| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* Bovine Research Australasia, Camden 2570, NSW, Australia
MC Franklin Laboratory, Department of Veterinary Science, University of Sydney, Australia
Church & Dwight Co., Inc., Arm & Hammer Animal Nutrition Group, Princeton, NJ 08543
1 Corresponding author: ianl{at}dairydocs.com.au
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Key Words: milk fever • meta-analysis • dairy cow • dietary cation-anion difference
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Meta-analysis, the quantitative analysis of previous studies, provides opportunities to investigate previously proposed hypotheses and to develop new hypotheses from large databases. In the study of disorders of relatively low incidence, such as milk fever, the availability of large numbers of observations provided by meta-analysis is useful for evaluating the effect on disease of factors that have physiologic effect. One such factor is DCAD. The DCAD equation cited by Ender et al. (1962) and used by Block (1984) {DCAD = (Na+ + K+) – (Cl– + S2–) [Equation 1]} is the most commonly used form of the equation. Horst et al. (1997) recommended that other anions and cations be included in the equation and proposed DCAD = (0.38 Ca2+ + 0.3 Mg2+ + Na+ + K+) – (Cl– + S2–) [Equation 2]. Goff (2000) proposed a variation of this equation based on the capacity of different salts to acidify urine and recommended DCAD = (0.15 Ca2+ + 0.15 Mg2+ + Na+ + K+) – (Cl– + 0.25 S2– + 0.5 P3–) [Equation 3]. Following research of Spears et al. (1985), who estimated that the absorption of sulfur from the gastrointestinal tract was 60% of dietary intake, Tucker et al. (1991) suggested that DCAD = (0.38 Ca2+ + 0.3 Mg2+ + Na+ + K+) – (Cl–+ 0.6 S2– + 0.5 P3–) [Equation 4]. One goal of the study was to determine, which, if any of these equations most accurately predicted risk of milk fever.
The optimal dietary concentration of calcium in prepartum diets is also contentious. Goff (2000) concluded that calcium concentration in prepartum diets had little influence on the incidence of milk fever when fed at levels above the daily requirements, approximately 30 g of calcium/d. Oetzel (2000) and Thilsing-Hansen et al. (2002) noted that the practice of feeding very low levels of calcium prepartum, <20 g/d, is effective in controlling hypocalcemia. However, Oetzel (2000) recommended a daily intake of 150 g of calcium/d in the prepartum diet, a calcium concentration of between 1.1 and 1.5% of DM, in conjunction with a dietary DCAD (based on Equation 1) of approximately –15 mEq/100 g of DM. However, this recommendation was not supported by his meta-analysis, which showed that the highest milk fever risk occurred with a dietary calcium concentration of 1.16% (Oetzel, 1991). Lean et al. (2003) suggested a prepartum intake of 60 g of calcium/d based on the seminal studies of Boda and Cole (1954) and Ramberg et al. (1976, 1984).
To investigate and attempt to clarify these areas of contention, data from the meta-analysis of Oetzel (1991) were updated with relevant, subsequent trials and reanalyzed. Oetzel (1991) concluded that his results substantiated the DCAD theory of milk fever prevention and concluded that the role of calcium in milk fever was nonlinear; extremely low and extremely high concentrations reduced the relative risk of milk fever. It was concluded that sulfate and sodium were the dietary electrolytes most closely linked with milk fever and were more important than DCAD or calcium concentration. Enevoldsen (1993) expressed concern at the methods used by Oetzel (1991), including the sparseness of data in regard to the number of covariate patterns, and reanalyzed Oetzel’s data. Enevoldsen (1993) similarly found that low dietary sulfur increased the risk of hypocalcemia, but disputed other conclusions of Oetzel (1991). In particular, calcium entered final models developed by Enevoldsen (1993) as a linear effect. Both Enevoldsen (1993) and Oetzel (1991) used fixed effects models for predicting outcomes, and it is now widely accepted that these models are vulnerable to over-dispersion associated with clustering of effects in trials. Random effects models, for example logistic normal regression, are preferred for such analyses (Dohoo et al., 2003).
