7 kg/d, 95% CI: 0.3, 1.0), thissoon vanished upon diagnosis. The milk yield of mas-

Figure 1. Effect of clinical mastitis (CM) on the lactation curve of primiparous cows (LSM from 3,681 lactations). The solid line withfilled circles represents a cow with 2 CM occurrences (generic); the dashed line with open squares represents a cow without CM (healthy).The arrows indicate median weeks in milk of CM. The dotted line portrays the estimated lactation curve of the CM-diseased cow if she hadremained CM free (potential). Standard errors for these estimates (first 43 dummy variables) were in the range of 0.16 to 0.17 kg of milkin primipara.


titic cows remained well below (P < 0.0001) that of theirnon-CM herdmates, and dropped further with a subse-quent episode. Figure 1 displays the potential lactationcurve of the mastitic cows if they had not contractedCM. Compared with CM-free cows, the first CM episodewas associated with a milk loss of 126 kg in the first 2mo, and the second episode with 160 kg. Considering


Figure 2. Effect of clinical mastitis (CM) on the lactation curve of multiparous cows (LSM from 6,699 lactations). The solid line withfilled circles represents a cow in the second lactation with 3 CM occurrences (generic); the dashed line with open squares represents a cowin the second lactation without CM (healthy). The arrows indicate median weeks in milk of CM. The dotted line portrays the estimatedlactation curve of the CM-diseased cow if she had remained CM free (potential). Standard errors for these estimates (first 43 dummyvariables) were in the range of 0.18 to 0.19 kg of milk in multipara.

the potential milk yield of CM-diseased cows, thesecows lost 164 and 198 kg of milk in the 2 mo after thefirst and second CM episodes, respectively.

In multiparous cows (Table 7) before CM, cows thatwould go on to develop this disease were producing 1.7kg/d more milk than their nonmastitic herdmates. Once

Table 4. Effects of the first 3 occurrences of generic clinical mastitis (CM) on milk yield in 3,681 parity 1cows on 5 New York State dairy farms1

First CM Second CM Third CM

95% CI3 95% CI 95% CIRelative time2 Estimate Estimate Estimate

Before CM 0.7 0.3 1.0Same week −3.8 −4.2 −3.4 −4.8 −5.5 −4.1 −1.8 −3.2 −0.4+1 wk −3.8 −4.3 −3.3 −3.9 −4.9 −2.9 −1.3 −3.1 0.5+2 wk −2.5 −3.1 −2.0 −2.4 −3.5 −1.2 0.6 −1.5 2.7+3 wk −1.7 −2.2 −1.1 −2.1 −3.3 −0.9 0.5 −1.7 2.8+4 wk −1.2 −1.7 −0.6 −2.2 −3.4 −0.9 −0.1 −2.5 2.2+5 wk −0.9 −1.5 −0.3 −1.5 −2.8 −0.3 −0.6 −2.9 1.7+6 wk −1.2 −1.7 −0.6 −1.6 −2.8 −0.3 0.5 −1.8 2.8+7 wk −1.0 −1.5 −0.5 −1.2 −2.4 0.0 2.0 −0.2 4.1+8 wk and after −1.0 −1.5 −0.6 −0.4 −1.4 0.6 2.0 0.2 3.8

1Estimates were obtained from a mixed model with an autoregressive covariance structure and randomherd effect. Clinical mastitis episodes were allowed to have simultaneous effects in the same time period.Values are kg/d of milk.

2Time of milk measurement in relation to disease occurrence.3CI = confidence interval.


