Materials and Methods

Horses—Horses admitted to the University of California's William R. Pritchard Veterinary Medical Teaching Hospital between 2002 and 2005 because of diarrhea, regardless of severity, and horses that developed diarrhea while hospitalized at the veterinary medical teaching hospital during the same period were eligible for inclusion in the study. Horses were enrolled prospectively in the study if they met the inclusion criteria.

For all horses included in the study, fresh fecal samples were collected at the time of admission (horses admitted because of diarrhea) or at the onset of diarrhea (horses that developed diarrhea while hospitalized) and submitted for microbial culture for C difficile and testing for C difficile antigen and toxin A, as described.¹³ Samples were refrigerated following collection, and all samples were processed within 12 hours after collection. For microbial culture, fecal samples were plated directly on cycloserine-cefoxitin fructose agar plates and were inoculated into cycloserine-cefoxitin fructose enrichment broth and subcultured on cycloserine-cefoxitin fructose agar plates after 24 to 48 hours of broth enrichment. Colonies were identified as C difficile on the basis of colony morphology, Gram-staining characteristics, and production of L-proline aminopeptidase, as described.^4,14 A commercial ELISA^b was used to test fecal samples for C difficile antigen and toxin A. Assays were performed in accordance with the manufacturer's instructions. The ELISA purportedly detects both the glutamate dehydrogenase antigen of C difficile and toxin A; for both the C difficile antigen and toxin A, the lower limit of sensitivity has been reported by the manufacturer to be 2 ng/mL. Fecal samples were also plated on xylose-lysine-tergitol 4 agar plates and inoculated in selenite enrichment broth for detection of Salmonella spp.

Horses were assigned to 1 of 4 groups on the basis of assay results and results of microbial culture. Group 1 consisted of horses for which ELISA results for both C difficile antigen and toxin A were positive and results of microbial culture for C difficile were positive or negative. Group 2 consisted of horses for which ELISA results for C difficile antigen were positive, ELISA results for toxin A were negative, and results of microbial culture were positive or negative. Group 3 consisted of horses for which ELISA results for both C difficile antigen and toxin A were negative and results of microbial culture were positive. Group 4 consisted of horses for which ELISA results for both C difficile antigen and toxin A were negative and results of microbial culture were negative.

Medical records review—Medical records of horses included in the study were reviewed, and information was obtained on signalment (age, sex, and breed); results of hematologic and serum biochemical testing, heart rate, respiratory rate, and rectal temperature at the time of admission or the onset of diarrhea; duration of hospitalization and duration of antimicrobial administration prior to the onset of diarrhea; total hospitalization time; and cost of hospitalization.

Statistical analysis—Categoric data were summarized as proportions. Continuous data were examined with the Kolmogorov-Smirnov test to determine whether they were normally distributed. Data that were normally distributed were summarized as mean and SD; data that were not normally distributed were summarized as median and range.

The χ² test of homogeneity was used to compare distributions for outcome (died or euthanized vs survived to discharge), fecal consistency (cow pie vs loose or watery), and sex among the 4 groups of horses. Kruskal-Wallis ANOVA followed by the Dunn multiple comparison test (for data that were not normally distributed) or 1-way ANOVA followed by the Tukey post test (for data that were normally distributed) was used to compare total hospitalization time, cost of hospitalization, age, results of hematologic and serum biochemical testing, and duration of antimicrobial administration prior to the onset of diarrhea among the 4 groups. Analyses were subsequently repeated with horses in groups 2 and 3 combined into a single group. Variables for which the F² test or ANOVA yielded a value of P < 0.20 were examined by means of univariate logistic regression comparing horses in group 1 with horses in group 4, horses in group 1 with horses in group 2 or 3, and horses in group 2 or 3 with horses in group 4. Variables for which the P value for the univariate logistic regression was < 0.02 were included in a multiple logistic regression model for each of the 3 comparisons. A backward elimination strategy was used, with factors for which the P value was > 0.10 removed. Interactions between variables in the final models were assessed by means of likelihood ratio tests. Results are presented as ORs and their 95% CIs.

Sensitivity and specificity of the ELISA for detecting C difficile antigen in feces were calculated by comparison with results of microbial culture of fecal samples and with results of microbial culture following broth enrichment. Agreement between results of the assay for C difficile antigen in feces and results of microbial culture was also assessed through calculation of the Cohen N statistic.

All analyses were performed with standard software.^c Values of P < 0.05 were considered significant.