The aim of this study was to develop random effect, predictive models to test hypotheses arising from recent reviews using the much larger number of studies and improved statistical methods available since the meta-analyses of Oetzel (1991) and Enevlodsen (1993).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
In all, 48 English-language, published papers were reviewed. Of the papers that were assessed, only those that satisfied specific predetermined criteria were included in the analyses. These criteria included evidence 1) that the trial was randomized, 2) that the populations studied were pregnant dairy cows that calved during the trial period, and 3) that enough detail on dietary composition was provided to allow calculation of the DCAD of the transition diet without use of book values for mineral composition of feeds. Trials were ineligible for inclusion in the analysis 1) if there was insufficient evidence of randomization, 2) if the trial was confounded with supplementary treatments such as bovine somatotropin, 3) if the paper provided insufficient data on the variables being measured, or 4) if hypocalcemia was induced in the study animals by use of either feed restriction or calcium binding agents (such as NaEDTA given intravenously in cows or zeolites in feed). Length of exposure to the prepartum transition diet was determined for each trial that reported this variable. Breed classification was recorded, and lactation number was calculated using methods described by Oetzel (1991). The 4 variations of the DCAD equation detailed previously were calculated where possible in units of mEq/100 g of DM and were tested in statistical models developed. The final data set was derived from 35 papers detailing 137 individual trials involving 2,545 observed calvings. Details of the trials investigated are presented in Appendix A
.
|
Statistical Analysis
Initial investigation was conducted using graphical display of variables plotted against milk fever incidence using logistic regression with a fixed effect of trial. Milk fever incidence was assessed by logistic regression using Statistix 8 (Statistix, 2003). Following univariate analyses, a process of model building was undertaken using forward and backward stepping and model fitting. Variables with a P < 0.3 in univariate analyses were allowed into later models for testing. This process was conducted for each of the different data sets. Fixed effect models were assessed on improvement in model fit using Hosmer-Lemeshow statistics, significance of variables entering, receiver-operator curves, and proportion of cases correctly predicted.
We used a systematic method of evaluating data univariately and subsequently with multivariate fixed and random effects models before using a subset of data that contained studies with detail of duration of exposure to the prepartum diet, but not including age or protein content of the diet. Investigation of variables in the different data subsets provided confidence that significant variables were not included in final models by chance; rather, these were very consistently significant in a large number of models developed with the different subsets of data. Final models were forward stepped and, interactions among variables and quadratic terms for included variables were sequentially included or excluded if not significant.
Following preliminary analysis, random effects models were used to evaluate these data (Egret, version 2.0.31, Cytel Software Corporation, Cambridge, MA). A random effects, logistic normal model was used based on the distribution of incidence of milk fever in trials, to develop final models, and interaction terms in the final models selected were then tested. Evidence of confounding was assessed from change in the coefficients when subsequent variables were entered. Biological plausibility of variables was also considered in the inclusion of variables that could have a quadratic response as well as any interactions between variables. The model for the response probability 
i is given by
 |
where 
is a constant >0, and 
N(0, 1). Unlike the beta-binomial and logistic binomial models, the logistic normal model yields a marginal likelihood for which there exists no closed form (Egret, Cytel Software Corp.). Logistic normal models are appropriate to use in analyses where data are clustered within a variable, for example litter, family, or in this case herd. In such situations, the effect of herd is included in the logistic normal model as a random effect.
The models were assessed for fit by examining graphs that showed residuals for each trial in the data set. Further examination of the final models was conducted by using linear regression to evaluate the predicted milk fever incidence derived from models against the observed milk fever incidence from the 37 trials excluded because of missing values from the original data used to generate the models. The missing values in these trials were replaced with mean values from the analyzed data. Residual plots from these models were evaluated.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
. Initial exploration of the data using graphs of variables studied plotted against milk fever incidence did not reveal any obvious cut points for any of the variables investigated. Therefore, all variables were treated as continuous to minimize the loss of information. Figure 1, for example, shows an essentially linear association between DCAD and milk fever risk.
|
|
). These models were further evaluated and modified by fitting random effects models to the originally defined variables.
|
and 4) were developed that had comparable predictive values and fit (Table 5). For both of these models, there was considerable evidence of overdispersion of data as indicated by a marked decrease in model deviance when random effects models were fitted. There was also a large amount of subject heterogeneity in data as shown by the large estimates of the variance component, 
2, relative to the standard error. Both these observations support use of random effects models to evaluate these data.
|
|
|
 |
 |
The reference breed in each of the models is the Holstein-Friesian breed. The adjustments for the coefficient in the logit models for breeds other than Holstein-Friesian are as follows: for the Jersey breed, 0.86 and 0.66 for Models 1 and 2 respectively, and for Norwegian and Swedish Red and White breeds, 1.49 for Model 1. Only the Jersey breed differed significantly from the Holstein-Friesian breed in risk of milk fever in Equation 1 (P < 0.05) and approached significance P = 0.09 in Equation 2. Risk of milk fever in cows of mixed breed or where no breed was recorded compared with the Holstein-Friesian breed did not approach significance in either model (P > 0.2).
