Milk fever and alert downer cows: Does hypophosphatemia affect the treatment response?

Abstract

The purpose of this prospective cohort study was to identify factors that place a dairy cow with uncomplicated milk fever (MF) at significant risk of becoming an alert downer cow (ADC) and to verify if these factors could be used to predict treatment outcome. Recumbent MF cows were examined before treatment and 52 were excluded due to complications. In all, histories and pretreatment serum samples were taken and the serum of 86 cows was analyzed for electrolyte levels (calcium, phosphorus, magnesium, and potassium). In total, 36 of the 86 samples were from ADCs and 50 from animals that responded to MF treatment (MFT). A binary-two-factor logistic model determined that a MF cow with a phosphorus pretreatment level of ≥ 0.9 mmol/L was 12 times more likely not to become an ADC than one with a phosphorus level < 0.9 mmol/L (CI: 6.3,23.1). Also, a binary multivariable logistic regression analysis showed that a MF cow with a pretreatment calcium level ≥ 1.7 mmol/L was 14 times more likely to become an ADC than one with a serum level < 1.7 mmol/L (CI: 2.0,98). Age and the other serum electrolytes were not statistically significant risk factors at the 0.05 level. The rigorous pretreatment examination and stringent adherence to protocol reduced ADC misclassification and fostered the strong association between single factor serum phosphorus levels and ADCs. By using a cutoff level of serum phosphorus at ≥ 0.9 mmol/L, a practitioner could correctly predict that 95% of the MFs would not become ADCs and, therefore, this level would be a useful pretreatment predictor.

Introduction

A downer cow (DC) is an animal that is unable to rise to a standing position (1). Downer cows (DCs) are a common presentation in the periparturient period. They can be divided into 3 categories: (a) cows that are unresponsive to standard hypocalcaemic or milk fever (MF) therapy and do not exhibit other complications but remain alert (ADCs); (b) alert recumbent animals that have traumatic musculoskeletal and nerve problems; (c) and recumbent animals that are affected with systemic diseases related to metabolic, toxic, alimentary, or neurologic conditions (1–3).

Calcium gluconate was established as one of the main medicaments for MF in 1930 (4). Mention of occurrences of DCs associated with milk fever treatment (MFT) failure began in the years preceding the outbreak of the 2nd World War (5). Since then, it has been reported that from 3.8% to 28.2% of all milk fever cases become alert downer cows (ADCs), with a case fatality of 20% to 67.0% (1). A retrospective study of case records at the Ontario Veterinary College, published in 1970, estimated that the percentage of cows treated for parturient paresis (MF) that became DCs was 4.5% (6).

There is no universal cause of the DC syndrome, but it is most frequently a sequela to MF where complications, such as muscle necrosis and nerve paralysis, can arise because of delayed or insufficient calcium (Ca) replacement. Additionally, secondary metabolic disorders involving phosphorus (P), magnesium (Mg), and potassium (K) deficits have been suggested as risk factors, but without direct evidence of their involvement (7). To better define therapeutic strategies and measure their benefits for ADCs, precise classification of recumbent animals into the (a) category and recording of biologically influential factors are needed before relevant associations on which to base therapy can be found (8).

The objectives of this study were to determine if pretreatment levels of serum electrolytes or age differs between ADCs and cows that respond to standard MFT and if these variables are factors that affect the response to MFT and could be used to predict if a MF cow will become an ADC.

Materials and methods

In total, 529 MF cows were examined and treated during a 3-year period (1986–1988) by a private veterinary practice in southeastern Quebec at the request of farmers. For every animal, the cause of recumbency was investigated before each treatment if a cow needed more than one. A diagnosis of MF relied on the history, signalment, and presence of clinical signs characteristic of hypocalcaemia. Other problems were identified either by careful physical examination, laboratory analyses, or necropsy. Complicated MF cases (n = 52), namely, cows with nutritional or locomotor ailments prior to calving, cows that had been recumbent in excess of 10 h prior to 1st treatment, cows with traumatic conditions suffered at or around parturition, and cows that were not given proper and timely care, were excluded from the study.