CM had occurred, the picture changed markedly. Forthe first episode, milk loss (P < 0.0001) occurred for 6wk after diagnosis. For the second and third episodes,milk loss (P < 0.0001) continued for 5 and 4 wk, respec-tively. Within each episode, milk loss was greatest im-mediately after diagnosis, and then tapered off in subse-

BAR ET AL.4650

Table 5. Effects of the first 3 occurrences of generic clinical mastitis (CM) on milk yield in 6,699 parity 2+cows on 5 New York State dairy farms1



Before CM 1.8 1.5 2.1Same week −4.8 −5.2 −4.5 −4.2 −4.5 −3.8 −3.1 −3.7 −2.5+1 wk −6.0 −6.4 −5.6 −5.0 −5.5 −4.5 −4.2 −5.0 −3.5+2 wk −3.6 −4.0 −3.2 −2.9 −3.4 −2.3 −2.8 −3.7 −1.9+3 wk −2.4 −2.8 −2.0 −1.9 −2.5 −1.3 −2.1 −3.1 −1.1+4 wk −1.6 −2.0 −1.2 −1.5 −2.1 −0.8 −1.6 −2.7 −0.6+5 wk −1.0 −1.4 −0.6 −0.9 −1.6 −0.3 −0.9 −2.0 0.1+6 wk −0.5 −0.9 −0.1 −0.5 −1.1 0.2 −0.8 −1.8 0.2+7 wk 0.0 −0.3 0.4 0.0 −0.6 0.6 −0.1 −1.1 0.8+8 wk and after 0.0 −0.3 0.3 0.0 −0.5 0.5 −0.1 −0.9 0.7

1Estimates were obtained from a mixed model with an autoregressive covariance structure and randomherd effect. Clinical mastitis episodes were allowed to have simultaneous effects in the same time period.Values are kg/d of milk.

2Time of milk measurement in relation to disease occurrence3CI = confidence interval.

quent weeks until production returned nearer to levelsof non-CM cows. Nonetheless, as seen in Figure 2 (re-sults from model [2]), even by the end of lactation, pro-duction of CM cows remained well below (P < 0.0001)that of their potential. After their first CM case, multip-arous cows were producing substantially less milk (P <0.0001) than their non-CM herdmates, even after morethan 8 wk after CM diagnosis. Compared with CM-freecows, the first CM episode was associated with a milkloss of 156 kg in the first 2 mo, the second episode with141 kg, and the third with 119 kg. Considering thehigher potential milk yield of CM-diseased cows (1.7kg/d), these cows lost 253, 238, and 216 kg of milk forthe above-mentioned period and cases, respectively.

If a cow experienced CM in her previous lactation,she produced less milk in her subsequent lactation. We

Table 6. Effects of the first 3 occurrences of generic clinical mastitis (CM) on milk yield in 3,681 parity 1cows on 5 New York State dairy farms1



Before CM 0.7 0.3 1.0Same week −3.8 −4.3 −3.4 −5.8 −6.6 −5.0 −2.9 −4.6 −1.2+1 wk −3.8 −4.3 −3.3 −4.5 −5.5 −3.5 −1.9 −3.9 0.1+2 wk −2.6 −3.1 −2.0 −2.9 −4.1 −1.8 0.4 −1.8 2.6+3 wk −1.9 −2.5 −1.4 −2.5 −3.7 −1.3 1.2 −1.1 3.4+4 wk −1.7 −2.3 −1.1 −2.6 −3.9 −1.3 0.5 −1.8 2.9+5 wk −1.4 −2.0 −0.8 −2.1 −3.5 −0.7 0.7 −1.6 3.0+6 wk −1.5 −2.1 −0.9 −1.5 −3.0 −0.1 1.0 −1.3 3.4+7 wk −1.1 −1.7 −0.6 −1.0 −2.4 0.4 1.6 −0.5 3.8+8 wk and after −1.4 −1.9 −0.9 −0.1 −1.3 1.2 1.0 −0.8 2.8

1Estimates were obtained from a mixed model with an autoregressive covariance structure and randomherd effect. Clinical mastitis episodes were not allowed to have simultaneous effects in the same time period.Values are kg/d of milk.



estimated this effect as 1.2 kg/d over the whole lactation(95% CI: 0.6, 1.7). There were no substantial differences(P > 0.05) either by number of occurrences of CM or bythe time when CM occurred in the previous lactation(data not shown).

