Monitoring and testing dairy herds for metabolic disease

Vet Clin Food Anim 20 (2004) 651–674

Monitoring and testing dairy herds for metabolic disease Garrett R. Oetzel, DVM, MS Department of Medical Sciences, School of Veterinary Medicine, University of Wisconsin–Madison, 2015 Linden Drive, Madison, WI 53706, USA

Metabolic disease incidence typically increases as milk production increases and as herds become larger. These factors favor the use of rigorous, quantitative monitoring of metabolic disease whenever possible. Fortunately, recent developments of herd-based tests plus new applications of some old tests are now available for use in routine herd monitoring and for investigating dairy herds with metabolic disease problems. This allows the herd consultant to make recommendations based on objective data rather than subjective impressions alone. This article will focus on strategies for testing and monitoring subacute ruminal acidosis (SARA), subclinical ketosis (SCK), and parturient hypocalcemia (clinical plus subclinical milk fever) in dairy herds. Quantitative data about these diseases have been published, and are the foundation for their use and interpretation on a herd basis. Additionally, these three disorders are gateway conditions for other metabolic disorders such as laminitis, displaced abomasum, impaired immune function, retained placenta, and cystic ovarian disease. Other metabolic diseases can be important problems in dairies (eg, hypomagnesemia, udder edema, hypokalemia, and so on), but these are less common disorders with limited published data. Biological and statistical basis for herd testing Interpreting test results for groups versus individual cows The interpretation of herd-based tests for metabolic diseases is very different than interpreting laboratory results for metabolites from individual cows. Test results from individual cows are interpreted by comparing the laboratory result to a normal range established by the laboratory that did

E-mail address: [email protected] (G.R. Oetzel). 0749-0720/04/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.cvfa.2004.06.006

652

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

the testing. Normal ranges are often derived by calculating a 95% confidence interval (or a similar statistic) of test results from 100 or more clinically normal animals. This approach is useful for making decisions about individual sick cows, but is not useful for interpreting test results when a subsample of the herd is tested and decisions will be made on a herd basis. Interpretation of herd-based test results requires an understanding of how metabolites affect cow performance (regardless of whether they are within the normal range or not), a statistically based approach to determining subsample sizes, and an emphasis on monitoring subclinical disease prevalence instead of clinical disease incidence. Interpreting herd proportions versus herd means Herd test results for metabolic diseases can be interpreted as either the mean test result of the subgroup sampled or as the proportion of animals above or below a certain cut point within the subsample. The biology of the metabolic disease tested determines which interpretive approach is the most appropriate. If a metabolite is associated with disease when it is either above or below a biologic threshold (cut point), then it should be evaluated as a proportional outcome. For example, ruminal pH
5.5 puts cows at risk for SARA, with subsequent rumenitis and other complications [1]. Ruminal pH values above 5.5 do not put cows at risk for SARA. Therefore, there is little value in interpreting mean ruminal pH results from a subsample of cows within a dairy herd. Instead, interpret the proportion of cows with ruminal pH below the cut point. SCK in dairy herds can be monitored by testing for b-hydroxybutyrate (BHB) or other ketone bodies in blood or milk. SCK is also a threshold disease, and cows are affected only when ketone concentrations are elevated. Lowering ketones below a threshold concentration is of little to no biologic significance to the cow. Therefore, herd-based BHB test results are interpreted on a proportional basis, and the mean concentration for the group of cows tested is not as useful as an outcome. Blood BHB concentration above 1400 lmol/L (14.4 mg/dL) is the most commonly used cut point for SCK. Early lactation cows with BHB concentrations above this cut point are at threefold greater risk to develop either clinical ketosis or displaced abomasum [2]. This cut point is considerably higher than the upper end of the typical laboratory normal reference range for individual cows. Nonesterified fatty acids (NEFA) concentrations in blood are an indicator of negative energy balance in prefresh cows. Elevated NEFA before calving are associated with increased risk for displaced abomasum after calving [3]. Low NEFA concentrations are not biologically important. A threshold above 0.400 mEq/L for cows between 2 and 14 days of actual calving has been suggested as the appropriate cut point. Again, we are not interested in the mean NEFA value from a group of prefresh cows, but rather in the proportion of cows above the cut point.

653

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

The incidence of parturient hypocalcemia (clinical plus subclinical milk fever) in a dairy herd can be evaluated by measuring serum calcium concentration within 12 to 24 hours after calving. Cut points of less than 2.0 mmol/L (8.0 mg/dL) total serum calcium [4] or less than 1.0 mmol/L (4.0 mg/dL) ionized calcium [5] have been used to define parturient hypocalcemia. Results are interpreted as the proportion of cows below the cut points. Besides defining the appropriate cut points for tests evaluated as proportion outcomes, it is also necessary to determine the alarm level for the proportion of animals above (or below) the described cut point. Because of normal biologic variation, a few individual cows are expected to be above (or below) the biologic threshold in any dairy. The alarm level is established from research results or clinical experience with these tests in herd settings. Suggested cut points and alarm levels for ruminal pH, BHB, and NEFA test results are listed in Table 1. Urinary pH in prefresh cows fed anionic salts is a useful test for herds with parturient hypocalcemia problems despite feeding supplemental anions before calving. The biologic threshold for this test is not one sided. Rather, there is an ‘‘optimal,’’ mid-range for urinary pH of about 6.0 to 7.0. Urinary pH values above or below this optimal range have adverse consequences. Therefore, an evaluation of the mean urinary pH in a group of prefresh cows is the most appropriate interpretation. Appropriate sample sizes for herd-based tests Herd-based testing is useful only when sufficient numbers of cows within the herd are tested. This gives reasonable confidence that the results (either a proportion or a mean) truly represent the entire population of eligible cows within the herd. Clinicians generally do not need to sample as many cows as a researcher would sample to achieve a 95% confidence (P \ .05) in the results. Rather, a 75% confidence interval is acceptable and more practical under most herd testing conditions. Lower confidence intervals are inherent in clinical decision making. When solving herd metabolic disease problems, there is no option to avoid making

Table 1 Cut points and alarm levels for herd-based metabolic disease tests evaluated as proportions Test

Cut point

Alarm level proportion

Associated risk

Ruminal pH


5.5

>25%

BHB NEFA

1400 lmol/L 0.400 mEq/L

>10% >10%

Subacute ruminal acidosis Subclinical ketosis Prepartum negative energy balance, fatty liver

Abbreviations: BHB, blood b-hydroxybutyrate; NEFA, plasma nonesterified fatty acids.

