Frequency and diagnostic reliability of laboratory variables in cows with traumatic reticuloperitonitis and type 1, 2, 3, 4 and 5 abomasal ulcer


 Background: A number of laboratory abnormalities occur in cows with traumatic reticuloperitonitis (TRP) as well as in those with abomasal ulcers (AU). This prompted us to compare the frequencies of laboratory abnormalities of healthy cows and cows with TRP and cows with abomasal ulcers and to calculate diagnostic sensitivities and specificities, predictive values and positive likelihood ratios for laboratory findings. The study included 182 healthy control cows, 503 cows with TRP, 94 cows with U1, 145 cows with U2, 60 cows with U3, 87 cows with U4 and 14 cows with U5. Haematocrit, total leukocyte count, concentrations of total protein, fibrinogen, urea, potassium and chloride, base excess and rumen chloride concentration were analysed.Results: Values outside the reference interval occurred in 2 to 24% of control cows (rumen chloride 2%, urea 6%, serum chloride 11%, haematocrit 13%, base excess 18%, fibrinogen 20%, total protein 21%, total leukocyte count 22% and potassium 24%), which made differentiation of healthy and ill cows difficult. Therefore, the variables best suited for distinguishing healthy and ill cows were rumen chloride and blood urea concentration. This was also supported by an LR+ of 14 to 27 for rumen chloride >30 mmol/l and 6 to 15 for blood urea >6.5 mmol/l in cows with ulcers. Urea also had a high diagnostic specificity and like rumen chloride was suited for differentiation of healthy and diseased cows. The urea concentration was >8.5 mmol/l in only 0.5% of controls, and the LR+ for a urea concentration >8.5 mmol/l ranged from 11 in cows with TRP to 128 in cows with U2. Except for cows with TRP, azotaemia was significantly more frequent in ill cows than in controls. Cows with U2 (70%) had urea concentrations >8.5 significantly more often than cows of the other groups, which may have been prerenal azotaemia attributable to hypovolaemia. Even though the groups of ill cows differed significantly with respect to several variables, no variables were identified to reliably differentiate the various groups.Conclusions: Isolated results are not suitable to distinguish among groups of ill cows and instead the history, the clinical findings and results of additional diagnostic techniques such as ultrasonography are required.


Background
Traumatic reticuloperitonitis (TRP) and abomasal ulcers are important intestinal disorders of cattle [1,2]. Abomasal ulcers are classi ed as type 1 (U1), 2 (U2), 3 (U3), 4 (U4) or 5 (U5) [3][4][5][6][7]. Type 1 ulcer is non-perforating and accompanied by minimal haemorrhage, and is subdivided into subtypes 1a, 1b, 1c and 1d [8,9]. Type 2 ulcer is characterised by the erosion of a large blood vessel causing intraluminal haemorrhage and anaemia. Type 3 ulcer is perforating and accompanied by localised peritonitis, U4 is perforating and accompanied by generalised peritonitis and U5 has perforated into the omental bursa. Laboratory ndings in cows with TRP [10], U1 [11], U2 [12], U3 [13], U4 [14] and U5 [7] were recently described in detail. Laboratory ndings reported in older studies of cows with TRP and abomasal ulcer have also been discussed [7,10]. Of 94 cows with U1, the majority had hypokalaemia (68%), positive base excess (60%) and azotaemia (51%) [11], and of 145 cows with U2, the most frequent laboratory ndings were azotaemia (89%), low haematocrit (82%), hypokalaemia (81%), hypoproteinaemia (74%) and metabolic acidosis (61%) [12]. The only laboratory abnormality that was commonly seen in 60 cows with U3 was hypokalaemia (75%) [13]. Haemoconcentration (69%) and azotaemia (56%) [14] were the main abnormalities in 87 cows with U4, and hypokalaemia (71%), haemoconcentration (57%) and metabolic acidosis (57%) were the most common abnormalities in 14 cows with U5 [7]. Hyper brinogenaemia (69%) and hyperproteinaemia (64%) were the major laboratory abnormalities in 503 cows with TRP [10]. The degree of increase in rumen chloride concentration also varied among the different diseases. The abovementioned studies showed that hypokalaemia and azotaemia are common to all groups, whereas haemoconcentration is typically seen in cows with U4 and U5, haemodilution in cows with U2, hyperproteinaemia in cows with TRP and hypoproteinaemia in cows with U2. Furthermore, the frequency with which the laboratory abnormalities occur varied greatly among the groups. The prognosis also differed signi cantly and was usually good in cows with TRP, often fair in cows with U2, but generally poor in cows with U4 and U5. Finally, TRP and abomasal ulcer require different therapeutic approaches. The diagnostic reliability of clinical signs in cows with TPR and abomasal ulcers was recently investigated [15]. From a diagnostic standpoint, it would be desirable to use also laboratory variables to aid in differentiation of cows with TRP and various types of abomasal ulcers. To investigate this, the frequencies of laboratory abnormalities recorded in cows with TRP and abomasal ulcer were compared and the parameters diagnostic sensitivity and speci city, predictive values and positive likelihood ratio (LR + ) calculated.