Estimates of Milk Loss Associatedwith Other Diseases

In addition to CM, retained placenta, metritis, keto-sis, DA, and pneumonia all reduced milk yield (P < 0.05)in primipara (Table 8). All of these diseases, exceptfor retained placenta, which can only occur right aftercalving, were associated with lower production (P <0.05) even before they were diagnosed. Retained pla-centa continued to have a negative effect (P < 0.0001) on


Table 7. Effects of the first 3 occurrences of generic clinical mastitis (CM) on milk yield in 6,699 parity 2+cows on 5 New York State dairy farms1



Before CM 1.7 1.4 2.1Same week −4.9 −5.2 −4.5 −5.6 −6.1 −5.1 −4.2 −5.0 −3.5+1 wk −5.9 −6.3 −5.6 −5.9 −6.4 −5.3 −4.8 −5.7 −3.9+2 wk −3.7 −4.1 −3.3 −3.6 −4.2 −3.0 −3.0 −4.0 −2.0+3 wk −2.5 −3.0 −2.1 −2.3 −3.0 −1.6 −2.1 −3.1 −1.0+4 wk −2.0 −2.4 −1.5 −1.5 −2.2 −0.8 −1.5 −2.5 −0.4+5 wk −1.5 −2.0 −1.0 −0.8 −1.5 0.0 −0.8 −1.9 0.3+6 wk −1.0 −1.5 −0.6 −0.4 −1.2 0.4 −0.6 −1.6 0.5+7 wk −0.7 −1.1 −0.2 0.0 −0.8 0.7 −0.1 −1.1 0.9+8 wk and after −0.7 −1.1 −0.2 0.3 −0.4 1.0 0.0 −0.8 0.8

1Estimates were obtained from a mixed model with an autoregressive covariance structure and randomherd effect. Clinical mastitis episodes were not allowed to have simultaneous effects in the same time period.Values are kg/d of milk.


production throughout the lactation period evaluated.Metritic and DA-diseased primipara produced less milk(P < 0.0001) than their herdmates for 5 wk after diagno-sis (and treatment). Pneumonia had a negative effect(P < 0.0001) on production until 7 to 8 wk after diagno-sis. Milk losses associated with ketosis continued forabout a month after diagnosis, at which time ketoticcows began to outproduce their nonketotic herdmates(P < 0.001).

Milk loss (P < 0.05) in multipara was associated withall diseases modeled (Table 9), in addition to that seenwith CM. Milk loss (P < 0.0001) occurred even beforediagnosis for metritic and DA-diseased multiparouscows. Losses associated with retained placenta, metri-tis, DA, and pneumonia continued for at least 7 to 8wk after diagnosis; they were especially large (P <0.0001) for DA, in which case the effect lasted through-out the lactation. Losses associated with ketosis were

Table 8. Effects of 5 other diseases on milk yield in 3,681 parity 1 cows on 5 New York State dairy farms1

Retained placenta Metritis Ketosis Displaced abomasum PneumoniaTime of milk

95% CI3 95% CI 95% CI 95% CI 95% CImeasurement2 Estimate Estimate Estimate Estimate Estimate

Before −3.0 −3.8 −2.2 −0.7 −1.5 0.0 −4.1 −4.9 −3.3 −1.4 −2.1 −0.6Same week −7.3 −8.1 −6.6 −3.2 −3.8 −2.6 −2.2 −3.0 −1.5 −11.1 −12.0 −10.1 −7.3 −8.2 −6.4+1 wk −3.7 −4.6 −2.9 −2.3 −3.0 −1.7 −1.4 −2.2 −0.6 −8.5 −9.6 −7.5 −5.5 −6.5 −4.5+2 wk −2.8 −3.6 −1.9 −2.2 −2.9 −1.6 −0.9 −1.7 0.0 −7.0 −8.1 −5.9 −4.4 −5.4 −3.3+3 wk −2.6 −3.4 −1.8 −1.7 −2.3 −1.0 −0.3 −1.1 0.6 −5.7 −6.8 −4.6 −4.3 −5.4 −3.2+4 wk −2.4 −3.2 −1.6 −1.2 −1.8 −0.5 0.4 −0.5 1.2 −4.3 −5.4 −3.2 −3.2 −4.3 −2.1+5 wk −1.9 −2.6 −1.1 −0.8 −1.5 −0.2 1.0 0.2 1.9 −2.5 −3.6 −1.4 −3.0 −4.1 −1.9+6 wk −1.5 −2.2 −0.8 −0.1 −0.7 0.5 1.4 0.6 2.2 −1.0 −2.0 0.1 −2.5 −3.6 −1.5+7 wk −1.2 −1.9 −0.6 0.1 −0.4 0.7 1.8 1.1 2.5 0.1 −0.9 1.0 −1.4 −2.4 −0.4+8 wk and after −0.6 −1.1 −0.1 0.5 0.1 0.9 1.8 1.2 2.4 0.7 −0.2 1.5 0.0 −0.9 0.8