All MF cases without complications in a 96-hour interval after calving (n = 477) were given a slow IV injection (1000 mL) of 1 of 3 commercial products: Cal-Dextro # 2 (Fort Dodge Laboratories, Fort Dodge, Iowa, USA), Norcalciphos (Norden Laboratories, Lincoln, Nebraska, USA), Maglucal Plus (Rogar/STB, London, Ontario). All cows still down 6–12 h after the 1st treatment were reexamined and given more Ca if they appeared dull or were still anorectic. Cows were recorded as: 1) positive responders (PRs), where the cow rose to a standing position after 1, 2, or 3 Ca treatments within 24 h or where the cow responded positively to the initial MFT, then relapsed, but responded to a 2nd or 3rd MFT also within 24 h, and 2) negative responders (NRs), where the cow was down for 24 h or longer after 2 or 3 standard MFTs. All the subjects in the negative response group were considered ADCs, because they remained in sternal recumbency, displaying a bright alert appearance and physiologically normal functions, except for their inability to stand for no apparent reason.

Pretreatment jugular blood samples were taken for determination of serum electrolyte levels by a provincial diagnostic laboratory. Randomization of sample analysis was impossible, because the herd owner was financially responsible for the treatment and laboratory costs, and decided which veterinary practice should treat the animal and if the sample should be analyzed. Most often, the blood was allowed to clot, and the serum was extracted and stored at 4°C until it was transported to the diagnostic laboratory. Electrolyte concentrations were measured by a colorimetric process (Kodak Ektachem DT 60 II analyzer; Eastman Kodak, Rochester, New York, USA). Normal reference values for interpretation were as follows: Ca (2.0–2.5 mmol/L), P (1.3–2.6 mmol/L), Mg (0.75–1.0 mmol/L), and K (3.9–5.8 mmol/L). A total of 86 complete serum analysis reports were obtained from Holstein-Friesian PRs (n = 50), and NRs, that is, ADCs (n = 36) that had not been medicated by the farmers or a veterinarian prior to the 1st examination. All cows received standardized care throughout the entire course of their condition and were treated for MF in accordance with the protocol.

Five independent variables (age, Ca, P, Mg, K) were examined. The non-Gaussian distribution of their values justified the use of nonparametric tests. A Wilcoxon rank-sum test compared the MFT response groups (PR and ADC) by using the variables’ medians. Correlations between age, Ca, and P were verified with Spearman’s rank correlation test (9). The epidemiologic unit in the study was the treated hypocalcaemic cow and the response was either that she became an ADC (treatment failure) or that she responded by returning to a standing position (success). Logistic regression analysis (LRA), using binomial univariable and multivariable maximum-likelihood model estimation, was performed to determine the association between the dependent variable (ADC or PR) and the independent variables (age, Ca, P), using statistical software (Stata/SE 8.0; Stata Corp LP, College Station, Texas, USA). The strength of the association was estimated by an odds ratio (OR) measure. For each of the 3 independent variables (age, Ca, P), a 4-level categorical variable was created by grouping the variable’s values by quartiles. Levels in each of the 4-level categorical variables were combined (collapsed) to reduce the number of levels (3 for P and 2 for Ca) in the regression model, if the variable level in the more saturated model had a Wald P-value greater than 0.05 and the Log-Likelihood Ratio test’s P-value was greater than 0.05 when it compared the likelihood value of the fuller model to that of the reduced model. The most parsimonial final model was selected, via backward elimination, with a Wald P-value of 0.05 as removal threshold, given an acceptable Log-Likelihood Ratio test value. Model fit was evaluated by Pearson’s and Hosmer-Lemeshow’s goodness-of-fit test (10,11). Traditional Mantel-Haenszel contingency table analysis was used to show the dose effect of serum P on ADC outcome and to provide a relative risk (RR) measure with positive (PPV) and negative (NPV) predictive values at different serum P levels.

Results

Excluded cases (n = 52) were conditions of neuromuscular (80%), alimentary (8%), toxic (8%), and metabolic (4%) origin. Treatment of uncomplicated MF cases (n = 477) resulted in 73 NRs, a 15.3% ADC outcome. The only major electrolyte abnormalities of the 86 serum samples analyzed involved Ca and P deficits. Comparisons between the PR and NR (ADC) groups showed that pretreatment level of serum P was the only 1 of 5 variables (age, Ca, P, Mg, K) examined by the Wilcoxon rank-sum test that had significantly different median values (P < 0.019) at the 0.05 level (Table 1). The Spearman rank correlation test was significant for Ca and P (rho = 0.49, P < 0.05), but not for the correlation between age and Ca or age and P (P > 0.05).

Table 1.

