Short-term mortality prediction based on biochemical blood markers in azotemic cattle with renal disease

Abstract

Background

Renal disease in cattle is often suspected based on urinalysis and blood biochemical analysis, but prognostic biomarkers for survival are lacking.

Hypothesis/Objectives

Identify blood and urine biomarkers predicting mortality or discharge in azotemic cattle.

Animals

Thirty-four azotemic adult cattle referred to the Clinic for Cattle of the National Veterinary School of Toulouse.

Methods

In our prospective cohort study, renal disease was suspected based on clinical signs and azotemia (plasma creatinine concentration > 228 μmol/L and urea concentration > 5 mmol/L) and confirmed by urinalysis or ultrasonography. When death or euthanasia occurred, confirmation was obtained by gross and histologic examination. To construct a decision tree for short-term prognosis, additional biochemical and cytological biomarkers were measured in urine and blood samples collected at hospitalization. Outcomes were defined as favorable (discharge) or unfavorable (death). Correlation analyses, univariate and multivariate statistics, receiver operating characteristic curves, and decision tree modeling were performed.

Results

Thirteen cattle survived and 21 died or were euthanized. Mortality was associated with higher plasma symmetric dimethylarginine (SDMA), creatinine, and urea concentrations, as well as hypocalcemia, hypochloremia, and lower hematocrit. Creatinine, SDMA, chloride, and albumin were the most consistent predictors of outcome. A decision tree combining creatinine and SDMA achieved 87.5% accuracy. Animals with plasma creatinine concentration < 605 μmol/L and SDMA < 33 μg/dL had the highest probability of survival.

Conclusions and clinical importance

Plasma creatinine, urea, SDMA, chloride, and albumin concentrations are promising short-term prognostic biomarkers in azotemic cattle. Combined SDMA and creatinine assessment may assist clinical decision-making.

Introduction

In cattle clinical practice, renal disorders such as pyelonephritis,^1,2 interstitial nephritis,³ glomerulonephritis,⁴ amyloidosis,⁵ and acute tubular necrosis^6,7 are suspected based on urinalysis⁸ and measurement of blood creatinine or urea concentrations.⁹

Although plasma creatinine and urea concentrations in cattle are influenced by multiple factors, which can decrease their diagnostic performance, these surrogate biomarkers of renal function and particularly glomerular filtration have been shown to be valuable in the diagnosis and monitoring of chronic kidney disease (CKD) or acute kidney injury (AKI) in cats, dogs, and horses.^10–13 Indeed, the International Renal Interest Society (IRIS) has established grading systems based on the severity of renal lesions using biochemical variables such as blood creatinine concentration, symmetric dimethylarginine (SDMA) concentration, and urine protein-to-creatinine ratio (UPC), as well as clinical signs for CKD,¹⁴ and blood creatinine concentration for AKI.¹³

In dogs and cats, some biochemical variables also are used as prognostic indicators. In cats with CKD, serum creatinine and inorganic phosphorus concentrations,¹⁵ lower blood hemoglobin concentration,¹⁶ metabolic acidosis,¹⁷ and increased fibroblast growth factor 23 (FGF23) concentrations¹⁸ are predictive of survival. In cats with AKI, low red or white blood cell counts,¹⁹ higher SDMA concentration,²⁰ and hypocalcemia¹³ are observed more frequently in non-survivors. In dogs, renal biomarkers can be categorized into those indicative of decreased glomerular filtration rate, such as blood creatinine and SDMA concentrations,^21,22 and those reflecting renal damage, such as albumin concentration. Changes in blood SDMA and creatinine concentrations also can be used to monitor dogs with CKD.²³

In cattle, fewer biomarkers are routinely used for diagnostic purposes and their value as prognostic indicators has not been specifically evaluated. Regardless of the severity of azotemia, treatment generally is implemented in the field without an estimation of the likelihood of success. The objective of our study was to identify biomarkers capable of predicting the short-term success or failure of therapeutic management in cases of renal disease in cattle. We hypothesized that SDMA could be of value in predicting the outcome in affected animals and that a combination of selected urine and blood biomarkers could help to support clinical decision to treat or not.

Materials and methods

The protocol for our prospective study was approved by the “Science et Santé Animale” ethics committee (SSA-2025-011_V3). Additional methodological details are provided in the Appendix.

Animals

Adult beef and dairy cattle referred to the Clinic for Cattle of the National Veterinary School of Toulouse (ENVT) between September 2023 and October 2025 were prospectively enrolled, regardless of sex or breed.

Inclusion criteria were clinical signs compatible with renal disease (primarily anorexia, with or without weight loss or dullness), azotemia (plasma creatinine concentration > 228 μmol/L and urea concentration > 5 mmol/L), and urinary tract involvement (dysuria and reddish or cloudy urine) with proteinuria. Proteinuria was detected using urine dipstick analysis, Heller’s reaction, and UPC; only cattle with UPC > 0.3 were included.²⁴

Blood samples were collected from the jugular vein into lithium heparin tubes and analyzed within 1 h at the ENVT Clinical Pathology Laboratory using a Catalyst One Chemistry Analyzer (IDEXX Laboratories, France).