1Estimates were obtained from a mixed model with an autoregressive (order 1) covariance structure and random herd effect. Values arekg/d of milk.

2Relative to disease recording.3CI = confidence interval.


evident until 5 to 6 wk after diagnosis (P < 0.0001).Milk fever-diseased cows produced less milk in the first2 wk after calving but greatly (P < 0.0001) outproducedtheir herdmates later in lactation.

DISCUSSION

Our main objective was to estimate the milk lossassociated with repeated occurrences of CM in high-producing dairy cows. The findings indicated that CM isfrequently a recurrent event, especially in multiparouscows. The same causative agent was involved in fewerthan half of the repeated CM cases. The milk loss associ-ated with repeated CM cases was slightly less severethan that caused by the first CM case.

We purposely chose to study large, high-producingdairy herds with low incidences of contagious mastitispathogens, because these are the farms that produce

BAR ET AL.4652

Tab

le9.

Eff

ects

of6

oth

erdi

seas

eson

mil

kyi

eld

in6,

699

pari

ty2+

cow

son

5N

ewY

ork

Sta

teda

iry

farm

s1

Mil

kfe

ver

Ret

ain

edpl

acen

taM

etri

tis

Ket

osis

Dis

plac

edab

omas

um

Pn

eum

onia

Tim

eof

mil

k95

%C

I395

%C

I95

%C

I95

%C

I95

%C

I95

%C

Im

easu

rem

ent2

Est

imat

eE

stim

ate

Est

imat

eE

stim

ate

Est

imat

eE

stim

ate

Bef

ore

−3.6

−4.7

−2.4

0.2

−0.5

0.8

−6.4

−7.2

−5.7

−0.1

−0.9

0.8

Sam

ew

eek

−2.1

−3.4

−0.8

−10.

6−1

1.2

−10.

0−6

.2−7

.0−5

.3−3

.3−3

.9−2

.6−1

5.5

−16.

3−1

4.6

−8.0

−9.0

−7.0

+1w

k−2

.0−3

.4−0

.5−7

.2−7

.8−6

.5−4

.5−5

.4−3

.6−3

.7−4

.3−3

.0−1

2.4

−13.

3−1

1.5

−6.6

−7.7

−5.5

+2w

k−2

.1−3

.5−0

.7−4

.8−5

.5−4

.2−4

.1−5

.0−3

.2−2

.5−3

.2−1

.8−1

0.5

−11.

4−9

.5−4

.8−6

.0−3

.6+3

wk

−0.5

−1.9

0.9

−3.5

−4.2

−2.9

−3.3

−4.2

−2.4

−1.9

−2.6

−1.2

−8.4

−9.3

−7.4

−4.9

−6.2

−3.6

+4w

k−0

.6−2

.00.

8−3

.0−3

.6−2

.4−2

.2−3

.2−1

.3−1

.4−2

.1−0

.6−6

.3−7

.3−5

.4−4

.3−5

.6−3

.0+5

wk

0.4

−1.0

1.7

−2.1

−2.7

−1.5

−1.7

−2.6

−0.8

−0.9

−1.6

−0.2

−5.2

−6.1

−4.2

−3.3

−4.6

−2.0

+6w

k1.

1−0

.12.

4−1

.3−1

.9−0

.8−1

.4−2

.2−0

.6−0

.5−1

.10.