Predicted incidence of milk fever can be calculated from each of the logit transformations using the following equation:
 |
The response to dietary calcium concentration, within the range of analyzed data, for Model 1 is presented in Figure 2. This was calculated using breed = Holstein-Friesian, magnesium = 0.4%, phosphorus = 0.4%, potassium = 1.6%, sodium = 0.3%, sulfur = 0.4%, chlorine = 1.8%, exposure = 14 d, and DCAD 1 = –21 mEq/100 g of DM. These factors were levels for those variables at which risk of milk fever was low.
|
, from trials included in the final model development. For example, missing values for calcium, sulfur, or chlorine were replaced with 1.02, 0.29, and 0.67% of DM, respectively. Of the 51 trials, 34 trials had one missing variable, 3 trials had 2 missing variables, 2 trials had 4 missing variables, 8 trials had 5 missing variables, and 4 trials had 7 missing variables. The 14 trials with >4 missing variables were excluded from the analysis, leaving 37 trials for analysis. The 2 models were fitted to the data, and predicted incidences of milk fever were retained and evaluated against the observed incidence for each trial. Linear regression weighted for trial size was used to assess the fit of models (Table 6). Both models provided estimated incidence rates that produced significant linear regression models; intercepts approximated zero, and slopes approached unity. Both models had standardized residuals that, when fitted against observed values, demonstrated appropriate model selection.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
The lack of data on dietary composition reported in studies, particularly with regard to dietary protein and carbohydrate fractions, and feed intake limited testing of hypotheses relating to these variables. In contrast to the model developed by Oetzel (1991), dietary crude protein was not included in final models. It did, however, approach significance in a number of preliminary fixed effects models (0.1 < P < 0.2). The lack of information that could be obtained on age of cows also lead to an inability to include age in final models developed. There is good evidence that increasing age increases risk of milk fever as a result of decreased intestinal calcium absorption and responsiveness to hypocalcemia (Hansard et al., 1954; Horst et al., 1978, 1990), reduced bone turnover (Parfitt, 1984), and decreased bone responsiveness to parathyroid hormone and vitamin D (Goff et al., 1991a). This lack of data is a limitation to the models presented, which would likely be improved by inclusion of age data. Nonetheless, meta-analysis of large data sets is a powerful tool for evaluating whether physiological effects, such as those of anionic DCAD treatments or efficiency of absorption of minerals, identified in smaller, more tightly controlled studies, have an effect on disease risk.
Two final models with good fit and good predictive ability are presented in this paper. The final models were developed using random effects models. This approach was supported by the observed decrease in variance in the random vs. fixed effects models and in the high estimates of variance relative to the standard deviations found in the final models. There are, consequently, several strong reasons for conducting this meta-analysis, specifically, the greater availability of data than available for Oetzel (1991), the use of random effects models that are required proper evaluation and presentation of these data, and the examination of new hypotheses, especially the value of different DCAD equations recently proposed.
The fit of random effects models is difficult to assess (Dohoo et al., 2003). In this case, however, expected incidence data produced by the models could be fitted against the observed incidence both from data used to produce the models and in the independent data provided in trials with missing data. The linear regression models (Tables 5 and 6) showed good fits, providing support for the final models developed. Although effects of age on risk of milk fever could not be included in the final models, this effect and the effect of breed of cow that was included in the model, are not immediately available to manipulation by farmers or their advisers. Consequently, the approximately 40% variance explained by the models developed is very satisfactory given that the outcome of interest, milk fever was not absolutely consistently nor comprehensively defined for cows in the studies.