654

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

a decision. Leaving herd management unchanged is an active decision—just as active as deciding to implement change. Given these factors, using a 75% confidence interval to determine minimum sample size is a reasonable compromise between statistical confidence, test costs, and practicality. Clinicians must constantly make decisions based on limited information, and in particular, must make herd-based decisions based on relatively small sample sizes. Researchers, on the other hand, can make ‘‘no decision’’ by concluding that the effect being studied was insignificant (P > 0.05). Thus, larger sample sizes are required to protect the researcher’s conclusion from random error. The impact of error in the conclusions drawn by a researcher is great. An erroneous conclusion could become dogma and a building block in the scientific process. The price is great when dogma is eventually refuted and scientific progress has to be torn down and rebuilt from new assumptions. In contrast, clinicians are not expected to make correct decisions even 95% of the time, and the penalty for making erroneous clinical decisions is typically not as great as it is for drawing erroneous research conclusions. Lower confidence intervals are also appropriate for clinical decisions because clinical decisions are made with more than just the herd-based testing data. Other clinical data (estimates of disease incidence rates, herd removal rates, milk production, milk components, and so on) are available from the herd. Final clinical decisions are made by subjectively combining these data with the herd testing results. The minimum sample size for herd-based tests with proportional outcomes is 12 cows. This minimum sample size gives reasonable confidence (75% or more) that the classification of the test results from the 12 cows sampled will correctly represent the true classification for the entire group. An example interpretation guide for ruminal pH testing results based on a sample size of 12 cows is presented in Fig. 1. The ability of this testing strategy to correctly categorize a group of cows within a herd for SARA is given in Fig. 2. Different statistical tests (taking into account variability around the mean) are applied to tests interpreted as means. For example, a minimum sample size of about eight cows is appropriate for urinary pH testing. Larger sample sizes are required when evaluating tests with proportional outcomes compared with mean outcomes. Sampling additional cows (beyond the initial minimum of 12 cows) is suggested when the results of a proportional outcome are very close to the alarm level. For example, a test result of 2/12 (16.7%) cows with ruminal pH \5.5 is below the alarm level of 25%. However, the alarm level is included within the confidence interval for the test result and the results are considered borderline (Fig. 1). Additional testing could be warranted, unless other clinical information from the herd strongly supports a negative classification for SARA. In contrast, if 5 of 12 (41.7%) of the cows tested had ruminal pH \5.5, then additional testing would not be warranted as long as other clinical data from the herd corroborate a herd diagnosis of SARA.

655

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

Number Positive Results (12 Cows Sampled)

7 6

Positive Herd Results (>4/12)

Alarm Level (25%)

5 4 Borderline Herd Results (2/12, 3/12 or 4/12)

3 2 1

Negative Herd Results (0/12 or 1/12)

0

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

Proportion Positive Fig. 1. Interpretation of ruminal pH test results using 75% confidence intervals and an alarm level of 25% for test results from 12 cows sampled from within a group 100 cows.

Cows to be sampled for these tests need to come from the appropriate ‘‘eligible’’ or ‘‘at risk’’ group. It is of no clinical value to test cows for a condition that they have little risk because of their current parity or stage of lactation. Appropriate eligible groups for herd-based tests for metabolic diseases are listed in Table 2. The size of the eligible group for testing has some, but limited influence on the appropriate sample size. More cows may be available for testing in larger herds, yet there is little statistical value in testing more animals just because the group size is larger. Statistical evaluation of the testing strategy shows that even when group size is large, the same sample size yields almost the exact same information about the group. In smaller herds, it may be possible to test the entire eligible group and still not have adequate sample size. For example, only the prefresh cows within about 3 weeks of expected calving date are eligible for urinary pH or NEFA testing. If there are only four cows in this category, then all four should be tested. However, a sample size of four cows is probably too small to be conclusive. So, additional cows should be tested as they move into the eligible group, and the group results interpreted only after about 8 (for urinary pH) or 12 (for NEFA) test results have been accumulated. If cows are repeatedly tested for NEFA or urinary pH as they approach calving, only the last test result before actual calving for that cow should be interpreted. Do not use multiple test results from the same cow to achieve minimum sample size goals.

656

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

Probability of Herd Classification (Negative, Borderline, or Positive)

1.00

0.80

0.60

0.40

0.20

0.00 5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

True Prevalence of Subacute Ruminal Acidosis Fig. 2. Probabilities of classifying a herd negative (triangle), borderline (circle), or positive (square) for subacute ruminal acidosis using a cut point of ruminal pH
5.5. The sample size was 12 cows from a group of 100, the alarm level was 25%, and the confidence interval was 75%. For example, a group with a 40% true prevalence of subacute ruminal acidosis would have a 1% chance of being classifed as negative, a 42% chance of being classifed borderline, and a 57% chance of being classified as positive for subacute ruminal acidosis.

Sample sizes larger than the recommended minimums are appropriate when a herd’s clinical problems are small or supporting evidence for a presumptive clinical diagnosis is weak. In this case, the clinician needs more than 75% confidence that the herd classification from the testing results correctly represents the entire group. This is practically easier in large herds, because there are more cows available to test and because the cost of testing is diluted across a larger number of cows. Additionally, the economic cost of a bad clinical decision based on small sample size may be greater in large herds. Table 2 Appropriate groups of cows eligible for different herd-based tests for metabolic diseases Test

Eligible group

Ruminal pH

Lactating cows, about 5 to 150 days in milk (focus testing on cows 5 to 50 days in milk in component-fed herds and cows 50 to 150 in milk in herds feeding a total mixed ration) Lactating cows, about 5 to 50 days in milk Pre-fresh cows, ideally 2 to 14 days from actual calving Prefresh cows that have been on an anionic diet for [24 hours

BHB NEFA Urinary pH

Abbreviations: BHB, blood b-hydroxybutyrate; NEFA, plasma nonesterified fatty acids.

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

657

Sources of error in herd-based testing Biologic tests can be very useful in supporting other clinical evidence of a metabolic disease problem on a dairy. Veterinarians have tremendous experience in collecting, analyzing, and interpreting the results of biologic tests. However, biologic test results do not stand alone in making herd-based decisions. Biologic test results are subject to errors from inadequate sample size, improper sample handling, inappropriate time of sample collection relative to feeding, and laboratory error. Thus, biologic test results should be supported by other herd data. For example, a finding of a high proportion of cows with low ruminal pH collected by rumenocentesis is corroborated by findings of low-fiber diets being consumed by the cows, thin cows in the face of high-energy diets, a high prevalence of laminitis-related lameness, or milk fat test depression. Without supporting evidence, however, the finding of low ruminal pH alone is very suspect and likely is in error (perhaps due to analytical problems in measuring pH of the ruminal fluid). Specific details for quantitative evaluation of key metabolic diseases are described in the following sections. These specific recommendations are based on the principles described above.