Results

Horses—A total of 292 horses met the criteria for inclusion in the study. Of these, 33 were assigned to group 1 (positive for both C difficile antigen and toxin A), 38 were assigned to group 2 (positive for C difficile antigen but negative for toxin A), 29 were assigned to group 3 (negative for both C difficile antigen and toxin A but results of microbial culture were positive), and 192 were assigned to group 4 (negative for both C difficile antigen and toxin A and negative for results of microbial culture). Eighteen breeds were represented; the 4 most common breeds were Thoroughbred (n = 71), Quarter Horse (62), Arabian (41), and Warmblood (33).

Age ranged from 1 day to 30 years and did not differ significantly among the 4 study groups. Eleven of the 33 (33%) group 1 horses, 15 of the 38 (39%) group 2 horses, 6 of the 29 (21%) group 3 horses, and 29 of the 192 (15%) group 4 horses were < 6 months old. Sex distribution did not differ significantly among the 4 study groups.

Clinical conditions included colic requiring surgical treatment (n = 84), colic requiring medical treatment (43), primary diarrhea without an identified underlying disease (46), neonatal disease (ie, sepsis, peripartum asphyxia, dysmaturity, and uroperitoneum; 44), orthopedic disorders (9), and lower airway disease or pneumonia (12). The remaining 54 horses had other conditions. None of the horses with primary diarrhea were assigned to group 1, in that all horses in group 1 had been treated with antimicrobials prior to the onset of diarrhea, except 1 neonatal foal.

Fecal consistency ranged from cow pie consistency to loose or watery. Twenty-nine of 33 (88%) horses in group 1, 35 of 38 (92%) horses in group 2, 23 of 29 (79%) horses in group 3, and 154 of 192 (80%) horses in group 4 had loose or watery feces. Overall, 241 of the 292 (82.5%) horses had loose or watery feces, although fecal consistency in individual horses was variable over time, ranging from cow pie consistency to watery on different days in 27 of these 241 (11.2%) horses. In contrast, 42 of 292 (14.4%) horses only ever had feces of cow pie consistency, with 4 of 33 (12%) of the horses in group 1, 2 of 38 (5%) horses in group 2, 6 of 29 (21%) horses in group 3, and 30 of 192 (16%) horses in group 4 only ever having feces of cow pie consistency. Eight of the 292 horses had feces described as loose with oil, and 1 horse was simply described as having diarrhea without further specification. Distribution of fecal consistency did not differ significantly among the 4 study groups.

Diarrhea was acute (≤ 7 days in duration prior to sample collection) in 284 of the 292 (97.3%) horses. One horse in group 2 and 5 horses in group 4 had chronic diarrhea (t 1 month in duration prior to entry into the study). One additional horse in group 4 had intermittent diarrhea for 11 days, and duration of diarrhea was not reported for 1 horse in group 3.

Clinical and clinicopathologic findings—Body temperature was significantly higher in group 1 horses (ie, horses positive for toxin A; mean ± SD, 38.4 ± 0.9°C [101.2 ± 1.6°F]), compared with group 4 horses (ie, horses without evidence of C difficile infection; 38.0 ± 0.6°C [100.4 ± 1.1°F]; P < 0.05) and horses in groups 2 and 3 combined (ie, horses with evidence of C difficile infection but negative for toxin A; 38.1 ± 0.6°C [100.6 ± 1.0°F]; P < 0.05). Band neutrophil count (median, 631.5 band neutrophils/PL [range, 0 to 3,577 band neutrophils/PL] vs 0.0 band neutrophils/PL [0 to 3,167 band neutrophils/PL]; P < 0.001), Hct (median, 36.3% [range, 23.4% to 65.9%] vs 30.9% [13.5% to 71.7%]; P < 0.01), and hemoglobin concentration (median, 13.1 g/dL [range, 9.4 to 25.5 g/dL] vs 12.1 g/dL [5.1 to 26.2 g/dL]; P < 0.05) were significantly higher in group 1 horses than in group 4 horses, and band neutrophil count was significantly (P < 0.05) higher in group 1 horses than in horses in groups 2 and 3 combined (median, 15.5 band neutrophils/PL [range, 0 to 2,999 band neutrophils/PL]). Plasma calcium concentration was significantly lower in group 1 horses than in group 4 horses (mean ± SD, 10.7 ± 1.0 g/dL vs 11.4 ± 0.9 g/ dL; P < 0.05), group 3 horses (ie, horses from which C difficile was cultured that were negative for antigen and toxin A; 11.6 ± 1.0 g/dL; P < 0.05), and horses in groups 2 and 3 combined (11.5 ± 1.1 g/dL; P < 0.01). Fibrinogen concentration was significantly (P < 0.05) higher in group 2 horses (ie, horses positive for C difficile antigen but negative for toxin A) than in group 3 horses (median, 500 mg/dL [range, 200 to 900 mg/dL] vs 400 mg/dL [100 to 500 mg/dL]), and icteric index (P < 0.05) and fibrinogen concentration (median, 500 mg/dL [range, 200 to 900 mg/dL] vs 300 mg/dL [100 to 1,300 mg/dL]; P < 0.01) were significantly higher in group 2 horses than in group 4 horses. Horses in groups 2 and 3 combined had a significantly (P < 0.001) higher icteric index than did horses in group 4. White blood cell count did not differ significantly (P = 0.12) among the 4 study groups.