Clinical and biological follow-up, therapeutic management, and outcome criteria

Urine cytology, urinary tract ultrasonography, and necropsy examination with histopathology were performed to further characterize renal disease. Blood SDMA concentration, venous blood gases, inflammatory markers, hematocrit, and electrolytes were assessed at admission and at the end of hospitalization.

Hydration status was clinically evaluated at admission.²⁵ Supportive treatment consisted of IV isotonic saline adjusted according to dehydration and electrolyte or acid–base abnormalities. Potassium chloride supplementation was added in cases of hypokalemia. Antibiotic treatment (amoxicillin or benzylpenicillin) was initiated when bacteriuria or pyuria was detected.

Animals were euthanized in cases of clinical deterioration, or persistent urinary (bacteriuria, hematuria, or pyuria), systemic (anorexia, dullness, and diarrhea), or biochemical abnormalities (sustained azotemia or proteinuria) despite treatment. Cattle were discharged after complete clinical recovery (return of appetite) and normalization of azotemia with no relapse 24 h after fluid discontinuation.

Based on short-term outcome, cattle were classified into 2 groups: unfavorable outcome (death or euthanasia) and favorable outcome (discharge after hospitalization).

Statistics

Statistical analyses were performed using R software (version 4.5.2). Continuous variables were summarized as medians and IQRs. Associations between variables and outcome (death vs survival) were assessed using Kruskal–Wallis or Fisher’s exact tests, receiver operating characteristic (ROC) curve analysis, and likelihood ratios.²⁶ Variables with univariate P < .25 were included in multivariable analyses. Missing data were handled using multiple imputation.²⁷ Penalized logistic regression with elastic net regularization and recursive partitioning classification trees were used to identify predictors associated with short-term mortality.^28,29

Results

Clinical follow-up

A total of 34 adult cattle were included, comprising 32 cows and 2 bulls, with a mean age of 7.8 years (range, 3-13 years). The majority were beef breeds (31/34), with Blonde d’Aquitaine being the most common breed represented (15/34) and some others (7 Aubrac, 4 Charolaise, 2 Gasconne, 2 Wagyu, and 1 Bazadaise). Only 3 animals belonged to dairy breeds (2 Simmental, 1 Brown Swiss). Among these 34 animals, 13 were discharged and considered recovered, whereas 21 died (euthanasia, n = 18; natural death, n = 3; Table S1). Pyelonephritis was diagnosed in 79% of cases (27/34), of which 52% (14/27) died. The remaining renal diseases diagnosed were amyloidosis (9%; 3/34), tubular necrosis (9%; 3/34), and glomerulonephritis (3%; 1/34). Clinical signs at admission were anorexia or hyporexia (100%, 34/34); underweight or poor physical condition (82%, 28/34), lethargy and low responsiveness to external stimuli (47%, 16/34), diarrhea (24%, 8/34), and brisket or submandibular edema (9%, 3/34) (Table 1). The mean duration of clinical signs before admission, as reported by the referring veterinarian or owner, was 23.3 ± 27 days. For discharged cattle, the mean duration of hospitalization, duration of antimicrobial treatment administered in ENVT clinics, and total volume of fluid were 17 ± 6.9 days, 10 ± 4.9 days, and 192 ± 88 L, respectively. For cattle that died, the mean duration of hospitalization was 6 ± 5.5 days, with a total fluid treatment volume of 180 ± 92 L per animal.

Table 1.

Clinical signs observed in 34 azotemic animals according to renal disease and treatment performed.

Diagnosis (number)	Outcome			Symptoms						Duration of hospitalization (days): mean [min–max]	Antibiotics and fluid therapy (mean [min, max] and proportion treated/total)
	Discharge	Euthanasia	Natural death	Anorexia/hyporexia	Loss of condition	Reduced alertness and apathy	Red or cloudy urine	Oedema (brisket or submandibular)	Diarrhea
Pyelonephritis/interstitial nephritis (27)	13	14	0	27	24	10	26	1	2	UF: 4.7 [1-14] F: 17 [6-25]	Amoxicillin (Vetrimoxin): 10/27 Benzylpenicillin (Intramicine, Shotapen): 17/27 Fluid therapy: 179 L [20-500]
Tubular necrosis (3)	0	0	3	3	0	3	0	0	3	UF: 6 [4-7]	Benzylpenicillin (Intramicine, Shotapen): 2/3 Fluid therapy: 66 L [55-75]
Amyloidosis (3)	0	3	0	3	3	2	0	1	2	UF: 5 [3-7]	Fluid therapy: 0-80 L
Glomerulonephritis (1)	0	1	0	1	1	1	1	1	1	UF: 20	Benzylpenicillin (Intramicine, Shotapen): 1/1 Fluid therapy: 105 L
Total	13	18	3	34	28	16	27	3	8	–	–

Abbreviations: UF = unfavorable, F = favorable.