Subacute ruminal acidosis SARA is diagnosed and prevented on a herd basis rather than on an individual cow basis [6]. Clinical signs in dairy herds affected with SARA may include low or fluctuating dry matter intakes, low body condition scores, diarrhea, nosebleeds, unexplained deaths due to chronic inflammatory diseases, unexplained high cull rates due to vague health problems, milk fat depression, and poor milk production in the second and greater lactation cows relative to the first lactation cows. None of these signs by themselves are diagnostic for SARA; however, considered together, they form the basis for a presumptive diagnosis of SARA in a herd. It is extremely useful to support a presumptive diagnosis of SARA in a herd with quantitative ruminal pH data. Testing herds for ruminal pH Ruminal pH is the definitive test for ruminal acidosis. Ruminal pH below about 5.5 for prolonged time periods is the root cause of the clinical signs observed in herds with SARA problems [1]. Evaluation of ruminal pH is challenging because of the difficulty in obtaining a sample for testing and because ruminal pH varies from day to day within herds and time of day within cow (Fig. 3). The methodology for collecting ruminal pH samples has been described in detail [7,8]. A potential source of error in ruminal pH measurements is the accuracy and correct calibration of the pH meter. A high-quality pH meter is

658

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

7.5 Meal #1

7.0

Ruminal pH

Meal #2

6.5

6.0

5.5

5.0 6:30

8:30 10:30 12:30 14:30 16:30 18:30 20:30 22:30 0:30

2:30

4:30

6:30

Time of Day Fig. 3. Example of the effect of time after feeding on ruminal pH for a cow fed a total mixed ration twice daily. [Data from Oetzel GR, unpublished data, 2004.]

recommended—pH paper is not accurate enough, and is influenced by the green color of the ruminal fluid. Keep in mind that pH meters do not work well when operated at cold temperatures. Bring all the ruminal fluid samples (in capped syringes with the air excluded) into a warm parlor or office to run the pH determinations during cold weather. Also, pH electrodes may become dry between uses. Soaking the electrode in a buffer solution before calibration improves accuracy. It is prudent to calibrate the meter twice (or more) before testing your samples. After the last calibration, put the pH 7 and pH 4 buffers back on the meter to verify that they read the correct pH. Necessary sample sizes for herd-based ruminal pH evaluation have been described in detail. [1]. A practical and statistically reasonable sample size for most herds is 12 animals per diet. If 5 or more of the 12 cows tested have a ruminal pH
5.5 (Fig. 1), then the group is considered to be at high risk for SARA, and if other clinical signs corroborate the diagnosis of SARA, the diet should be modified to reduce the risk for SARA. This testing scheme works very well for herds with high (>30%) or low (\15%) prevalences of cows with low ruminal pH (Fig. 2). It does not have the necessary statistical power to be used as a means of ‘‘fine-tuning’’ diets for optimal ruminal pH. This would require much larger sample sizes and more frequent testing than would be practical. Herds with intermediate prevalences (16.7%–33.3%) of low ruminal pH may require additional testing, or may be classified as either at high or low

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

659

risk for SARA based on additional herd information. Immediate dietary intervention is probably not critical in herds with intermediate prevalences, so it is not unreasonable to take some additional time to gather more herd information. Ruminal pH sampling should be done around the time of the expected lowest point (nadir), because its purpose is to identify cows with low ruminal pH. In component-fed herds, the nadir in ruminal pH occurs about 2 to 4 hours postfeeding. In total mixed reaction-fed herds, the nadir in ruminal pH occurs about 6 to 8 hours postfeeding (or later). The Food Animal Production Medicine Section at the University of Wisconsin–Madison has used ruminal pH testing to help classify herds for SARA for over 8 years. Of the 737 cows we have tested, 20% had ruminal pH below 5.5, and 23% of the 57 herds we evaluated were classified as being at high risk for SARA.

Subclinical ketosis It is difficult to subjectively assess the degree of SCK problems that a herd may be experiencing. Clinical ketosis rates (as determined by dairy producers) are of extremely limited value at all in assessing the true ketosis status of a herd. Herds vary dramatically in their definition of clinical ketosis and in their ability to detect clinical signs in early lactation cows. Producers in smaller herds tend to overestimate the incidence of clinical ketosis [9], and (based on my own clinical observations) producers in larger herds tend to underestimate the incidence of clinical ketosis. Therefore, it is essential to make clinical decisions based on the measured prevalence of SCK in a herd instead of attempting to rely on the dairy producer’s perception of the incidence of clinical ketosis. Herds with SCK problems in early lactation cows also tend to have increased incidence of displaced abomasum and increased herd removals in the first 60 days in milk. Affected herds may also have a higher proportion (>40%) of cows with milk fat to true protein percentages below 0.70 at first test after calving [10]. However, none of these clinical findings are definitive evidence for a SCK problem in a herd. A quantitative evaluation of the prevalence of SCK is extremely useful in most dairy herds. The incidence of SCK in a herd (which requires repeated measures of blood BHB in early lactation) is impractical to determine in a commercial dairy setting. Fortunately, the incidence of SCK may be inferred from its prevalence. When both SCK incidence (blood BHB tests at 1, 2, 3, and 6 weeks after calving) and average early lactation SCK prevalences were measured in a large field study [11], the cumulative incidence of SCK (45%) was about 2.2 times the average prevalence of SCK (20%). In the same study, the reported incidence of clinical ketosis was only 1.5% (this is

660

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

probably an example of underrepresenting of the clinical ketosis case rate, given that a SCK prevalence of 20% is relatively high). Strategy for evaluating herds for subclinical ketosis The ‘‘gold standard’’ test for SCK is blood BHB. This ketone body is more stable in blood than acetone or acetoacetate [12]. The mostly commonly used cut point for SCK is 1400 lmol/L (14.4 mg/dL) of blood BHB. Early lactation cows with blood BHB concentrations above this cut point are at threefold greater risk to develop displaced abomasum or clinical ketosis, and cows with blood BHB concentrations above 2000 lmol/L are at risk for reduced milk yield [2]. Some studies use a slightly lower cut point (1200 lmol/ L) of blood BHB for defining SCK. The exact cut point chosen usually has a minor effect on the interpretation of herd-based results. Clinical ketosis generally involves much higher levels of BHB (3000 lmol/L or more). The alarm level for the proportion of cows above the cut point of 1400 lmol/L of blood BHB has not been well defined. Published research studies show an average SCK prevalence of about 15% (Tables 3–5). Based on this information, I suggest using 10% as the alarm level for herd-based SCK testing. An example interpretation guide for BHB testing based on this alarm level is presented in Fig. 4. My own clinical experience in testing problem herds for clinical ketosis suggests that the 10% alarm level is achievable and appropriate. In the last 7 years I have evaluated the SCK status of 766 cows in 56 herds as part of my regular clinical service to dairy producers. This is a convenience sampling of herds (not a random sample), and many of these herds had already been identified as potential ketosis problem herds based on clinical signs. The overall prevalence of SCK in these herds was 15.0%, and 34% of the herds evaluated had SCK prevalence below 10% (suggesting that \10% is an achievable goal). Thirty-seven of the herds I have screened for SCK had sufficient sample size to allow for categorization according to the scheme presented in Fig. 4. Eight herds (21%) were classified as negative for SCK (ie, zero cows with SCK of 12 or more cows tested), 14 herds were classified as borderline, and 15 herds were classified as positive for SCK (three or more cows with SCK of 12 or more cows tested). Each of the 15 positive herds had other clinical evidence to corroborate the SCK diagnosis. About 3.7% of the cows I have tested for SCK had very high blood BHB concentrations (above 3000 lmol/L). Most of these cows should have been identified as clinically ketotic but were not (I only test cows for SCK that have not been previously identified as sick). Four herds (of 37 herds with sufficient sample size to be categorized) had more than one cow in the very high BHB category. All four herds were already classified as SCK problem herds. These results suggest that they also had an apparent problem of inadequate disease recognition in early lactation cows. This is a particular