Hospitalization time and cost—Median duration of hospitalization prior to the onset of diarrhea was 4 days for horses in group 1, 1 day for horses in group 2, 2 days for horses in group 3, and 1 day for horses in group 4. Duration of hospitalization prior to the onset of diarrhea was significantly longer for horses in group 1 than for horses in groups 2 (P < 0.01) and 4 (P < 0.001). Five of 33 (15%) group 1 horses, 11 of 38 (29%) group 2 horses, 8 of 29 (28%) group 3 horses, and 52 of 192 (27%) group 4 horses were admitted for treatment of diarrhea. The remaining 28 (85%) horses in group 1, 27 (71%) horses in group 2, 21 (72%) horses in group 3, and 140 (73%) horses in groups 4 developed diarrhea while hospitalized. Median total hospitalization time was 10 days for horses in group 1, 8 days for horses in group 2, 7 days for horses in group 3, and 6 days for horses in group 4. Total hospitalization time was significantly longer for horses in group 1 than for horses in groups 2 (P < 0.05), 3 (P < 0.05), and 4 (P < 0.001). For horses that survived, median hospitalization time after development of diarrhea was 6 days for group 1 horses, 5 days for group 2 horses, 4 days for group 3 horses, and 4 days for group 4 horses; hospitalization time for surviving horses after development of diarrhea did not differ significantly among study groups. Cost of hospitalization was significantly (P < 0.01) higher for group 1 horses than for group 4 horses.

Total hospitalization time for surviving horses and hospitalization time prior to development of diarrhea for all horses were significantly (P < 0.01) longer for horses in group 1 than for horses in groups 2 and 3 combined. Hospitalization cost (P < 0.01) and total hospitalization time (P < 0.05) were significantly longer for horses in groups 2 and 3 combined than for horses in group 4.

Antimicrobial use—Thirty-two of the 33 (97%) horses in group 1 had received antimicrobials prior to the onset of diarrhea. By comparison, 30 of 38 (79%) horses in group 2, 19 of 29 (66%) horses in group 3, and 92 of 192 (48%) horses in group 4 had received antimicrobials prior to the onset of diarrhea. In each group, a wide variety of antimicrobials had been administered, and no single antimicrobial predominated. However, metronidazole had been administered to more horses in group 1 (14/33 [42%]) than to horses in group 2 (8/38 [21%]), 3 (4/29 [14%]), or 4 (11/192 [6%]), and horses in group 1 were more likely (OR = 8.3; 95% CI, 3.7 to 18.7; P < 0.001) to have received metronidazole prior to the onset of diarrhea as were horses in the other 3 groups.

Median duration of antimicrobial administration prior to the onset of diarrhea was 4 days for horses in group 1, 3 days for horses in group 2, 2.5 days for horses in group 3, and 1 day for horses in group 4 and was significantly (P < 0.01) longer for horses in group 1 than for horses in group 4. Antimicrobial administration had been discontinued in 8 (24%) horses in group 1, 4 (11%) horses in group 2, 2 (7%) horses in group 3, and 12 (6%) horses in group 4 immediately prior to the onset of diarrhea.

Results of multivariate analysis—Factors that were retained in the final multivariate logistic regression model comparing groups 1 and 4 included duration of hospitalization prior to the onset of diarrhea, duration of antimicrobial administration prior to the onset of diarrhea, the interaction between duration of hospitalization and duration of antimicrobial administration prior to the onset of diarrhea, band neutrophil count, total bilirubin concentration, rectal temperature, total hospitalization time for horses that survived, and total hospitalization cost.