Median values of the variables [age, calcium (Ca), phosphorus (P), magnesium (Mg), potassium (K)] compared between the positive response (PR) and negative response (NR-ADC) groups in the study

Variables	Age (years)		Ca (mmol/L)		P (mmol/L)		Mg (mmol/L)		K (mmol/L)
Response	PR	ADC	PR	ADC	PR	ADC	PR	ADC	PR	ADC
Median	8.0	8.1	1.41	1.40	.85a	.50a	1.11	1.03	4.20	4.14
Cows (n)	50	36	50	36	50	36	50	36	50	36
P-value	0.387		0.378		0.019		0.260		0.326

ª Median values with superscripts are significantly different at the 0.05 level (Wilcoxon Rank-Sum Test)

Binary logistic regression models used in this study determined that there was a lack of confounding and that the interactions between the variables (age, Ca, P) were not statistically significant (Wald P > 0.05 and Log-Likelihood ratio P > 0.05), so they were eliminated from the models. In univariable models that examined each variable individually, both Ca and P values were statistically significant (P < 0.05), but age was not (P > 0.39). The categorical variables of importance for the final binary multivariable logistic model are presented in Table 2. It shows that the odds for a MF cow treated with standard hypocalcaemia therapy of becoming an ADC was negatively (negative coefficients) associated with pretreatment serum P levels and positively (positive coefficients) associated with pretreatment serum Ca levels. Indeed, a cow with a pretreatment P level of ≥ 0.9 mmol/L was 143 (1/OR) times less likely to become an ADC than if she had an initial P serum level of < 0.7 mmol/L. The likeliness of becoming an ADC at the medium level (0.9 > P ≥ 0.7 mmol/L) decreased 25-fold (143 vs 5.7). Conversely, a cow with a pretreatment serum Ca level of ≥ 1.7 mmol/L was 14 times more likely to become an ADC than if her initial Ca level was < 1.7 mmol/L. The Mantel-Haenszel analysis produced a simpler model, analogous to a binary-two-factor logistic regression, using a 2-level P category as the independent variable to predict MFT outcome. The effect of different cutoff levels of the pretreatment serum P values on the relative risk (RR) and also the positive and negative predictive values (PPV-NPV) for ADC outcome is shown in Table 3. A cow with a pretreatment serum P level < 0.9 mmol/L was 12 times more likely to become an ADC than if her initial P level was ≥ 0.9 mmol/L (P < 0.02, CI: 6.3, 23.1).

Table 2.

Results from the logistic regression model describing the associations between the dependent variable ADC (MFT failure) and the independent pretreatment reduced (P) phosphorus (n = 3) and (Ca) calcium (n = 2) levels of the study

Variable/Level	Coefficient	Odds Ratio	P-value	95% Confidence Intervals
Phosphorus (mmol/L)
(H) P ≥ 0.9	24.91	0.007	0.000	0.03–0.89
(M) 0.9 > P ≥ 0.7	21.75	0.174	0.036	0.0001–0.11
(L) P < 0.7	Referencea	1
Calcium (mmol/L)
(H) Ca ≥ 1.7	2.64	14	0.008	2.0–98
(L) Ca < 1.7	Referencea	1
Constantb	0.172		0.571
Correct predictions for the model: 73.2% (cutpoint is 0.5 or 50.0%)
Model’s PPV = 61.8% and NPV = 93.6% at the P ≥ 0.9 mmol/L levelc

^aComparison mode: High level compared to Low, and Medium compared to Low

^bIntercept for the complete model

^c(PPV) positive predictive value, (NPV) negative predictive value

ADC — alert downer cow

MFT — milk fever treatment

Table 3.

Mantel-Haenszel contingency table calculations of the relative risk and the predictive measures for alert downer cow (ADC) outcome at different cutoff levels of pretreatment serum levels of phosphorus (P) for cows with milk fever treated with standard commercial calcium preparations

Pretreatment Phosphorus (mmol/L)	Relative Risk (RR)	95% Confidence Intervals	Positive Predictive Value (PPV)	Negative Predictive Value (NPV)
0.9	12.04	6.3, 23.1	55%	95.5%
0.8	3.7	3.1, 4.40	57.4%	84.3%
0.7	3.13	2.2, 4.50	59.2%	81.1%
0.65	2.5	1.50, 4.1	57.4%	77%
Cutoff level 0.9 compares two groups: cows with phosphorus (P) levels ≥ 0.9 mmol/L and cows with levels < 0.9 mmol/L (same for other levels)
Correct predictions for the model at the 0.9 level: 65.1% (cutpoint is 0.5 or 50.0%)