Test type/study Acetest tablet Nielen et al, 1994 Keto Test Osborne et al, 2002 Ketostix, trace (5 lmol/L) Carrier et al, 2003 Oetzel, 2004 Ketostix, small (15 lmol/L) Carrier et al, 2003 Oetzel, 2004 Ketostix, moderate (40 lmol/L) Carrier et al, 2003 Oetzel, 2004

Herds tested

% SCK

Total samples

True positives

False negatives

False positives

18

11.3%

124

14

0

45

1

18.2%

159

28

1

1 6

7.0% 12.0%

741 83

47 9

1 6

7.0% 12.0%

741 83

1 6

7.0% 12.0%

741 83

True negatives

Sensitivity

Specificity

65

100%

59%

52

78

97%

60%

5 1

101 18

588 55

90% 90%

85% 75%

41 8

11 2

31 6

658 67

79% 80%

96% 92%

26 7

26 3

7 2

682 71

50% 70%

99% 97%

Abbreviations: SCK, subclinical ketosis, defined by blood b-hydroxybutyrate 1400 lmol/L. Data from Refs. [18–20] and Oetzel GR, unpublished data from clinical herd investigations at the University of Wisconsin–Madison; 2004.

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

Table 3 Sensitivity and specificity of urine cowside tests compared to blood b-hydroxybutyrate (cut point of 1400 lmol/L)

661

662

Test type/study Utrecht powder: Nielen et al, 1994 Geishauser et al, 1998 KetoCheck powder (trace): Geishauser et al, 1998 Carrier et al, 2003 Bioketone powder (trace): Geishauser et al, 1998

Blood BHB cut point

Herds tested

% SCK

Total samples

True positives

False negatives

False positives

True negatives

Sensitivity

Specificity

1400 1200

18 25

10.3% 16.4%

185 529

17 37

2 50

7 0

159 442

89% 43%

96% 100%

1200 1400

25 1

16.4% 7.5%

529 878

24 28

63 38

0 9

442 803

28% 42%

100% 99%

1200

25

16.4%

529

24

63

0

442

28%

100%

Abbreviations: BHB, b-hydroxybutyrate, lmol/L; SCK, subclinical ketosis, defined by blood b-hydroxybutyrate 1200 or 1400 lmol/L. Data from Geishauser T, Leslie K, Kelton D, Duffield T. Evaluation of five cowside tests for use with milk to detect subclinical ketosis in dairy cows. J Dairy Sci 1998;81(2):438–43 and Refs. [19,20].

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

Table 4 Sensitivity and specificity of cowside milk nitroprusside powders compared to blood b-hydroxybutyrate (cut point of 1200 or 1400 lmol/L)

Table 5 Sensitivity and specificity of a cowside milk b-hydroxybutyrate strip compared to blood b-hydroxybutyrate (cut point of 1400 lmol/L) Test type/study

% SCK

Total samples

True positives

21 1 17 39

11.9% 7.6% 17.2% 10.2%

469 883 221 1573

51 59 34 144

8 21 1 1 17 53

8.4% 11.9% 16.5% 7.6% 17.2% 12.6%

190 469 248 883 221 2246

8 21 5 1 17 52

8.4% 11.9% 27.2% 7.6% 17.2% 12.1%

190 469 235 883 221 1998

False negatives

False positives

True negatives

Sensitivity

Specificity

5 8 4 17

182 100 36 318

231 716 147 1094

91% 88% 89% 89%

56% 88% 80% 77%

14 45 39 50 33 233

2 11 2 17 5 49

31 99 65 54 32 345

143 314 142 762 151 1619

88% 80% 95% 75% 87% 83%

82% 76% 69% 93% 83% 82%

12 33 47 20 17 129

4 23 17 47 21 112

14 42 29 10 5 100

160 371 142 806 178 1657

75% 59% 73% 30% 45% 54%

92% 90% 83% 99% 97% 94%

663

Abbreviations: BHB, b-hydroxybutyrate; SCK, subclinical ketosis, defined by blood b-hydroxybutyrate 1400 lmol/L. Data from Geishauser T, Leslie K, Ten Hag J, Bashiri A. Evaluation of eight cow-side ketone tests in milk for detection of subclinical ketosis in dairy cows. J Dairy Sci 2000;83(2):296–9; Oetzel GR, unpublished data from clinical herd investigations at the University of Wisconsin–Madison, 2004; Jorritsma R, Baldee SJC, Schukken YH, Wensing T, Wentink GH. Evaluation of a milk test for detection of subclinical ketosis. Vet Quart 1998;20(3):108–10; Duffield TH, LeBlanc S, Bagg R, Leslie K, Ten Hag J, Dick P. Effect of a monensin controlled release capsule on metabolic parameters in transition dairy cows. J Dairy Sci 2003;86(4):1171–6; and Refs. [18,20].

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

Milk BHB strip (50 lmol/L) Geishauser et al, 2000 Carrier et al, 2003 Oetzel, 2004 Pooled data (by cow) Milk BHB strip (100 lmol/L) Jorritsma et al, 1998 Geishauser et al, 2000 Osborne et al, 2002 Carrier et al, 2003 Oetzel, 2004 Pooled data (by cow) Milk BHB strip (200 lmol/L) Jorritsma et al, 1998 Geishauser et al, 2000 Duffield et al, 2003 Carrier et al, 2003 Oetzel, 2004 Pooled data (by cow)

Herds tested

664

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

Number Positive Results (12 Cows Sampled)

6 5 4

Positive Herd Results (>2/12)

Alarm Level (10%)

3 2 Borderline Herd Results (1/12 or 2/12)

1 0

Negative Herd Result (0/12)

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

Positive Cows in Herd (%) Fig. 4. Interpretation of blood b-hydroxybutyrate test results using 75% confidence intervals and an alarm level of 10% for test results from 12 cows sampled from within a group 50 cows.