The likelihood that toxin A would be detected (ie, that a horse would be assigned to group 1 vs group 4) increased significantly as duration of hospitalization prior to the onset of diarrhea increased (OR of being in group 1 for every 1-day increase in hospitalization time = 1.30; 95% CI, 1.10 to 1.54; P = 0.002). However, the likelihood that toxin A would be detected did not differ significantly as duration of antimicrobial administration prior to the onset of diarrhea increased (OR of being in group 1 for every 1-day increase in antimicrobial administration time = 1.13; 95% CI, 0.99 to 1.30; P = 0.07). The significant (P = 0.019) interaction between duration of hospitalization and duration of antimicrobial administration indicated that the increase in likelihood that toxin A would be detected, associated with an increase in duration of 1 factor, was even greater as duration of the other factor increased.

The likelihood that toxin A would be detected increased significantly as band neutrophil count increased (OR of a horse being in group 1 for every increase in band neutrophil count of 100 band neutrophils/PL = 1.12; 95% CI, 1.03 to 1.21; P = 0.005). However, the likelihood that toxin A would be detected did not differ significantly as total bilirubin concentration increased (OR of a horse being in group 1 for every increase in total bilirubin concentration of 1 mg/dL = 1.29; 95% CI, 0.99 to 1.67; P = 0.058).

The likelihood that toxin A would be detected increased significantly as rectal temperature (OR of a horse being in group 1 for each 1°F increase in rectal temperature = 1.94; 95% CI, 1.28 to 2.92; P = 0.002) and total hospitalization time for surviving horses (OR of a horse being in group 1 for each 1-day increase in total hospitalization time = 1.16; 95% CI, 1.03 to 1.34; P = 0.010) increased. However, the likelihood that toxin A would be detected did not differ significantly as total hospitalization cost increased (OR of a horse being in group 1 for every $500 increase in total hospitalization cost = 1.09; 95% CI, 0.99 to 1.19; P = 0.067).

Factors that were retained in the final multivariate logistic regression model comparing group 1 with groups 2 and 3 combined included duration of hospitalization prior to the onset of diarrhea, band neutrophil count, calcium concentration, total hospitalization time for horses that survived, and rectal temperature.

For horses with evidence of C difficile infection, the likelihood that toxin A would be detected (ie, that a horse would be assigned to group 1 as opposed to group 2 or 3) increased significantly as duration of hospitalization increased prior to the onset of diarrhea (OR of a horse being in group 1 as opposed to group 2 or 3 for each 1-day increase in hospitalization time = 1.39; 95% CI, 1.15 to 1.68; P < 0.001). Likelihood that toxin A would be detected did not differ significantly as band neutrophil count (OR of a horse being in group 1 as opposed to group 2 or 3 for every increase in band neutrophil count of 100 band neutrophils/PL = 1.06; 95% CI, 1.00 to 1.14; P = 0.07) or calcium concentration (OR of a horse being in group 1 as opposed to group 2 or 3 for every increase in calcium concentration of 1 mg/dL = 0.68; 95% CI, 0.44 to 1.05; P = 0.08) changed. The likelihood that toxin A would be detected increased significantly as total hospitalization time for horses that survived (OR of a horse being in group 1 as opposed to group 2 or 3 for each 1-day increase in total hospitalization time = 1.32; 95% CI, 1.10 to 1.57; P = 0.002) and rectal temperature (OR of a horse being in group 1 as opposed to group 2 or 3 for each 1°F increase in rectal temperature = 1.87; 95% CI, 1.12 to 3.11; P = 0.02) increased.

Factors that were retained in the final multivariate logistic regression model comparing groups 2 and 3 combined with group 4 included total bilirubin concentration, potassium concentration, heart rate, and total hospitalization cost.

The likelihood of C difficile infection without detection of toxin A (ie, that a horse would be assigned to group 2 or 3 vs group 4) did not differ significantly as total bilirubin concentration (OR of a horse being assigned to group 2 or 3 as opposed to group 4 for every increase in total bilirubin concentration of 1 mg/dL = 1.19; 95% CI, 0.98 to 1.44; P = 0.07) or potassium concentration (OR of a horse being assigned to group 2 or 3 as opposed to group 4 for an increase in potassium concentration of 1 mEq/L = 1.50; 95% CI, 0.94 to 2.40; P = 0.09) increased. However, the likelihood of being in group 2 or 3 as opposed to group 4 increased significantly as heart rate (OR of a horse being in group 2 or 3 as opposed to group 4 for every increase in heart rate of 10 beats/min = 1.29; 95% CI, 1.14 to 1.46; P < 0.001) and total hospitalization cost (OR of a horse being in group 2 or 3 as opposed to group 4 for every $500 increase in total hospitalization cost = 1.12; 95% CI, 1.05 to 1.18; P = 0.001) increased.