Histopathology, ultrasonography findings at admission, and diagnosis

Gross and microscopic lesions were evaluated in 62% (21/34) of cases that died or were euthanized. Histopathology was regarded as the examination that could most reliably establish the final diagnosis. Four conditions have been identified with pyelonephritis with or without interstitial nephritis being the most common (14/21), followed by tubular necrosis (3/21), amyloidosis (3/21), and glomerulonephritis (1/21). Interstitial nephritis was present only in association with pyelonephritis in 42% (6/14) of pyelonephritis cases. Only 1 of the 3 cows with amyloidosis and the 1 cow with glomerulonephritis exhibited clinical signs consistent with nephrotic syndrome (edema, severe hypoalbuminemia, and proteinuria; Table 1).

Kidney ultrasonography identified renal enlargement and dilatation of the renal calyces compatible with pyelonephritis in 79% (27/34) of cases. This suspicion was confirmed by histopathologic examination when performed on the necropsied animals (14/27 cases). No abnormalities could be observed for the remaining admitted cases, except 2 of the amyloidosis cases in which an increase in size of the kidneys was suspected.

For the 13 cows that recovered and were discharged back to their farms, the final diagnosis was pyelonephritis. This conclusion was based on a combination of the ultrasound findings (see above), gross clinical characteristics of the urine (pyuria or hematuria), biochemical and cytologic urine data (proteinuria, bacteriuria), despite the absence of histopathologic examination.

Blood and urine biomarkers

The most frequent blood electrolyte disorder was hyponatremia (Na ≤ 135 mmol/L) observed in 59% (20/34) of cattle, followed by hyperphosphatemia (Pi ≥ 2.8 mmol/L) in 38% (13/34), hypocalcemia (totCa ≤ 2 mmol/L) in 29% (10/34), and hyperkalemia (K ≥ 5.6 mmol/L) in 9% (3/34). Hyperproteinemia (>75 g/L) was observed in 56% (19/34) of animals, hypoproteinemia (≤67 g/L) in 9% (3/34), and hyperfibrinogenemia (>7 g/L) in 21% (7/34). The most frequent acid–base disturbance was metabolic alkalosis (pH > 7.45, bicarbonate > 29 mmol/L) in 17.6% (6/34) of cattle, followed by metabolic acidosis (pH < 7.35, bicarbonate < 19 mmol/L) in 12% (4/34) of cases.

Urinalysis identified USG < 1.020 in 85% (29/34) of cases, hematuria (cytology) in 79% (27/34), glucosuria in 44% (15/34), and alkaluria (pH > 8.7) in 41% (14/34). Cytology examination identified crystalluria in 35% (12/34) of cattle, pyuria in 59% (20/34) and casts in 24% (8/34).

In all cattle included in the study, UPC values and urinary protein concentrations markedly exceeded the upper reference limits (0.252 and 211 mg/L, respectively).²⁴

Biomarker correlations

All significant correlations (Figure 1) could be grouped into several physiopathological clusters. Strong positive associations were observed among several blood biomarkers, including creatinine, urea, SDMA, phosphate, and magnesium, and negative associations with ionized calcium and total calcium. Indeed, plasma creatinine concentration was strongly positively correlated with plasma urea (Spearman rho = 0.79, P < .00001), phosphate (rho = 0.63, P < .003) concentrations and moderately positively correlated with plasma SDMA (rho = 0.40, P = .04) and magnesium (rho = 0.63, P < .001) concentrations. Urea also showed a strong positive correlation with phosphate (rho = 0.80, P < 10⁻⁶). Ionized calcium concentration was moderately to strongly negatively correlated with urea (rho = −0.72, P < .001), phosphate (rho = −0.7, P = .004), creatinine (rho = −0.7, P = .02), and SDMA (rho = −0.56, P = .01) and positively correlated with chloride (rho = −0.70, P = .002). Total calcium concentration was moderately positively correlated with ionized calcium (rho = 0.59, P = .008), potassium (rho = 0.38, P = .04), and negatively correlated with phosphate (rho = −0.54, P = .004) and urea (rho = −0.40, P = .04).

Figure 1.

Correlation matrix of blood and urine biochemical analytes in 34 azotemic cattle. Abbreviations: B = blood, U = urine.

Furthermore, the relationship between proteinuria and glomerular injury was evident from the positive correlations between proteinuria and UPC, USG, and urea (rho = 0.8, P < .001; rho = 0.72, P = .01; and rho = 0.47, P = .003, respectively).

Urine variables also were interrelated: USG was moderately to strongly positively correlated with creatinuria (rho = 0.41, P = .01) and proteinuria (rho = 0.72, P = .01). Blood lactate concentration showed moderate positive correlations with urea (rho = 0.53, P = .02), creatinine (rho = 0.50, P = .01), proteinuria (rho = 0.62, P < .0001), and albumin (rho = 0.37, P = .04).

Hemoglobin and hematocrit were strongly positively correlated (rho = 0.9, P < 10⁻⁶). Total proteins were moderately to strongly positively correlated with albumin (rho = 0.68, P < .0001), fibrinogen (rho = 0.45, P = .03), and potassium (rho = 0.58, P < .001).