concern in larger dairies with group feeding of early lactation cows. It is difficult but important for group-fed herds to develop protocols to monitor feed intake or measures of ketosis in early lactation cows. Prompt identification of these cows allows for individual cow treatment with glucose precursors, and alerts the producer to underlying nutritional problems. The SCK testing strategy described here is designed to identify herds with either very high or very low prevalence of SCK (Fig. 5). It is not intended to ‘‘fine tune’’ or optimize a transition cow feeding program for SCK prevention. An evaluation of early lactation cows for SCK requires testing most or all of the eligible cows in small to medium-sized herds. In larger herds, a suitable sample size may be obtained on a single herd visit. Results of blood BHB testing can also be useful for discerning the underlying cause(s) of the ketosis problem. SCK caused by preexisting fatty liver late in the prefresh period (type II ketosis [13]) tends to cause elevated BHB concentrations in the first 5 to 15 days in milk. These cows often have other manifestations of fatty liver, including immune suppression and lack of response to ketosis treatment. Cows with type I ketosis typically do not become ketotic before about 3 to 6 weeks in milk [14]. However, cows with type II ketosis may have persistent SCK that lingers into the third week or more of lactation, so it is difficult to discern which type of ketosis underlies the SCK in these cows. A larger sample size may be required to clearly evaluate the days in milk of cows with SCK. Examples of herd test results typical of either type II or type I ketosis are presented in Table 6.

665

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

Probability of Herd Classification (Negative, Borderline, or Positive)

1.00

0.80

0.60

0.40

0.20

0.00 0%

5%

10%

15%

20%

25%

30%

35%

40%

True Prevalence of Subclinical Ketosis Fig. 5. Probabilities of classifying a herd negative (triangle), borderline (circle), or positive (square) for subclinical ketosis using a cut point of blood b-hydroxybutyrate (BHB) 1400 lmol/L. The sample size was 12 cows from a group of 50, the alarm level was 10%, and the confidence interval was 75%. For example, a group with a 25% true prevalence of subclinical ketosis would have a 2% chance of being classifed as negative, a 38% chance of being classifed borderline, and a 60% chance of being classified as positive for subclinical ketosis.

Blood b-hydroxybutyrate sampling requirements The BHB test can be performed on serum samples, and there are no special sample handling requirements. However, blood samples for BHB should not be collected from the mammary vein. Mammary vein blood is lower in BHB because the udder tends to extract BHB but releases acetoacetate [15]. Blood BHB concentrations typically increase after feeding [16,17]. Consistent sampling at 4 to 5 hours after the start of feeding has been suggested to capture peak BHB concentrations [16]. The postfeeding peak in serum BHB concentrations is likely due to ruminal production of butyric acid. Excess amounts of butyric acid (either from ruminal production or from silage) are easily converted to BHB in the wall of the rumen. Overview of cowside tests for subclinical ketosis A variety of cowside tests are available for SCK monitoring of dairy herds. However, none of the cowside tests have perfect sensitivity and specificity compared with blood BHB (see Tables 3–5). Therefore, the gold standard SCK test (blood BHB) is the most accurate for herd monitoring, and is particularly warranted for investigating herds with presumptive SCK.

666

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

Table 6 Examples of dairy herds with days in milk of cows with high blood b-hydroxybutyrate concentrations (cut point of 1400 lmol/L) suggestive of either Type II or Type I subclinical ketosis Type II Ketosis example herd: Cow 1902 2910 4176 3503 5293 6576 6624 3548 3553 4019 1709 4081 6662 6476 6681 4040 3579 4109

Days in milk

BHB (lmol/L)

5 5 8 9 9 10 11 13 13 13 14 14 14 19 23 26 31 32 Group Summary:

1410 2490 4200 670 1730 1930 1100 1070 700 400 2530 650 1580 1310 1820 920 800 660 8/18

Type I ketosis example herd: Cow Sheila Susan Lynn Sparkle Swish Dimples Gracie Marcy Diane Tootsie Sasha Olive Merry Kristyn Morgan April Twinkle Sarah

Days in milk

BHB (lmol/L)

5 5 6 6 7 8 10 10 11 11 15 19 20 24 25 31 35 38 Group Summary:

740 870 450 770 950 530 390 1190 820 830 970 2260 1710 1520 2970 2220 450 2230 6/18

Abbreviations: BHB, Blood b-hydroxybutyrate (values over the 1400 lmol/L cut point are bolded). Data from Oetzel, GR. Unpublished data from clinical herd investigations at the University of Wisconsin–Madison, 2004.

Cowside ketosis tests have the advantages of lower cost, less labor, and immediate results when compared with blood BHB testing. This makes them particularly useful for making (or excluding) a clinical diagnosis of ketosis in individual, sick cows. However, testing herds for SCK requires a very different testing strategy compared with diagnostic decision making for sick cows. Cowside urine tests for subclinical ketosis Urine can be evaluated for cowside ketosis testing; however, it is much more difficult to collect a urine sample than a cowside milk sample. Even with considerable effort, some cows inevitably fail to urinate within a reasonable time period and cannot be tested at all. In research trials, urine samples are not usually collected from 100% of eligible cows. An example is a recent study in which urine samples were successfully collected from only 64% of eligible cows [18]. This is a substantial practical limitation on farms and greatly increases labor costs for testing.

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

667

Urine acetoacetate can be evaluated quantitatively by nitroprusside tablets (Acetest; Bayer Corp. Diagnostics Division, Elkhart, Indiana). This test has excellent sensitivity but poor specificity (Table 3) [19]. This makes it a useful test for evaluating individual sick cows (for whom a false positive result is preferred to a false negative one), but not very useful for herd-based monitoring. A dipstick designed for evaluating milk BHB has been evaluated for use with urine [18], despite lacking a label for use with urine. As for the urine tablets, this test has good sensitivity but poor specificity (Table 3). The higher cost of these strips compared with other urine ketone tests makes them impractical for use on urine, although they are an excellent cowside test for milk BHB, as described later. The best test for cowside urine ketone evaluation is a semiquantitative dipstick (Ketostix; Bayer Corp. Diagnostics Division, Elkhart, Indiana) that measures acetoacetate. Urine ketone tests, on the whole, have a reputation for very poor specificity; however, recent data suggest that poor specificity may not be a problem with the Ketostix. The urine dipstick had very good specificity (and sensitivity) compared with the blood BHB test (Table 3) in a recent study [20]. I found similar sensitivity but slightly lower specificity with this test in my own comparisons of the urine dipstick to blood BHB from herd investigation data (Table 3). Prolonged contact of urine with the reagent may explain some of the false positive results obtained with the urine ketone tablet or the milk BHB strip. The label for the urine dipsticks states that the test result should be interpreted exactly 15 seconds after contact with the urine sample. Results were read within five seconds in one study [20], and this study reported the highest specificity results for a urine test. Interestingly, results for urine testing with the Ketostix suggest that lower concentrations (eg, ‘‘small’’) should not be ignored if the purpose of the test is to identify cows with SCK that might benefit from treatment for ketosis. Because oral treatment with glucose precursors is generally inexpensive and safe, it is most appropriate to use a low cut point for urine ketones in making individual cow treatment decisions. At a cut point of ‘‘small,’’ only about 2% of urine test negative cows have SCK, and about 43% of urine test positive cows do not have SCK (calculated from pooled data presented in Table 3). Cowside milk tests for subclinical ketosis Cowside milk tests have tremendous advantages over urine cowside tests for ease of collection and for assurance that all eligible cows can be tested. However, milk tests are generally not as sensitive as urine tests in detecting SCK. Nitroprusside powders (Utrecht powder, KetoCheck powder) can be used to qualitatively test milk acetoacetate. However, these tests generally have very poor sensitivity for SCK compared with blood BHB (see Table 4),