Discussion

The results from the 86 animals that were part of 529 post-parturient recumbent Holstein-Friesian cows treated by the practice confirmed that both pretreatment Ca and P are important risk factors to the ADC outcome, but that Mg, K, and age were not. The multivariable regression analyses retained pre-treatment Ca levels, although the Wilcoxon rank-sum test was negative for statistical difference between the groups (Table 1). The multivariable models separated the Ca levels by response groups and pretreatment P levels, thereby including covariate effects (Table 2). These results agreed with the literature that Ca was important in the development of the DC syndrome, and of ADCs (12). Also, they confirm that a multivariable approach can be used to interpret a cow’s serum electrolyte profile after calving to better assess her health or, in this case, her response to MFT (13). Low Ca and high P cows are less likely to become ADCs and, as demonstrated in Table 3, the higher the P levels the better the prediction to not becoming an ADC. For a practitioner, the 2-level P model gave the best negative predictive value, 95.5%, to determine which animal will not become an ADC.

To our knowledge, this study is the first to report significant differences in pretreatment serum P levels between PRs and ADCs. Also, it is the first to demonstrate that hypophosphatemia is an important contributing factor to the ADC and to show a dose response relationship of P on MFT outcome (Table 3). Such a role had been alleged following blood examinations of MF cows treated by udder insufflation (14). Later, studies on MFT suggested that a rise in plasma P was necessary for recovery, but reports of favorable responses were anecdotal and not from controlled trials (4,5,15,16). An interesting study by Barlet and Davicco (17) did demonstrate that rapid recovery of ADCs requires a sustained increase in Ca, P, and 1-α hydroxycholecalciferol.

Few prospective studies reported in the veterinary literature have presented a detailed treatment protocol for MF cows, defined the MFT response groups as clearly, classified recumbent cattle as rigorously, and examined the pretreatment serum electrolytes of ADCs so broadly as has this study. Björsell et al (18) found no difference between responders and nonresponders, based solely on measures of Ca in serum. Waage (19), in a retrospective study, made the same observation without measuring P levels in plasma and reported lower Mg levels in plasma for DCs that were more frequently dystocic and, therefore, would have been excluded from this study. A normal serum Ca level is to be expected in cows with the downer syndrome caused by traumatic lesions in the absence of metabolic disorders (20). Close monitoring of cases in this study, as well as of those encountered subsequently, has consistently shown that normal Ca and P at the initial pretreatment serum evaluation can occur in cows suffering from unobserved trauma around parturition. This could explain the higher average Ca and P values for DCs in studies where we have noted possible classification errors (5,21). Differences in the electrolyte values between our response groups and those presented by Fenwick (22–24), where DCs were defined differently and 33% or more of the animals were down for 12 h or longer before 1st treatment, can be explained because the majority of our cows were seen early and considered mild MF cases not recumbent in excess of 10 h.

There were potential sources of bias in our study. The greatest proportion of blood samples came from animals in the nonresponder (ADC) group. Farmers who requested analyses for responders may have chosen animals with high pretreatment serum P levels, but we think this is improbable because the sampled cows were not clinically different from those not sampled and the only way to identify high or low P is to analyze the blood before treatment. Furthermore, subsequent to this study during which blood sampling was continued in the practice, no particular pattern of selecting for high or low P in either response group was observed. Some farmers could have been calling our clinic for a certain type of MF and others for another type, or farmers with high producing herds may have been choosing our practice to treat their MFs. This latter clientele had been educated to recognize the initial signs of MF and understood the necessity of requesting prompt veterinary care. Most often, cows were found in early stage II of MF with pre-treatment Ca and P levels for responders similar to those for the mild cases reported by Robertson (4). The majority of the herds that contributed data to the study had production levels at or above the provincial average (25,26) and the production level of most the of MF treated cows was between 7000 kg and 9000 kg of milk per year. Farm effects were not examined and may have biased our results. Although dry cow management and feeding practices varied from farm to farm, in our clinical opinion, there were no differences between farms that enabled us to anticipate one type of response more than another. Anionic diets were not being fed in any of the herds during the study period and cows that received injectable or oral products susceptible of altering the pretreatment serum electrolytes were excluded.

The limitations of this study were not having a greater proportion of serum samples from the group of responders and not having analyzed production records and farm management styles. Future studies are virtually impossible, but if attempted, they should include more animals, practices, and regions.

In conclusion, given the biases and the study limitations, we think our results are valid. Pretreatment P levels influence the response to MFT and the knowledge of this level can aid a practitioner in determining the appropriate treatment. CVJ