Identification of potential prognostic indicators of renal disorders

Based on univariate comparisons between cattle that died and survived and ROC analysis, the prognostic value of blood and urine biomarkers associated with kidney disorders is presented in Tables 2 and 3. Blood concentrations of SDMA, urea, and creatinine were 1.6- to 2.2-fold higher in animals that died (natural or euthanasia) compared with those that recovered (Table 2). In addition, plasma calcium (total and ionized) as well as chloride concentrations and hematocrit were significantly lower in non-surviving animals (on average 22%, 15%, and 27% lower, respectively). Plasma concentrations of SDMA, urea, creatinine, chloride, total calcium, and albumin were identified as significant prognostic indicators, with the highest area under the curve (AUC > 0.70; Table 3). At the selected thresholds, all variables except urea showed positive likelihood ratios (LR+) ranging from 3.04 to 6.21, indicating moderate to good prognostic performance for predicting death. Plasma ionized calcium, lactate, sodium, and potassium concentrations as well as UPC also exhibited LR+ values > 3 but were associated with poor negative likelihood ratios (LR− > 0.6). High sensitivity (Se > 0.9) for predicting death was achieved with plasma creatinine (≥605 μmol/L) and urea (≥25.3 mmol/L) concentrations, whereas high specificity (Sp > 0.85) was observed for plasma potassium (≥5.3 mmol/L), sodium (≤124 mmol/L), chloride (≤84 mmol/L), albumin (≤30 g/L), and lactate (≥3.5 mmol/L) concentrations.

Table 2.

Univariate analysis comparing blood and urine biomarkers according to animal outcome, presented as median and IQR; ^*: Reference intervals were derived from the reference intervals used at the clinical pathology laboratory (ENVT), from published data.^24,30,31

Biomarker		Standard ranges*	Natural death-euthanasia (median [IQR]) (n = 21)	Return to owner (median [IQR]) (n = 13)	P-value (Mann–Whitney test)
Blood	SDMA (μg/dL)	6-16	39.00 [33.00, 58.00]	24.00 [16.00, 32.75]	.01
	Hematocrit (%)	24-39	24.50 [21.25, 26.35]	33.50 [27.50, 35.22]	.05
	Fibrinogen (g/L)	<5	9.48 [6.67, 11.50]	7.89 [6.60, 9.89]	.7
	Total calcium (mmol/L)	2-2.5	1.80 [1.60, 2.05]	2.30 [1.95, 2.35]	.03
	Ionized calcium (mmol/L)	1-1.25	0.86 [0.70, 1.01]	1.11 [0.96, 1.18]	.04
	Urea (mmol/L)	0.3-3.1	45.00 [37.00, 61.80]	19.70 [14.80, 37.50]	.02
	Creatinine (μmol/L)	88-228	1078.00 [906.00, 1546.00]	532.00 [421.00, 596.00]	.01
	Magnesium (mmol/L)	0.8-1.2	1.00 [0.90, 1.40]	1.00 [0.80, 1.38]	.75
	Phosphate (mmol/L)	1.5-2.8	3.10 [2.30, 3.95]	2.70 [1.80, 3.90]	.52
	Total proteins (g/L)	67-75	85.00 [75.00, 94.00]	96.00 [90.00, 108.00]	.09
	Albumin (g/L)	28-34	30.50 [27.25, 35.00]	35.00 [32.50, 37.00]	.06
	L-Lactate (mmol/L)	<5	1.70 [1.20, 7.60]	1.00 [0.80, 1.40]	.17
	pH	7.35-7.45	7.44 [7.41, 7.45]	7.43 [7.36, 7.48]	.92
	Bicarbonate (mmol/L)	19-29	29.95 [27.33, 39.85]	31.80 [26.35, 38.90]	.81
	Hemoglobin (g/dL)	8.2-13	9.05 [8.75, 10.32]	11.05 [10.15, 11.95]	.14
	Sodium (mmol/L)	135-150	130.50 [119.70, 136.10]	130.75 [128.12, 136.00]	.37
	Potassium (mmol/L)	3.8-5.6	4.38 [3.50, 5.10]	4.42 [3.51, 4.77]	.83
	Chloride (mmol/L)	95-110	82.00 [74.80, 94.55]	96.55 [89.92, 102.98]	.02
	UPC	0.035-0.135	4.55 [2.50, 8.69]	2.66 [1.48, 3.42]	.26
Urine	Creatinine (mg/L)	458-3617	465.00 [402.00, 626.50]	567.50 [435.75, 863.75]	.36
	Proteins (mg/L)	67-211	2521.00 [1499.00, 5406.50]	1432.00 [1184.75, 2786.00]	.4
	Urine specific gravity	1.02-1.045	1.01 [1.01, 1.01]	1.01 [1.01, 1.02]	.7
	pH	7-8.7	8.50 [8.00, 9.00]	8.00 [7.65, 8.50]	.31

Abbreviations: ENVT = National Veterinary School of Toulouse, SDMA = symmetric dimethylarginine, UPC = urine protein-to-creatinine ratio.

Table 3.

Univariate analysis of the diagnostic performance of blood biomarkers in predicting outcome.