668

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

and cannot be recommended as tests for herd-based monitoring. They have some (but limited) value as cowside tests for diagnostic decisions for individual cows. The most promising cowside milk ketone test is a semiquantitative milk BHB test strip manufactured by Sanwa Kagaku Kenkyusho Co., Ltd. (Nagoya, Japan). This test strip is marketed under various names (KetoTest, Ketolac BHB, and Sanketopaper) in different parts of the world. It is not commercially marketed in the United States, although it may be imported into the United States from Canada (CDMV, St. Hyacinthe, Quebec) and costs about $2.00 (USD) per strip. Results of numerous studies evaluating the sensitivity and specificity of the milk BHB test strip compared with blood BHB results are presented in Table 5. My own clinical experience with this test (221 cows from 17 herds) corroborates previously published results. When used at the cut point of 100 lmol/L, this test is about 83% sensitive and 82% specific. For individual cow testing, the 50 lmol/L cut point provides better sensitivity (89%) but has a false positive rate of 69% (calculated from pooled data presented in Table 5). Increasing the cut point to 200 lmol/L reduces test sensitivity to 54% (Table 5). At this higher cut point the test is of little value for diagnosing ketosis in individual sick cows but has potential use for herdbased evaluations, as discussed later. The cowside milk BHB test strip has limited value for herd-based monitoring of SCK. Blood BHB test results are much more reliable for this purpose, and immediate cowside results are not particularly critical for herdbased testing (as they are for individual sick cow diagnosis). The imperfect sensitivity and specificity of the milk BHB test distort the prevalence of SCK in a herd. The true herd prevalence of SCK may be either higher or lower than the prevalence measured by the milk BHB test strip, depending on the cut point chosen (Table 7). The degree of disparity between SCK prevalence determined by the milk BHB strip versus the blood BHB test also depends on the true prevalence of SCK. The best cut point for herd monitoring when using the milk BHB strip appears to be 200 lmol/L. At this cut point the prevalence of test-positive results is similar to the true prevalence, allowing the same alarm level for SCK prevalence (10%) to be used for both tests. Unfortunately, milk BHB test strip prevalence changes little as true prevalence increases (Table 7), rendering the test practically useful only for identifying herds with a very high prevalence of SCK. Results from my herd investigations illustrate the difficulty in using the milk BHB test strip for herd-based monitoring. In nine herds I had sufficient sample size to categorize the herd for SCK using both milk BHB (200 lmol/L cut point, 10% alarm level) and blood BHB (1400 lmol/L cut point, 10% alarm level). Categorization of five herds was the same using either test method. However, two herds classified positive by blood BHB were classified negative by the milk BHB test strip, and both herds had apparently high SCK prevalences (44% and 24%). The classification of two

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

669

Table 7 Expected test positive prevalences for milk ketone BHB test strip results at different true herd prevalences and test strip cut points Herd SCK prevalence category Low

Alarm level

Milk BHB 50 lmol/L (89% sensitivity, 77% specificity): True prevalence 7.5% 10.0% Milk strip test positive prevalence 28.0% 29.6% Milk BHB 100 lmol/L (83% sensitivity, 82% specificity): True prevalence 7.5% 10.0% Milk strip test positive prevalence 22.9% 24.5% Milk BHB 200 lmol/L (54% sensitivity, 94% specificity): True prevalence 7.5% 10.0% Milk strip test positive prevalence 9.6% 10.8%

Moderate

High

15.0% 32.9%

30.0% 42.8%

15.0% 27.8%

30.0% 37.5%

15.0% 13.2%

30.0% 20.4%

Abbreviations: BHB, b-hydroxybutyrate; SCK, subclinical ketosis, defined as blood BHB 1400 lmol/L.

other herds was different for the milk BHB test strip compared with the blood BHB test; however, this was not as concerning because these two herds had intermediate prevalences of SCK. Some of the limitations of using the milk BHB test for herd monitoring might be overcome by substantially increasing the number of cows sampled. This should not be a practical problem given the lower cost and ease of handling milk samples compared with serum samples. Results of repeated testing with the milk BHB strip within the same herd could then be evaluated by control charting. Cowside milk ketone tests have not been evaluated for the effect of sample collection method (strip milk sample versus proportional milk sample from the entire milking) and for time of sample collection relative to feeding. A better understanding of these potential effects could improve the usefulness of the milk BHB test in diagnosing and monitoring ketosis. Nonesterified fatty acid testing for prepartum negative energy balance The NEFA test is used to evaluate the presence of negative energy balance before calving [21]. Cows should stay in positive energy balance up until the last 24 to 48 hours before calving. Negative energy balance is expected in milking cows, so the NEFA test is highly variable and very difficult to evaluate after calving (use the blood BHB test instead after calving). Elevated NEFA concentrations in prefresh cows are associated with increased risk for displaced abomasum after calving [3]. Based on the physiology of NEFA, it is best positioned as a secondary test in a herd already known to have a high incidence of SCK. The NEFA testing helps determine whether the postpartum ketosis is due to precalving negative energy balance. There is only limited value in conducting NEFA