The decision to retain 2 models for presentation reflected strong evidence in univariate assessment that a low DCAD (assessed by Equation 1) was associated with a low incidence of milk fever. Exploratory analysis indicated that this was, clearly, the best of the DCAD models for explaining milk fever incidence. During the process of model development, DCAD Equations 2, 3, and 4 did not approach significance in either the fixed or random effect models. Milk fever risk declined linearly with DCAD assessed by Equation 1 (Figure 1), in contrast with the nonlinear response of urinary pH to lower DCAD (McNeill et al., 2002). In Model 1, an increase in DCAD in the precalving diet from –25 to +25 mEq/100 g increases risk of milk fever by 110%. There was no additional gain in model predictions or fit from the inclusion of dietary concentrations of multivalent ions other than sulfur or from the inclusion of absorption coefficients in DCAD equations. The inclusion of calcium, magnesium, or phosphorus in DCAD equations to predict milk fever risk must, therefore, be questioned. In both models, the second-order effect of calcium, contrasted with a linear effect predicted by DCAD Equations 2, 3, and 4. Increasing dietary concentrations of magnesium reduce milk fever risk markedly, again in contrast to the predicted effect of the DCAD Equations 2, 3 and 4. Similarly, increasing phosphorus concentrations increase milk fever risk, whereas a protective role is proposed in DCAD Equations 2, 3 and 4. These observations reinforce the difference between models designed to predict DCAD with those designed to predict milk fever. Strategies designed to solely manipulate DCAD to alter acid-base status are not necessarily those best designed to minimize milk fever risk.
The final models are similar in modeling risk of milk fever but differ in a critical aspect, specifically, the roles of chloride and sodium. Neither sodium nor chloride was included in Model 2, but both were present in Model 1 through inclusion of DCAD (Equation 1). This is surprising, as there is a substantial body of evidence, both experimental and from field observations, to support an important role of these ions in the pathogenesis of milk fever. High dietary sodium content is a risk factor for milk fever (Dishington, 1975; Goff and Horst, 1997; Goff et al., 1991a; Oetzel, 1991), and there is also anecdotal evidence that sodium bicarbonate fed before calving increases milk fever incidence. Conversely, increasing dietary chloride intake reduces the risk of milk fever (Oetzel et al., 1988; Gaynor et al., 1989; Goff and Horst, 1998; Goff et al., 1991a; Tucker et al., 1991). The lack of chloride or sodium in Model 2 may reflect a lack of reporting of chloride or sodium values for many early trials originally analyzed by Oetzel (1991). These concentrations were often calculated from book values by Oetzel (1991) and were subsequently used in this study. There is a greater coefficient of variation in feed chloride concentrations than in feed sulfur concentrations. The effect of nondifferential errors in measurement is to drive hypotheses toward the null, potentially resulting in lack of significance for important factors. Consequently, we suggest that inclusion of chloride and sodium through DCAD in Model 1 is preferred over the omission of these in Model 2.
Potassium and sulfur both had a highly significant and substantial effect on milk fever incidence. Higher potassium concentrations greatly increased milk fever incidence, and higher sulfur concentrations strongly and linearly reduced milk fever risk. These observations are consistent with the proposed effects of strong ions on milk fever incidence mediated through the DCAD (Block, 1984; Goff et al., 1989; Tucker et al., 1991; Goff and Horst, 1997; Constable, 1999; Goff, 2000). It is possible also that potassium may have an effect on milk fever mediated by a negative effect on magnesium absorption (Suttle and Field, 1967); however, statistical interactions between potassium and magnesium concentrations were not significant, providing no support for this interaction playing a critical role in the pathogenesis of milk fever. The role of sulfur in prevention of milk fever has been contentious. The effect of sulfur on milk fever risk was linear in both Models 1 and 2, suggesting the possibility of an effect not related to DCAD; however, such mechanisms are yet to be demonstrated. Ramberg et al. (1996) suggested several mechanisms by which sulfur could reduce the risk of milk fever, including an acidogenic effect, if gastrointestinal absorption of SO42– was preferential over Mg2+, and a possible laxative effect of magnesium sulfate that might increase bicarbonate losses in the feces. It will be difficult to determine whether any such physiological actions of sulfur, apart from a role in DCAD, influence risk of milk fever independently of the role in DCAD.