Parameters	Threshold	Sensitivity	Specificity	Youden index	LRPa	LRNb	AUC	P-value
Creatinine (μmol/L)	≥605	0.90 [0.70-0.99]	0.77 [0.46-0.95]	0.67	3.92	0.12	0.77	.0001
SDMA (μg/dL)	≥33	0.76 [0.53-0.92]	0.75 [0.44-0.94]	0.51	3.12	0.32	0.77	.006
Urea (mmol/L)	≥25.3	0.90 [0.70-0.99]	0.62 [0.32-0.86]	0.52	2.35	0.15	0.74	.002
Chloride (mmol/L)	≤84.5	0.63 [0.39-0.82]	0.89 [0.59-0.99]	0.51	6.21	0.42	0.72	.01
Total calcium (mmol/L)	≤1.85	0.66 [0.42-0.84]	0.77 [0.47-0.95]	0.43	3.04	0.45	0.70	.03
Albumin (g/L)	≤30	0.50 [0.28-0.72]	0.88 [0.59-0.99]	0.38	4.88	0.57	0.70	.04
Total proteins (g/L)	≤91	0.64 [0.41-0.83]	0.74 [0.44-0.94]	0.38	2.55	0.48	0.68	.05
Ionized calcium (mmol/L)	≤0.83	0.52 [0.29-0.73]	0.80 [0.5-0.96]	0.32	3.41	0.61	0.65	.13
Hematocrit (%)	≤25	0.59 [0.36-0.79]	0.74 [0.43-0.93]	0.33	2.42	0.56	0.64	.12
Hemoglobin (g/dL)	≤9.3	0.62 [0.38-0.81]	0.75 [0.45-0.94]	0.37	2.62	0.51	0.63	.08
L-Lactate (mmol/L)	≥3.5	0.44 [0.24-0.67]	0.86 [0.57-0.98]	0.30	3.71	0.64	0.62	.13
UPC	≥6.3	0.45 [0.24-0.68]	0.85 [0.55-0.98]	0.30	3.67	0.65	0.59	.14
Fibrinogen (g/L)	≥10.37	0.49 [0.27-0.71]	0.77 [0.46-0.95]	0.26	2.21	0.67	0.55	.19
Phosphate (mmol/L)	≥2.3	0.78 [0.55-0.93]	0.44 [0.18-0.73]	0.22	1.40	0.50	0.56	.23
Sodium (mmol/L)	≤124.4	0.44 [0.23-0.67]	0.90 [0.61-0.99]	0.34	4.74	0.62	0.55	.06
Potassium (mmol/L)	≥5.3	0.29 [0.12-0.53]	0.92 [0.63-1.00]	0.21	3.65	0.77	0.51	.19

^aLR+: positive likelihood ratio: indicates how much more likely a positive test result is in individuals that died compared to those that survived.

^bLR−: negative likelihood ratio: indicates how much less likely a negative test result is in individuals that died compared to those that survived.

Abbreviations: SDMA = symmetric dimethylarginine, UPC = urine protein-to-creatinine ratio.

Multivariate analysis

Results from penalized logistic regression models applied to imputed datasets enabled the identification of the most relevant variables associated with individual outcomes, while accounting for small sample size and collinearity among explanatory variables. The most consistently selected predictors across imputed datasets were SDMA (selection frequency = 1.00), chloride (1.00), albumin (0.98), and creatinine (0.90), indicating a strong and stable association with survival. Plasma urea (0.86), total protein (0.84), lactate (0.76), hematocrit (0.72), and potassium (0.72) concentrations showed moderate selection frequencies, suggesting a potentially relevant but less robust contribution. In contrast, UPC (0.58), ionized calcium (0.48), total calcium (0.42), and phosphate (0.36) concentrations were infrequently selected and therefore were considered less reliable predictors when accounting for the other variables in the model, likely reflecting their moderate to high correlation with other predictors including urea, creatinine, and SDMA.

Decision tree

Based on variable selection provided by multivariate models, the most relevant decision trees derived from this cattle cohort were constructed across all imputed datasets (Figure 2). The model integrated plasma creatinine and SDMA concentrations measured at admission as key biomarkers for predicting short-term clinical outcome (discharge versus natural death or euthanasia). The mean overall accuracy was 87.5% (95% CI, 71.7-96.3) representing significant improvement over the no-information rate (P = .001). The mean sensitivity and specificity of the decision tree were 1.00 (95% CI, 0.83-1.00) and 0.67, respectively (95% CI, 0.36-0.90).

Figure 2.

Decision tree for predicting outcome in cattle with renal azotemia. Outcomes were defined as discharge to owner after hospitalization and euthanasia or death. Thresholds correspond to plasma concentrations of biomarkers analyzed at admission. Abbreviation: B = blood.

Among animals with plasma creatinine concentration below 605 μmol/L, an SDMA concentration ≥ 33 μg/dL was associated with a 67% probability of death (representing 9% of the study, 3/34), whereas an SDMA < 33 μg/dL was associated with 100% discharged (26% of the population, 9/34). Among the 22 cattle with plasma creatinine concentration ≥ 605 μmol/L, 19 died (86.4%) and no additional relevant biomarkers improved the prediction of survival within this group.