670

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

testing in herds with a low incidence of SCK, because increased risk for SCK is the main result of high NEFA before calving. Michigan workers have described a NEFA cut point of 0.400 mEq/L in prefresh cows between 2 and 14 days from calving. NEFA concentrations normally rise in the last 48 hours before calving, so results from cows that calve this soon after the sample was collected are difficult to interpret and should either be discarded or interpreted with caution. The alarm level for the proportion of cows with elevated NEFA concentrations within a group has not been precisely defined. In my own NEFA testing (245 cows in 29 herds), I have found a mean prevalence of 17.6% elevated NEFA concentrations in cows that calve between 2 and 14 days after testing. I suggest using 10% as a reasonable alarm level for herdbased NEFA testing. Because this is the same alarm level as for blood BHB in postfresh cows (10%), the interpretation of NEFA results is the same as outlined for blood BHB in Figs. 4 and 5. In small dairy herds, the number of prefresh cows eligible for NEFA testing is small, so all potentially eligible cows may need to be tested. Samples may need to be frozen and submitted as a group when actual calving dates are known, and about 12 or more samples have been accumulated. In my clinical experience with herd-based NEFA testing, I have obtained adequate sample size during a single herd visit for only 10 herds of the 29 herds I have attempted to evaluate. Each herd with adequate NEFA sample size had over about 400 total cows. In large dairy herds, only a portion of the prefresh group may be subsampled for NEFA screening. In this case, select the cows that appear to be the closest to calving (based on due dates and visual observation), but avoid those cows in which calving appears to be imminent. In my experience, only about 75% of cows identified for NEFA testing using these criteria will actually calve 2 to 14 days after sample collection. Thus, at least 16 cows should be sampled to have 12 or more valid samples after actual calving dates are known. I have found it extremely useful in some investigations to collect NEFA samples from cows in the maternity pen as well as the prefresh pen (but still not sampling cows starting to give birth). Many of these cows will not calve for several more days, and they are at very high risk for elevated NEFA concentrations because of the move to a new pen. Moving cows to a different pen about 3 and 9 days before actual calving appears to be particularly difficult for them. The negative impact of a pen move just before calving is multiplied if the maternity pen provides inadequate access to feed, water, and resting space. Concentrations of NEFA reach their nadir about 4 to 5 hours postfeeding [16], and peak just before the next feeding. The best approach, therefore, is to sample just before feeding to capture the peak value. The difference between peak and nadir values is probably influenced by the availability of feed throughout the day and the relative size of the meals

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

671

consumed by the cows. Sampling at a consistent time relative to feeding is recommended, and especially when monitoring prefresh cow NEFA concentrations in a herd over time. It is important to keep the plasma samples cool or frozen from the time they are collected from the cow until the time they are received at the laboratory for analysis. Otherwise, some of the triglycerides normally present in the sample may degrade to NEFA and falsely (but slightly) elevate the test results. If a high proportion of elevated NEFA concentrations are detected, then attention should be focused on increasing total energy intake in the prefresh group. This may require increasing the energy density of the prefresh diet, increasing prefresh dietary nonfiber carbohydrate content, improving diet palatability, increasing bunk space, increasing feeding frequency, or increasing daily feed refusals. Reducing pen moves near calving, and improving maternity pen facilities may be even more important than diet changes. Herd-based NEFA testing is costly because of the practical difficulties in accumulating adequate sample size and handling the samples properly. The information provided by NEFA test results may support recommendations for needed changes in prefresh cow management and may also corroborate a diagnosis of type II ketosis in some herds. But in many other herds, the cost of prefresh NEFA testing cannot be justified because herd recommendations can be adequately supported by clinical observations and the days in milk pattern of the postfresh blood BHB results.

Milk fever Herd monitoring for parturient hypocalcemia Both clinical milk fever and parturient hypocalcemia incidence rates can be monitored in dairy herds. Parturient hypocalcemia can be defined as low blood total calcium (\2.0 mmol/L or 8.0 mg/dL) or low blood ionized calcium (\1.0 mmol/L or 4.0 mg/dL), with or without clinical signs of hypocalcemia. Parturient hypocalcemia is a risk factor for subsequent displaced abomasum [22]. Because the duration of parturient hypocalcemia is extremely short (about the first 48 hours after calving), its incidence is monitored instead of its prevalence [4]. Limited data are available to assist in determining an alarm level for parturient hypocalcemia. Two studies with multiparous Holstein cows [5,23] record the incidence of both clinical milk fever and parturient hypocalcemia. In both studies, cows were fed control diets without anionic salts added (average dietary cation–anion difference [DCAD] of 258 mEq/kg of [Na þ K]  [Cl þ S]) and with anionic salts added (average DCAD of 73 mEq/ kg). The addition of anionic salts reduced the incidence of clinical milk fever from 18.5% to 7.7% and the incidence of parturient hypocalcemia from 50.0% to 28.2%. These data suggest that a reasonable alarm levels are

672

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

30% parturient hypocalcemia and 8% clinical milk fever in multiparous Holstein cows. Primiparous cows are at very low risk for low blood calcium around calving. There is little value in including first lactation cows in a monitoring scheme for parturient hypocalcemia, and monitoring of clinical milk fever is necessary only for multiparous cows. Because clinical milk fever incidence is typically expressed as a percentage of the entire herd, incidence rates should be adjusted for the proportion of calvings that were from first lactation animals. For example, if the alarm level for clinical milk fever is 8.0% in multiparous cows and the 35% of the calvings are from first lactation animals, then the alarm level for clinical milk fever in the herd is about 5.2% (8.0%  0.65). It is difficult to monitor the incidence of parturient hypocalcemia in dairy herds. The best time to collect blood samples is about 12 to 24 hours after calving. In most situations the blood samples must be collected by on-farm personnel rather than by the herd veterinarian. The farm then needs either a means of separating the serum (or plasma) and storing it. Additionally, an arrangement for prompt pickup of whole blood samples for transport to an analytical laboratory (or an on-farm system for analysis of blood calcium) is required. Urinary pH for monitoring anion dose Dietary acidification is an effective method of preventing both clinical and subclinical hypocalcemia [24]. The degree of dietary acidification is related to urinary pH [25]. Several studies suggest that DCAD and milk fever prevention are optimal at urinary pH values of about 6.0 to 7.0. If mean urinary pH is high, you do not know how close the group might be to optimal acidification [26]. In small dairy herds it may be necessary to check all eligible prefresh cows for urinary pH. Results from about eight cows or more should be accumulated before calculating a mean value, interpreting the result, and possibly adjusting the diet. The most practical strategy for obtaining adequate sample size is to have on-farm personnel do the testing. Urinary pH can be determined satisfactorily with pH paper—a calibrated pH meter is not required. The effect of time postfeeding on urinary pH is small when access to feed is good throughout the day [27]. If feed access is not good throughout the day for prefresh cows, then you have identified a problem that is much bigger than any concerns about dietary acidification and urinary pH. Mean urinary pH values above about 7.0 (when acidifying diets are supposedly being fed) indicate that the acidification is not optimal. A common culprit is undetected increased potassium content of one of the forages. Either more anionic salts or less high DCAD forage should be fed. A small change in actual DCAD consumed by the cows could dramatically move the urinary pH values into the optimal range [26].