Methods Animals
A total of 1,085 cows including 182 healthy controls, 503 cows with TRP, 94 cows with U1, 145 cows with U2, 60 cows with U3, 87 cows with U4 and 14 cows with U5 were used. All animals were privately owned and transported to the Veterinary Teaching Hospital of the University of Zurich for clinical examination. The sample size differed between groups and was dictated by the case load, which varied depending of the incidence of the disease in the population. The frequency and diagnostic reliability of the clinical ndings of these cows were the subject of a recent publication [15].
Tentative and de nitive diagnoses of TRP and type 1, 2, 3, 4 or 5 abomasal ulcer, treatment and response to treatment The criteria for tentative and con rmed diagnoses of TRP and abomasal ulcers, inclusion and exclusion criteria, and treatment and response to treatment of the cows in this study were recently published [15].

Statistical analysis
The program SPSS Version 25 was used for analysis. The Shapiro-Wilk test showed that only total protein and urea concentrations had normal distribution and therefor the medians and 25th and 75th percentiles were calculated. The medians underwent one-factor analysis of variance and pair-wise comparison using the Kruskal-Wallis test, and frequency distributions of all variables were calculated for the controls and six disease groups. Laboratory results were arbitrarily divided into appropriate ranges (see Table 2, example haematocrit; 27-37%, ≤ 20 and > 20%, < 27 and ≥ 27%, ≤ 37 and > 37%, ≤ 44 and > 44%), and the ranges compared among groups using the chi-square and the Bonferroni post-hoc tests. A P-value < accompanied by decreased (< 60 g/l), normal (60-85 g/l) and elevated total protein concentration (> 85 g/l), and the occurrence of decreased, normal and elevated total protein concentration accompanied by decreased (< 4 g/l), normal (4-7 g/l) and elevated (> 7 g/l) brinogen concentration were investigated.

Potassium concentration
The median potassium concentration ranged from 3.4 to 4.2 mmol/l (Table 1). Except for cows with U5, all cows from the other disease groups had signi cantly lower potassium concentrations than controls. Seventy-six percent of controls had a potassium concentration in the reference interval (Table 2, Fig. 6) and 23 and 0.5% had hypokalaemia with concentrations of < 4 and < 3 mmol/l, respectively. In cows of all disease groups, concentrations < 4 (68 to 81% of cows) and < 3 mmol/l (13 to 25% of cows) were signi cantly more common than in controls. The LR + for a potassium concentration < 3 mmol/l ranged from 23 to 46 in all disease groups (Table 3).

Base excess
The median base excess ranged from 0.3 to 3.6 mmol/l (Table 1). Eighty-two percent of controls had normal values of -4 to + 4 mmol/l ( Table 2, Fig. 8), 6% had values lower and 12% had values greater than the reference interval. A base excess smaller than − 4 and − 8 mmol/l was recorded in 6 and 0.5% of controls, respectively (Table 2); values lower than − 4 mmol/l were signi cantly more common in cows with U2 (21%) and U4 (18%), and values lower than − 8 mmol/l were signi cantly more frequent in cows with U2 (11%) than in controls. The LR + for a base excess lower than − 8 mmol/l was 21, 11 and 14 in cows with U2, U4 and U5, respectively (Table 3). A positive base excess greater than + 4 and + 7 mmol/l was signi cantly more frequent in cows with TRP, U1, U2, U3 and U4 (greater than + 4 mmol/l in 27 to 43%; greater than + 7 mmol/l in 14 to 28% including cows with U5) than in controls. The LR + for a base excess greater than + 7 mmol/l ranged from 26 to 54 for all groups (Table 3).