2−4

.0−4

.9−3

.1−2

.4−3

.6−1

.1+7

wk

1.6

0.4

2.7

−0.6

−1.1

−0.1

−1.2

−1.9

−0.4

−0.2

−0.8

0.4

−3.0

−3.8

−2.2

−1.3

−2.4

−0.1

+8w

kan

daf

ter

1.6

0.7

2.5

−0.1

−0.5

0.3

−0.2

−0.8

0.4

−0.2

−0.7

0.3

−1.9

−2.6

−1.2

−0.5

−1.5

0.6

1 Est

imat

esw

ere

obta

ined

from

am

ixed

mod

elw

ith

anau

tore

gres

sive

(ord

er1)

cova

rian

cest

ruct

ure

and

ran

dom

her

def

fect

.V

alu

esar

ekg

/dof

mil

k.2 R

elat

ive

todi

seas

ere

cord

ing.

3 CI

=co

nfi

den

cein

terv

al.


most of the milk in industrialized countries. The inci-dence of CM in our sample herds, and the estimatedmilk loss associated with this disease, demonstrate thatdespite the success of control programs against conta-gious mastitis pathogens, CM remains a serious, eco-nomically limiting disease in the dairy industry. Clini-cal mastitis affects cows that generally have high milkproduction potential. This finding is in agreement withprevious studies (Seegers et al., 2003; Wilson et al.,2004). It also has a long-term effect on future milkproduction. In the current study, even one CM episodein the lactation was associated with a loss of approxi-mately 400 kg of milk in the next lactation. Houben etal. (1993) found this carryover effect only after 3 ormore CM episodes, but their study involved cows pro-ducing only 7,000 kg of milk per lactation and onlyapproximately 500 CM cases (vs. more than 3,000cases).

Several parameterization schemes for estimating re-peated CM cases were possible. We presented 2 andprefer the CM coding scheme in which the carryovereffect of previous CM case(s) was included in the currentCM (model [2]), because it results in a better model fitand is simpler to implement in economic models.

Taking into account the higher milk production ofcows before the first CM episode and the long-termmilk loss of the first case, the effect of subsequent CMepisodes on milk production was less severe than thatcaused by the first case. Several explanations for thiseffect are feasible. Bradley and Green (2001) postulatedthat agents with less pathogenicity are more likely in-volved in repeated cases. These more “udder-adaptive”agents, after a CM episode has occurred, are not com-pletely cleared from the udder. They may exist for sometime at undetectable levels. The levels of these patho-gens may then rise above the detection threshold andeven cause symptoms of CM, thus resulting in a re-peated CM case. Such an effect could be a result ofacquired immunity after a first CM case (Paape et al.,2002), leading to a less severe response (at least inmilk yield) in subsequent episodes. Another possibleexplanation for this decreasing effect is that the milkloss is relative and not absolute. This theory can explainwhy our estimates are higher than those found in lowerproducing cows (Rajala-Schultz et al., 1999). Trans-forming our estimates into percentages of actual milkyield, the estimates from both studies, and the effectof repeated cases, make our results more similar.

A third explanation is that observational studiesmade in commercial herds inevitably suffer from vari-ous selection and misclassification biases (Kleinbaumet al., 1982). A selection bias will be present becausecows suffering from high milk losses after CM are culledas a result of low production, so that the resulting CM


milk loss is actually biased toward the null. This isespecially so for the repeated CM cases. A misclassifi-cation bias is present if a CM cow is not diagnosed assuch. In this case we attributed the effect of undetectedCM cases to the long-term milk loss of the previous case.The exceptionally large data set needed to estimate theeffects of repeated CM episodes makes the wish to dosuch estimations in research herds (with a forced “nocull” policy) a prohibitively costly endeavor.

The milk losses obtained in this study are slightlyless for the first case of CM than those found in 2 otherNew York State dairy herds with comparable produc-tion levels (Wilson et al., 2004). One reason for thediscrepancy is that in the previous study, the effectof repeated cases was attributed to the first CM case.Repeated cases occurred in approximately 30% of cowswith a first CM case; hence, this more correct productionloss accounting is quite substantial.