Discussion

In our study, azotemia was reliably linked to renal involvement as confirmed by macroscopic lesions and histopathology in 21 of the 34 cases (61.7%). For the remaining 13 discharged animals, the final diagnosis of kidney injury can be considered reasonably reliable based on a combination of clinical, ultrasonographic, and urinalysis findings. However, the diagnostic accuracy is lower because of the absence of gross or histopathologic examination. Differentiation among pyelonephritis, interstitial nephritis, and other infectious renal diseases was not possible.

The clinical course of disease was highly variable. The duration of renal disease before admission was reported as predominantly subacute to chronic, with a median of 23.3 days (range, 2-90 days). However, the interval between the onset of clinical signs and admission could only be estimated, because of the often vague nature of early clinical signs, intensive breeding conditions (cow-calf herds), and reliance on the breeder’s observational capacity. It is likely that a small number of cases met the classic definition of chronic kidney disease (ie, irreversible progressive renal disease of more than 3 months duration with variable degrees of proteinuria) commonly used in veterinary medicine.^10,12,14,17 Nevertheless, distinguishing between CKD and AKI based on criteria routinely applied in humans or companion animals remains difficult and challenging in food animals.

Pyelonephritis and interstitial purulent nephritis was the most frequent diagnosis, accounting for 79.4% of cases (27/34), and is consistent with the recognized prevalence of this condition in cattle.³² Compared with previous studies on pyelonephritis in cattle,^1–3,33 several notable differences were observed, particularly regarding the frequency of hypokalemia, hyponatremia, and overall survival rate. These discrepancies may be attributable to differences in cohort definition and inclusion criteria. In previous studies,^1,2 only a proportion of cows with pyelonephritis were azotemic (27% and 53%, respectively), because of probable unilateral lesion. From a prognostic perspective, the assessment of blood test results therefore appears to be more relevant than assessing lesions using ultrasonography.

Given the small sample size, our findings should be interpreted with caution, because our model estimates are inherently unstable and subject to sampling variability. Consequently, our results are exploratory in nature and primarily serve to generate hypotheses rather than provide confirmatory evidence. Increased plasma concentrations of creatinine, urea, and SDMA, as well as hypocalcemia, hypochloremia, and hypoalbuminemia were associated with increased risk of death (natural or euthanasia) despite supportive treatment. Similar associations have been reported in other species and in cattle with unrelated diseases. In cats with chronic kidney disease, non-survivors tend to have higher creatinine, urea, and albumin concentrations than survivors.^12,15,17,19 In cattle, higher creatinine and urea concentrations and marked hypochloremia (mean, 82.5 mmol/L) have been described in animals that died after surgery for right abomasal displacement (mean serum creatinine concentration, 157.5 μmol/L; mean serum urea concentration, 15.9 mmol/L), compared with survivors (mean serum creatinine concentration, 91 μmol/L; mean serum urea concentration, 8 mmol/L).³⁴ Likewise, downer dairy cows with increased serum creatinine concentrations were less likely to survive than cows with normal results.³⁵ In cattle, serum or plasma creatinine concentration is affected by muscle mass, which varies according to breed, sex,³⁶ and myositis for downer cows whereas urea concentration depends on dietary factors, particularly crude protein intake.³⁷ For this reason, the decision tree established in our study should be applied exclusively to renal diseases and is not transferable to other conditions.

Other biomarkers previously identified as prognostic indicators or as useful for monitoring renal disease in cats and dogs were not confirmed in our study. These included hyperphosphatemia,^15,17 and increased UPC in cats with CKD.¹⁶ In cats with CKD, each 1-unit increase in serum phosphorus concentration has been associated with an 11.8% increase in the risk of death.¹⁵ In cattle, hyperphosphatemia also has been reported in non-surviving cows undergoing surgery for right abomasal displacement.³⁸ In our study, the lack of a significant difference between survivors and non-survivors is likely attributable to limited statistical power resulting from small sample size and substantial within-group variability in these variables. For instance, a UPC ≥ 6.3 was associated with a positive likelihood ratio (LR+) of 3.67 and acceptable specificity (Sp = 0.85) but showed low sensitivity (Se = 0.45). Although UPC was higher in non-surviving cows than in survivors, the difference was not significant (P = .26), and the statistical power of the study was low (0.37). A sample size approximately 3 times larger would be required to achieve a statistical power > 0.8 at a significance level of α = 0.05.

Our study, an increase in SDMA above 33 μg/dL appeared to be a useful predictive variable when associated with plasma creatinine concentration (≥605 μmol/L). These findings are consistent with the use of plasma SDMA or creatinine concentrations in monitoring disease progression and staging in dogs, cats, and horses.¹¹ In dogs and cats, a strong correlation between SDMA concentration and glomerular filtration rate (GFR) has been demonstrated.^39,40 In dogs with non-azotemic AKI, the area under the ROC curve of SDMA at first examination for predicting disease severity was 0.73 (95% CI, 0.51-0.95).²¹ In cattle, SDMA recently has been investigated in healthy animals^41,42 and a reference interval has been established.³¹ However, no correlation between SDMA and GFR has yet been demonstrated in this species. In cattle, specific GFR evaluation relies on a limited number of studies.^43–45

The strong positive correlations observed between plasma urea, creatinine, and SDMA concentrations, and both magnesium and phosphorus concentrations, as well as their negative correlation with calcium, are consistent with changes associated with chronic renal disorders in dogs, cats, or humans. These observed correlations could be explained by decreased GFR, which leads to retention of minerals such as phosphorus and magnesium and increased renal excretion of calcium in azotemic cattle. These alterations could be explained by the development of renal secondary hyperparathyroidism, resulting from the combined effects of phosphate retention and relative hypocalcemia, which would stimulate calcium-sensing receptors on parathyroid cells.^33,45 However, the exact mechanisms and their temporal sequence still need to be elucidated, particularly in relation to the primary cause, the specific renal lesion, and hormonal regulation.