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

673

Mean urinary pH values below about 6.0 are uncommon, and indicate overacidification of the diet. Reduced dry matter intake (with secondary negative energy balance before calving) is likely when urine is overacidified. A common culprit in these situations is overzealous supplementation of anionic salts in the face of a milk fever outbreak. An undetected high chloride content in one of the forages could also contribute to the low urinary pH. In either case, the dose of anions added to the prefresh diet should be reduced; otherwise, dry matter intake of the prefresh group may be unnecessarily suppressed. Summary Clinical impressions of metabolic disease problems in dairy herds can be corroborated with herd-based metabolic testing. Ruminal pH should be evaluated in herds showing clinical signs associated with SARA (lame cows, thin cows, high herd removals or death loss across all stages of lactation, or milk fat depression). Testing a herd for the prevalence of SCK via blood BHB sampling in early lactation is useful in almost any dairy herd, and particularly if the herd is experiencing a high incidence of displaced abomasum or high removal rates of early lactation cows. If cows are experiencing SCK within the first 3 weeks of lactation, then consider NEFA testing of the prefresh cows to corroborate prefresh negative energy balance. Finally, monitoring cows on the day of calving for parturient hypocalcemia can provide early detection of diet-induced problems in calcium homeostasis. If hypocalcemia problems are present despite supplementing anionic salts before calving, then it may be helpful to evaluate mean urinary pH of a group of the prefresh cows. Quantitative testing strategies based on statistical analyses can be used to establish minmum sample sizes and interpretation guidelines for all of these tests. References [1] Garrett EF, Pereira MN, Nordlund KV, Armentano LE, Goodger WJ, Oetzel GR. Diagnostic methods for the detection of subacute ruminal acidosis in dairy cows. J Dairy Sci 1999;82(6):1170–8. [2] Duffield TF. Effects of a monensin controlled release capsule on energy metabolism, health, and production in lactating dairy cattle. Thesis dissertation. Guelph (Ontario): University of Guelph; 1997. [3] Cameron REB, Dyk PB, Herdt TH, Kaneene JB, Miller R, Bucholtz HF, et al. Dry cow diet, management, and energy balance as risk factors for displaced abomasum in high producing dairy herds. J Dairy Sci 1998;81(1):132–9. [4] Oetzel GR. Effect of calcium chloride gel treatment in dairy cows on incidence of periparturient diseases. J Am Vet Med Assoc 1996;209(5):958–61. [5] Oetzel GR, Olson JD, Curtis CR, Fettman MJ. Ammonium chloride and ammonium sulfate for prevention of parturient paresis in dairy cows. J Dairy Sci 1988;71(12):3302–9. [6] Oetzel GR. Clinical aspects of ruminal acidosis in dairy cattle. In: Proceedings of the 33rd Conference of the American Association of Bovine Practitioners, Rome (GA); 2000. p. 46–53.

674

G.R. Oetzel / Vet Clin Food Anim 20 (2004) 651–674

[7] Nordlund KV, Garrett EF. Rumenocentesis: a technique for collecting rumen fluid for the diagnosis of subacute rumen acidosis in dairy herds. Bovine Practit 1994;28(1):109–12. [8] Nordlund KV, Garrett EF, Oetzel GR. Herd-based rumenocentesis: a clinical approach to the diagnosis of subacute rumen acidosis. Compend Contin Educ Pract Vet 1995;17(8):S48–56. [9] Simensen E, Halse K, Gillund P, Lutnaes B. Ketosis treatment and milk yield in dairy cows related to milk acetoacetate levels. Acta Vet Scand 1990;31(4):433–40. [10] Duffield T, Bagg R. Herd level indicators for the prediction of high-risk dairy herds for subclinical ketosis. In: Proceedings of the 35th Conference of the American Association of Bovine Practitioners, Rome (GA); 2002, p. 175–6. [11] Duffield TF, Sandals D, Leslie KE, Lissemore K, McBride BW, Lumsden JH, et al. Efficacy of monensin for the prevention of subclinical ketosis in lactating dairy cows. J Dairy Sci 1998;81(11):2866–73. [12] Tyopponen J, Kauppinen K. The stability and automatic determination of ketone bodies in blood samples taken in field conditions. Acta Vet Scand 1980;21(1):55–61. [13] Holtenius P, Holtenius K. New aspects of ketone bodies in energy metabolism of dairy cows: A review. J Vet Med A 1996;43(10):579–87. [14] Herdt TH. Ruminant adaptation to negative energy balance: influences on the etiology of ketosis and fatty liver. Vet Clin North Am 2000;16(2):215–30. [15] Kronfeld DS, Raggi F, Ramberg CF Jr. Mammary blood flow and ketone metabolism in normal, fasted, and ketotic cows. Am J Physiol 1968;215(1):218–27. [16] Eicher R, Liesegang A, Bouchard E, Tremblay A. Influence of concentrate feeding frequency and intrinsic factors on diurnal variations of blood metabolites in dairy cows. In: Proceedings of the 31st Conference of the American Association of Bovine Practitioners. Rome (GA); 1998. p. 198–202. [17] Manston R, Rowlands GJ, Little W, Collis KA. Variability of the blood composition of dairy cows in relation to time of day. J Agric Sci (Camb) 1981;96(6):593–8. [18] Osborne TM, Leslie KE, Duffield T, Petersson CS, Ten Hag J, Okada Y. Evaluation of KetoTest in urine and milk for the detection of subclinical ketosis in periparturient Holstein dairy cattle. In: Proceedings of the 35th Conference of the American Association of Bovine Practitioners. Rome (GA); 2002. p. 188–9. [19] Nielen M, Aarts MGA, Jonkers AGM, Wensing T, Schukken YH. Evaluation of two cowside tests for the detection of subclinical ketosis in dairy cows. Can Vet J 1994;35(4): 229–32. [20] Carrier J, Stewart S, Godden S, Fetrow J, Rapnicki P. Evaluation of three cow-side diagnostic tests for the detection of subclinical ketosis in fresh cows. In: Proceedings of the 2003 Four-State Nutrition Conference at La Crosse (WI). Ames (IA): Midwest Plan Service; 2003. p. 9–13. [21] Herdt TH. Variability characteristics and test selection in herd-level nutritional and metabolic profile testing. Vet Clin North Am 2000;16(2):387–403. [22] Massey CD, Wang C, Donovan GA, Beede DK. Hypocalcemia at parturition as a risk factor for left displacement of the abomasum in dairy cows. J Am Vet Med Assoc 1993;203(6): 852–3. [23] Joyce PW, Sanchez WK, Goff JP. Effect of anionic salts in prepartum diets based on alfalfa. J Dairy Sci 1999;80(11):2866–75. [24] Oetzel GR. Use of anionic salts for prevention of milk fever in dairy cattle. Compend Contin Educ Pract Vet 1993;15(8):1138–46. [25] Vagnoni DB, Oetzel GR. Effects of dietary cation-anion difference on the acid-base status of dry cows. J Dairy Sci 1998;81(6):1643–52. [26] Oetzel GR. Management of dry cows for the prevention of milk fever and other mineral disorders. Vet Clin North Am 2000;16(2):369–86. [27] Goff JP, Horst RL. Effect of time after feeding on urine pH determinations to assess response to dietary cation-anion adjustment. J Dairy Sci 1998;81(Suppl 1):44 [abstract 173].