Rumen chloride
The median rumen chloride concentration ranged from 13 to 29 mmol/l ( Table 1). Except for cows with U5, all disease groups had signi cantly higher rumen chloride concentrations than controls. Ninety-eight percent of controls had rumen chloride concentrations in the reference interval ( Table 2, Fig. 9) and 2% had concentrations exceeding 30 mmol/l. The percentage of cows from all disease groups with concentrations > 30 mmol/l (13 to 44%) was signi cantly greater compared with controls. Except for TRP, the LR + for a rumen chloride concentration > 30 mmol/l was greater than 10 and ranged from 14 to 27 (Table 3).

Total protein concentration and haematocrit
The combined analysis of total protein concentration and haematocrit showed that 79% of cows with U2 had a decreased haematocrit, which was accompanied by hypoproteinaemia (30-59 g/l) in 73% and normoproteinaemia in 6% (Fig. 10A). In the other disease groups, a decreased haematocrit occurred in no more than 17% of the cows and was accompanied by normo-, hypo-or hyperproteinaemia. Most cows with a normal haematocrit had normoproteinaemia, some had hyperproteinaemia whereas hypoproteinaemia was rare (Fig. 10B). Most cows with increased haematocrit had normoproteinaemia and only rarely hypo-or hyperproteinaemia (Fig. 10C).
Total protein and brinogen concentrations Seventy-ve percent of cows with U2 had hypoproteinaemia, which was accompanied by normo brinogenaemia in 42%, hypo brinogenaemia in 32% and hyper brinogenaemia in 1% (Fig. 11A). Of the cows with normoproteinaemia (Fig. 11B), normo brinogenaemia was most common in controls and cows with U1, U2, U3 and U5, and hyper brinogenaemia was most common in cows with TRP and U4. Most of the controls and cows with U1 with hyperproteinaemia had normo brinogenaemia and most of the cows with TRP, U2, U3 and U4 with hyperproteinaemia had hyper brinogenaemia (Fig. 11C).