Our study had some limitations that should be acknowledged when interpreting the results and extrapolating them to field conditions. First, the relatively small sample size (n = 34 cattle), case recruitment, and characteristics of the study population may limit external validity. Only adult cattle, predominantly females and of beef breeds, were included. The clinical course of disease was highly variable, often subacute to chronic, although it may have been shortened in some cases. Second, pyelonephritis with interstitial purulent nephritis was the predominant final diagnosis (27/34 cases) and apparent recovery (13/34 cases) was systematically associated with this condition. Moreover, this condition was very frequently observed in cases resulting in death or euthanasia (21/34 cases). Nevertheless, pyelonephritis with interstitial purulent nephritis is among the most common renal diseases in cattle,³² supporting the clinical relevance of the study population.

Additional limitations relate to the methodology. Diagnostic accuracy was unequivocal in cases of death or euthanasia, because confirmation relied on macroscopic and histopathologic findings. In contrast, in cases of apparent recovery, the diagnosis remains less certain despite the combined use of clinical examination, urinalysis, blood analyses, and ultrasonography. Finally, treatment selection and decisions regarding euthanasia were based on a standardized protocol. However, as in routine clinical practice, a limited degree of subjectivity cannot be entirely excluded.

Conclusion

Our study provided evidence for the potential use of short-term prognostic biomarkers in azotemic cattle under field conditions. Among the evaluated variables, plasma SDMA, creatinine, urea, chloride, and albumin concentrations emerged as the most consistent indicators associated with survival. A combined assessment of some of these biomarkers, particularly plasma SDMA (<33 μg/dL) and creatinine concentrations (<605 μmol/L), may assist practitioners in guiding decision-making regarding whether to initiate treatment in cattle with renal disease. However, these findings should be interpreted with caution because of the limited sample size and the exploratory nature of the statistical approach. Additional prospective studies in larger and more diverse populations are required to validate these biomarkers, clarify their pathophysiologic relevance, and improve their clinical applicability in nephrology of cattle.

Abbreviations

AKI

acute kidney injury

AUC

area under the curve

totCa

total calcium

CKD

chronic kidney disease

FGF23

fibroblast growth factor 23

GFR

glomerular filtration rate

IRIS

International Renal Interest Society

LR +

positive likelihood ratio

LR−

negative likelihood ratio

ROC

receiver operating characteristic

SDMA

symmetric dimethylarginine

UPC

urine protein-to-creatinine ratio

USG

urine specific gravity

Appendix

Measurement methods and laboratory procedures

Blood samples were centrifuged immediately at 500 g for 15 min (Gusto Mini centrifuge, Fisherbrand, Heathrow Scientific, IL, USA). Symmetric dimethylarginine and biochemical (total protein, fibrinogen, albumin, total and ionized calcium, phosphate, and magnesium) analyses were performed using the Catalyst One Chemistry Analyzer (IDEXX Laboratories, Montpellier, France). Blood gas and electrolyte analyses were performed using a Stat Profile Prime+ analyzer (Nova Biomedical, Waltham, MA, USA), according to manufacturers’ recommendations and laboratory internal quality-control procedures. These variables were systematically assessed at admission and at the end of hospitalization to evaluate treatment efficacy.

Azotemia thresholds (plasma creatinine concentration > 228 μmol/L and urea > 5 mmol/L) were based on reference intervals established at the ENVT Clinical Pathology Laboratory from 199 healthy cattle (50 female dairy cows and 149 beef cattle) originating from 119 herds located throughout western, central, and southwestern France, in accordance with American Society for Veterinary Clinical Pathology (ASVCP) recommendations⁴⁶ and using Reference Value Advisor freeware.⁴⁷ These thresholds are consistent with published data,³⁰ including a Belgian study³⁶ reporting breed-related differences in creatinine concentrations.

Proteinuria was initially detected using a dipstick test (Combur-Test, Cobas, Roche Diagnostics, France), confirmed by Heller’s reaction, and quantified by calculating the urine protein-to creatinine ratio (>0.25),²⁴ using pyrogallol red (U/CSF Protein, ThermoFisher Scientific, Waltham, MA, USA) and enzymatic methods (Creatinine Enzymatic, ThermoFisher Scientific), respectively, with an automated analyzer and manufacturer-provided reagents (Indiko, ThermoFisher Scientific).