Although we included other production diseases onlyas potential confounders in our models, the fact thatwe had daily milk yield records, the size of the studydata set, and the statistical procedure used make theestimated milk losses associated with these diseases avaluable contribution in evaluating their economic im-portance.

Because estimations of the effects of repeated CMhave rarely been addressed in previous studies, thecurrent study on the milk losses associated with re-peated CM episodes is an important step forward inhelping dairies assess the profitability of individualcows as they progress through lactation and overallherd life.

ACKNOWLEDGMENTS

The USDA (Cooperative State Research, Education,and Extension Service) Award No. 2005-35204-15714provided the funding for this study. The authors wantto thank the owners and personnel of the 5 dairies,


and the personnel of the Ithaca and Canton RegionalLaboratories, Quality Milk Production Services, fortheir valuable cooperation during the study.

REFERENCES

Barkema, H. W., Y. H. Schukken, T. J. Lam, M. L. Beiboer, H. Wil-mink, G. Benedictus, and A. Brand. 1998. Incidence of clinicalmastitis in dairy herds grouped in three categories by bulk milksomatic cell counts. J. Dairy Sci. 81:411–419.

Bradley, A. J., and M. J. Green. 2001. Adaptation of Escherichia colito the bovine mammary gland. J. Clin. Microbiol. 39:1845–1849.

Dopfer, D., H. W. Barkema, T. J. Lam, Y. H. Schukken, and W.Gaastra. 1999. Recurrent clinical mastitis caused by Escherichiacoli in dairy cows. J. Dairy Sci. 82:80–85.

Grohn, Y. T., J. J. McDermott, Y. H. Schukken, J. A. Hertl, andS. W. Eicker. 1999. Analysis of correlated continuous repeatedobservations: Modelling the effect of ketosis on milk yield in dairycows. Prev. Vet. Med. 39:137–153.

Grohn, Y. T., D. J. Wilson, R. N. Gonzalez, J. A. Hertl, H. Schulte,G. Bennett, and Y. H. Schukken. 2004. Effect of pathogen-specificclinical mastitis on milk yield in dairy cows. J. Dairy Sci.87:3358–3374.

Houben, E. H. P., A. A. Dijkhuizen, J. A. M. van Arendonk, and R.Huirne. 1993. Short- and long-term production losses and repeat-ability of clinical mastitis in dairy cattle. J. Dairy Sci.76:2561–2578.

Kleinbaum, D. G., L. L. Kupper, and H. Morgenstern. 1982. Epidemio-logic Research. John Wiley and Sons, New York, NY.

Paape, M., J. Mehrzad, X. Zhao, J. Detilleux, and C. Burvenich. 2002.Defense of the bovine mammary gland by polymorphonuclearneutrophil leukocytes. J. Mammary Gland Biol. Neoplasia7:109–121.

Rajala-Schultz, P. J., Y. T. Grohn, C. E. McCulloch, and C. L. Guard.1999. Effects of clinical mastitis on milk yield in dairy cows. J.Dairy Sci. 82:1213–1220.

SAS Institute. 2006. SAS OnlineDoc 9.1.3. SAS Institute Inc.,Cary, NC.

Seegers, H., C. Fourichon, and F. Beaudeau. 2003. Production effectsrelated to mastitis and mastitis economics in dairy cattle herds.Vet. Res. 34:475–491.

Sviland, S., and S. Waage. 2002. Clinical bovine mastitis in Norway.Prev. Vet. Med. 54:65–78.

Wilson, D. J., R. N. Gonzalez, J. A. Hertl, H. Schulte, G. Bennett, Y.Schukken, and Y. Grohn. 2004. Effect of clinical mastitis on thelactation curve: A mixed model estimation using daily milkweights. J. Dairy Sci. 87:2073–2084.

Zadoks, R. N., H. G. Allore, H. W. Barkema, O. C. Sampimon, G. J.Wellenberg, Y. T. Grohn, and Y. H. Schukken. 2001. Cow andquarter-level risk factors for Streptococcus uberis and Staphylo-coccus aureus mastitis. J. Dairy Sci. 84:2649–2663.

effect of repeated episodes of generic clinical mastitis on milk yield in dairy cows

Documents