Discussion
Various studies have described the sensitivity, speci city and predictive value of total protein [19], total protein and brinogen [20] and brinogen, serum amyloid A and haptoglobin concentrations [21] for the diagnosis of cows with TRP; however, the results were compared with ndings in ill cows and not healthy controls. Therefore, the calculated diagnostic parameters differ greatly from ours and a direct comparison is not possible. The laboratory variables used for comparison of healthy cows and cows with TRP, U1, U2, U3, U4 and U5 were selected from a comprehensive panel that included haematological, serum biochemical, blood gas and rumen uid analyses and urinalysis. For a variable to be included in the study, it had to have been measured in the majority of cows and lie outside the reference interval in a large number of ill cows. Interestingly, numerous variables were outside the reference intervals in healthy cows including rumen chloride (2%), urea (6%), serum chloride (11%), haematocrit (13%), base excess (18%), brinogen (20%), total protein (21%), total leukocyte count (22%) and potassium (24%). This was despite the fact that the reference intervals of several variables had been modi ed and expanded. According to Constable and co-workers [22], healthy animals are assumed to have values within a certain range referred to as normal range or reference interval, and ill animals are assumed to have values outside this interval. A reference interval is established by using the results of a large number of healthy animals and usually includes 95% of the animals. Calculation of a reference interval is straightforward when the values have a normal distribution but requires transformation of non-normal data. In the present study, the variables had non-normal distributions with the exception of total protein and urea concentrations. Therefore, calculation of the reference interval assuming a normal distribution (mean ± 2 standard deviations) would result in an exaggerated interval that would include a considerable number of ill animals, erroneously considered healthy. When reference intervals published in the veterinary literature were used to assess healthy cows, the following values were outside the reference interval: haematocrit (41%), total protein (40%), serum chloride (26%), base excess (58%) and rumen chloride (64%). Therefore, the reference intervals of some variables were slightly expanded upward (total protein), downward (serum chloride and rumen chloride) or upward and downward (haematocrit, base excess). When healthy cows have values outside the reference intervals, they may be incorrectly diagnosed as ill (false positive). This problem can be mitigated by considering a combination of variables when making a diagnosis or to disregard certain values below the reference interval [22]. The extent of the deviation from the reference interval should also be considered because small deviations are less likely to re ect a disease than large ones. This can be quanti ed using the LR + , which expresses the association between a test result and the presence of a target disease [22]. The LR + expresses how much more likely an individual with the disease is to have a positive test result than if the individual is disease-free [23]. The LR + was calculated for all variables in the present study because it is an overall measure of the e ciency of a diagnostic test; it combines the diagnostic sensitivity and speci city and is not affected by the prevalence of the disease [22,23]. The latter characteristic was of particular importance in the present study because the exact prevalence of TRP and abomasal ulcers is not known.
Another drawback of using reference intervals is that not all diseased animals have values outside the interval. When a single variable with a value within the reference interval is used to make a diagnosis, a sick animal may be erroneously identi ed as not having the disease (false negative). This can be mitigated by shrinking the reference interval but in doing so the rate of false-positives increases. It would be better to consider groups of variables that all have diagnostic utility for the disorder in question. This was done in the present study by combining haematocrit and total protein, and total protein and brinogen concentrations.
Rumen chloride was the best variable for the differentiation of healthy and diseased cows because only 2% of controls had values outside the reference interval. This was also supported by an LR + of 14 to 27 in cows with ulcers. An LR + of 14 means that the likelihood of increased rumen chloride in an ill cow is 14 times more likely to occur than in a cow without an ulcer. The usefulness of rumen chloride for differentiating the different disease groups was limited. Only cows with TRP had rumen chloride concentrations < 40 mmol/l signi cantly less often (6%) than cows with U1 and U3. From a practical standpoint, this means that TRP is unlikely in a cow with a marked increase in rumen chloride concentration.
Although of slightly less importance than rumen chloride, urea had a high diagnostic speci city and was well suited for differentiation of healthy and diseased cows. The urea concentration was > 6.5 mmol/l in only 6% and > 8.5 mmol/l in only 0.5% of controls, and the LR + for a urea concentration > 8.5 mmol/l ranged from 11 in cows with TRP to 128 in cows with U2. Except for cows with TRP, azotaemia was signi cantly more frequent in ill cows than in controls. Cows with U2 (70%) had urea concentrations > 8.5 signi cantly more often than cows of the other groups, which may have been prerenal azotaemia attributable to hypovolaemia. Analogous to increased rumen chloride, an increase in urea concentration (above 8.5 mmol/l) is rare (6%) in cows with TRP because this is a localised disease with no direct systemic effects.
Even though 13% of controls had a haematocrit outside the expanded reference interval of 27 to 37%, none had a haematocrit > 44%. Likewise, no more than 5% of cows with TRP, U1, U2 and U3 had values > 44%, which was in contrast to cows with U4, in which 33% had a haematocrit > 44%.
Thus, in cows with a high haematocrit, TRP or type 1, 2, 3 or 5 abomasal ulcers are unlikely. A haematocrit ≤ 20% occurred in 70% of cows with U2 and in only 0 to 7% of cows in the other groups and thus allowed differentiation of cows with U2.
Leukocytosis was a poor criterion for differentiating healthy and diseased cows and among cows of the different disease groups. The utility of leukopenia for this purpose was similar except that a total leukocyte count < 5000/µl was seen in 35% of cows with U4 but in only 4 to 14% of cows of the other groups. Persistent leukopenia is always a serious nding. Leukopenia in cows with U4 can be interpreted as a sequel to sepsis and leukocyte consumption [14]. Bone marrow production of leukocytes is insu cient in the face of severe in ammation because mobilisation of the marrow stem cell pool is slower in cattle than in other species [24].
A potassium concentration < 3 mmol/l is almost always a sign of disease because it occurred in only 0.5% of controls. However, hypokalaemia was non-speci c and all disease groups were affected. The principal cause of hypokalaemia is decreased forage intake associated with illness [25].
A total protein concentration > 85 g/l was measured in 45% cows with TRP, which was signi cantly more frequent than in cows of the other groups.

Conclusions
Depending on the variable, 2 to 24% of healthy cows had values outside the reference intervals, which made the differentiation of healthy and diseased cows di cult based on laboratory variables alone. Rumen chloride and urea concentrations provided the best selectivity because they were outside the reference intervals in only 2 and 6% of healthy cows, respectively. Although there were signi cant differences between groups of cows with respect to several laboratory variables, there were no variables that reliably differentiated cows with different diseases. Instead of using a speci c laboratory variable for diagnosis, a combination of multiple variables must be considered along with the history and clinical ndings and often the results of additional techniques such as ultrasonography.