Urine cytology and ultrasonography

Urine samples were collected by catheterization in cows or spontaneous urination in bulls. Urine sediment examination was performed within 2 h of collection by light microscopy (Nikon E200, Tokyo, Japan). Granular casts and crystals were evaluated on 10 low power fields, whereas epithelial cells, erythrocytes, and leukocytes were quantified on 10 high power fields. A semi-quantitative bacterial score (0-2) was assigned by a single operator (C.T.).

Ultrasonographic examination of the urinary tract was performed using a 5-MHz linear transducer (ExaPad, IMV Imaging, France). The left kidney and pelvic urinary tract were evaluated transrectally, whereas the right kidney was assessed transcutaneously through the right paralumbar fossa and last intercostal space.¹ Renal cortical thickness, renal diameter, and calyceal dilatation were specifically assessed. Renal cortical thickness and vertical renal diameter were measured to assess renal enlargement, and particular attention was paid to the renal calyces, especially in cases of suspected dilatation consistent with pyelonephritis.

Necropsy and histopathology

In case of natural death or euthanasia, necropsy was systematically performed, with detailed characterization of urinary tract lesions. Kidney samples were fixed in 10% buffered formalin, preferentially from macroscopically affected areas, processed using standard paraffin embedding and hematoxylin–eosin staining, and examined by light microscopy. Final histopathologic diagnoses were established by a board-certified veterinary pathologist (M.L.). No kidney biopsies were performed in live cattle for histopathologic examination.

Therapeutic management

During initial clinical examination at admission, hydration status was assessed based on eyeball recession and skin tent duration.²⁵ Supportive treatment consisted of IV fluid administration adjusted to the estimated degree of dehydration and initial electrolyte and acid–base status. Isotonic saline (0.9% NaCl) was administered at a rate of 20 mL/kg/h, corresponding to approximately 20-60 L per animal per day. In cases of hypokalemia, potassium chloride (KCl) was added to the fluids at a concentration of 1-2 g/L.²⁵ Antibiotic treatment (amoxicillin—15 mg/kg body weight every 48 h or benzylpenicillin—20 mg/kg body weight once daily) was initiated when bacteriuria or pyuria was confirmed by urinary cytology. Treatment regimens followed the marketing authorization of the respective commercial products. No bacteria isolation was performed.

Detailed statistical procedures

Spearman correlation coefficients were calculated to examine associations among continuous variables. Univariate comparisons between outcome groups (death vs survival at the end of hospitalization) were performed using Kruskal–Wallis tests for non-normally distributed continuous variables, and Fisher’s exact tests for categorial variables. For each continuous variable, ROC curves were generated, and the Youden J index was used to determine optimal thresholds. Sensitivity (Se), specificity (Sp), and positive and negative likelihood ratio (LR+ and LR−, respectively) as well as AUC values were calculated to quantify associations with the binary outcome. Positive and negative likelihood ratios express how much a test result above or below a given threshold increases or decreases the likelihood of the outcome (death) based on its distribution in affected and unaffected individuals.²⁶ Multivariable analysis then was performed after selection of variables with a univariate P-value < .25 to avoid excluding potentially relevant predictors. Missing data were handled using multiple imputation by predictive mean matching implemented in the mice package (version 3.19.0),²⁷ generating 50 imputed datasets. The imputation model included all predictors and the outcome (survival status) to preserve underlying relationships, and predictive mean matching was used to ensure plausible values for continuous variables. Penalized logistic regression with elastic net regularization (α = 0.5) was applied using the glmnet package²⁸ (version 4.1-10), to account for the high predictor-to-sample ratio and collinearity among variables. Within each imputed dataset, the optimal penalty parameter (λ) was selected using leave-one-out cross-validation, minimizing the binomial deviance. The modeling procedure was repeated across all imputed datasets, and variables were evaluated based on their frequency of selection, providing an assessment of the robustness of their association with the outcome variable. Given the penalized modeling framework and variable selection procedure, no conventional statistical inference (eg, P-values or confidence intervals) was performed, and results should be interpreted as exploratory. Finally, a multivariable classification tree was constructed on each imputed dataset, using recursive partitioning (rpart package)²⁹ to identify key predictors and interactions, with a minimum split of 5 observations, a complexity parameter of 0.01, and cross-validation across all observations. The mean overall accuracy, sensitivity, and specificity were computed across all trees.

Author contributions

Charly De Campos (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing—original draft, Writing—review & editing), Manon Goncalves-Brandão (Investigation), Emeline Cote (Investigation), Marie-Noëlle Lucas (Investigation), Fabien Corbière (Formal analysis, Software), Catherine Trumel (Investigation, Methodology, Project administration, Resources), and François Schelcher (Conceptualization, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing)

Conflicts of interest

The authors declare no conflicts of interest.

Funding

This work was partially supported by Ecoles Nationales Vétérinaires de France (ENVF) (Clinical Research projects—BIOREBOV) and by ENVT (Bonus Qualité Recherche projects).

Off-label antimicrobial declaration

The authors declare no off-label use of antimicrobials.

Institutional animal care and use committee or other approval declaration

The protocol was approved by the “Science et Santé Animale” ethics committee (SSA-2025-011_V3).

Human ethics approval declaration

The authors declare human ethics approval was not needed.