The second-order effect of calcium present in both models strongly supports the concept that either low dietary calcium percentage (Boda and Cole, 1954; Goings et al., 1974; Wiggers et al., 1975) or high dietary calcium percentage (Lomba et al., 1978; Oetzel et al., 1988) fed prepartum reduces milk fever risk. The effect of low calcium diets on calcium homeostasis is well established and has been reviewed in depth (Goings et al., 1974; Green et al., 1981). Goff and Horst (1997), however, proposed that observed benefits of lowering dietary calcium prepartum to approximately 50 g/d in preventing milk fever were, in part, the result of a reduction of dietary potassium rather than a stimulatory effect of low dietary calcium on calcium homeostasis. Those researchers cited evidence that dietary calcium needed to be restricted to <20 g/d to sufficiently stimulate the calcium homeostatic mechanism that prevents milk fever. It was also found that when the effects of calcium were separated from the effects of the strong dietary cations, calcium consumption above the dietary requirements of the cow had little effect on the incidence of milk fever. These postulates are not supported by this study, which shows that the effects of calcium and potassium were independent and that the effect of calcium content of the diet in the final models on milk fever risk is quadratic. The effect of raising dietary calcium concentration from 0.5 to 0.6% while maintaining all other variables, as predicted by Model 1, is to increase the risk of milk fever by approximately 37%. An increase in calcium concentration from 0.5 to 1.0% in the precalving diet would increase the risk by 327%.
The process by which very high dietary calcium concentrations prepartum may reduce milk fever risk is unclear. It has been suggested that higher concentrations of calcium increase uptake of calcium by passive absorption and may counteract the depletion of calcium associated with low DCAD diets (Lean et al., 2003). The hypercalciuric effect of low DCAD diets (Vagnoni and Oetzel, 1998; van Mosel et al., 1993) may lower readily available bone calcium and, hence, bone calcium reserves available for mobilization after calving. Feeding higher dietary calcium concentrations prepartum may prevent this. This hypercalcuric effect may be exacerbated with increased duration of exposure to a low DCAD diet prepartum. Longer exposure to a prepartum transition diet would, therefore, increase the incidence of milk fever as predicted in both models. Although increased urinary calcium loss on low DCAD diets has been demonstrated (van Mosel et al., 1993; Vagnoni and Oetzel, 1998), an effect of duration of exposure to the low DCAD diet on milk fever risk has not been established in trial work. Exposure was a consistently significant variable in models developed with this variable and acted to substantially modify coefficients for calcium and magnesium. Increasing exposure to the diet before calving from 20 to 30 d increased risk of milk fever by 42%. Therefore, it was necessary to include this variable in models, in contrast to those previously developed by Oetzel (1991) and Enevoldsen (1993). Further, an interaction term between the quadratic effect of calcium and exposure was significant in many models (e.g., Table 2) and approached significance (P < 0.15), but was not included in the final models. This term suggested that short exposures to higher concentrations, >1.5%, of calcium increase risk of milk fever, whereas longer term exposures to high concentrations of calcium decrease risk. Given, that there are fewer studies that fed >1.5% calcium in the prepartum diet than fed <0.5%, we consider that the observed quadratic response in milk fever risk to calcium concentrations should be tested in experimental protocols that also examine length of exposure to the prepartum diet.
The regression equations calculated also showed a very strong association between higher concentrations of magnesium in the diet and a lower incidence of milk fever. For example, an increase in magnesium concentration from 0.3 to 0.4% of DM, while maintaining the other variables, would result in an approximate 62% decrease in milk fever risk. There are sound physiological bases for a protective role of magnesium in the pathogeneses of milk fever. Apart from the role of magnesium in the synthesis of proteins, RNA, and DNA, many enzymes such as those involved in the dephosphorylation of ATP and kinases are magnesium dependent (Martens and Schweigel, 2000). Magnesium is critical in the release of parathyroid hormone and in the synthesis of 1,25-dihydroxycholecalciferol. In hypomagnesaemic states, kidney and bone are less responsive to parathyroid hormone as a result of the reduced synthesis of the intermediary cellular messengers adenylate cyclase and phospholipase C (Sampson et al., 1983; Goff, 2000). Metabolic acidosis, such as that induced by a low DCAD prepartum diet, may also increase renal excretion of magnesium (Horst and Jorgensen, 1974; Fredeen et al., 1988). Experimental evidence supports a key role for magnesium in calcium homeostasis. Wang and Beede (1992) found that nonpregnant, nonlactating cows fed a diet high in magnesium had lower renal calcium excretion than those fed a diet low in magnesium. Contreras et al. (1982) demonstrated that hypomagnesemic cows were less able to respond to an acute drop in plasma calcium secondary to intravenous administration of NaEDTA than were normomagnesemic cows, and van de Braak et al. (1987b) also found that calcium mobilization rates at parturition were slower when cows were fed a prepartum diet deficient in magnesium.
Phosphorus was also a significant predictor of milk fever, as increasing phosphorus concentrations increased milk fever risk. For example, increasing the phosphorus concentration from 0.3 to 0.4% in the pre-calving diet would increase risk of milk fever by 18%. There is a sound physiological basis for this relationship. Although phosphorous concentrations are not as tightly regulated as calcium, both are closely related to plasma PO4 concentrations regulated directly by 1,25 (OH) vitamin D3 and indirectly by the parathyroid hormone/calcium negative feedback loop (Goff, 1999). However, in rats, hyperphosphatemia can inhibit the activity of renal 25-hydroxyvitamin D 1 alpha-hydroxylase despite increased synthesis of parathyroid hormone thereby reducing production of 1,25 OHD3 sufficiently to cause hypocalcemia (Tallon et al., 1996; Silver et al., 1999; Masuyama et al., 2000). In cattle, there is evidence that a prepartum diet high in phosphorus can have a negative impact on calcium homeostasis possibly by the same pathways (Julien et al., 1977; Kichura et al., 1982; Barton et al., 1987).
The increased risk of milk fever for Jersey and Norwegian and Swedish Red and White cows compared with Holstein-Friesian cows was not surprising, as these risks are well recognized. Many of the studies involving Norwegian and Swedish Red and White cows were older, and breeding programs to select for lower disease risks may alter the magnitude of this risk.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Received for publication May 30, 2005. Accepted for publication September 6, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  G. F. Tremblay, Z. Nie, G. Belanger, S. Pelletier, and G. Allard Predicting timothy mineral concentrations, dietary cation-anion difference, and grass tetany index by near-infrared reflectance spectroscopy J Dairy Sci, September 1, 2009; 92(9): 4499 - 4506. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. J. Lean, A. R. Rabiee, T. F. Duffield, and I. R. Dohoo Invited review: Use of meta-analysis in animal health and reproduction: Methods and applications J Dairy Sci, August 1, 2009; 92(8): 3545 - 3565. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Charbonneau, P. Y. Chouinard, G. F. Tremblay, G. Allard, and D. Pellerin Timothy silage with low dietary cation-anion difference fed to nonlactating cows J Dairy Sci, May 1, 2009; 92(5): 2067 - 2077. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Siciliano-Jones, P. W. Jardon, M. Kucerak, and M. B. de Ondarza Case Study: Early Lactation Production, Body Condition, and Incidence of Disease in Holstein Cows Fed a Low-Potassium Diet Alone or Supplemented with Chloride Prepartum Professional Animal Scientist, December 1, 2008; 24(6): 661 - 667. [Abstract] [PDF]
|
|
|
|
|
|
G. B. Penner, G. F. Tremblay, T. Dow, and M. Oba Timothy Hay with a Low Dietary Cation-Anion Difference Improves Calcium Homeostasis in Periparturient Holstein Cows J Dairy Sci, May 1, 2008; 91(5): 1959 - 1968. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. F. Duffield, A. R. Rabiee, and I. J. Lean A Meta-Analysis of the Impact of Monensin in Lactating Dairy Cattle. Part 1. Metabolic Effects J Dairy Sci, April 1, 2008; 91(4): 1334 - 1346. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. P. Woodrow, C. J. Sharpe, N. J. Fudge, A. O. Hoff, R. F. Gagel, and C. S. Kovacs Calcitonin Plays a Critical Role in Regulating Skeletal Mineral Metabolism during Lactation Endocrinology, September 1, 2006; 147(9): 4010 - 4021. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. R. Roche and D. P. Berry Periparturient climatic, animal, and management factors influencing the incidence of milk Fever in grazing systems. J Dairy Sci, July 1, 2006; 89(7): 2775 - 2783. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Husband, R. Vecqueray, and M. Green Calcium in transition diets and its effect on milk fever. Vet Rec., June 3, 2006; 158(22): 771 - 772. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