Outcome—Eight of 33 (24%) horses in group 1, 6 of 38 (16%) horses in group 2, 4 of 29 (14%) horses in group 3, and 23 of 192 (12%) horses in group 4 were euthanized, and 1 horse in group 2 died. Outcome did not differ significantly (P = 0.25) among individual groups. Review of the medical records suggested that 34 horses were euthanized because of a perceived poor prognosis for recovery and that only 5 horses, all in group 4, were euthanized because of financial concerns. Reasons for euthanasia could not be determined for 1 horse in group 2 and 1 horse in group 4. Even when the 5 horses euthanized for financial reasons and the 2 horses for which reasons for euthanasia could not be determined were excluded, there was still no significant (P = 0.1) difference in outcome among groups.

Mortality rate for horses in group 1 (8/33 [24%]) was significantly (P = 0.04) higher than mortality rate for horses in the other 3 groups (26/252 [10%]), when horses euthanized for financial or unknown reasons were excluded. Horses in group 1 were significantly more likely to die or be euthanized than were horses in the other 3 groups combined (OR, 2.78; 95% CI, 1.14 to 6.80; P = 0.04). On the other hand, mortality rate for horses in groups 1, 2, and 3 combined (ie, horses with any evidence of C difficile infection; 17/100 [17%]) was not significantly (P = 0.055) different from mortality rate for horses in group 4 (ie, horses without evidence of C difficile infection; 17/186 [9%]), when horses euthanized for financial or unknown reasons were excluded.

Reasons for euthanasia of horses in group 1 included colitis or complications of colitis (ie, disseminated intravascular coagulation, laminitis, or renal disease; n = 6), severe hypoxic or ischemic disease in the foal (1), and refractory colic (1). Reasons for euthanasia of horses in group 2 included colitis or complications of colitis (n = 2), postcastration evisceration (1), sepsis (1), gastric rupture (1), and peritonitis (1); the horse that died had complications of colitis. Reasons for euthanasia of horses in group 3 included colitis or complications of colitis (n = 3) and neoplasia (1). Reasons for euthanasia of horses in group 4 included colitis or complications of colitis (n = 9), refractory or persistent colic (6), peritonitis (3), neoplasia (2), respiratory distress (1), an abdominal abscess (1), and unknown reason (1).

Sensitivity and specificity of the assay for C difficile antigen—For 25 of the 33 (75%) horses in group 1, results of microbial culture of fecal samples were positive following direct plating and following enrichment in broth; for 4 (12%) horses, results of microbial culture were negative following direct plating but positive following enrichment in broth; and for the remaining 12 (12%) horses, results of microbial culture were negative following direct plating and following enrichment in broth. For 15 of the 38 (39%) horses in group 2, results of microbial culture of fecal samples were positive following direct plating and following enrichment in broth; for 12 (31%) horses, results of microbial culture were negative following direct plating but positive following enrichment in broth; and for the remaining 11 (29%) horses, results of microbial culture were negative following direct plating and following enrichment in broth. For 3 of the 29 (10%) horses in group 3, results of microbial culture of fecal samples were positive following direct plating and following enrichment in broth, and for the remaining 26 (90%) horses, results of microbial culture were negative following direct plating but positive following enrichment in broth. For all 192 horses in group 4, results of microbial culture were negative following direct plating and following enrichment in broth.

When results of microbial culture following broth enrichment were used as the gold standard, sensitivity and specificity of the C difficile antigen assay were 66% (56/85) and 93% (192/207), respectively. When results of microbial culture following direct plating were used as the gold standard, sensitivity and specificity of the C difficile antigen assay were 93% (40/43) and 88% (218/249), respectively. There was good agreement between results of the C difficile antigen assay and results of microbial culture following broth enrichment (N = 0.62; P < 0.001) and following direct plating (N = 0.63; P < 0.001).

Other infectious agents—Three of the 33 horses in group 1 and 2 of the 38 horses in group 2 were also positive for Salmonella spp. Two foals in group 3 and 2 foals in group 4 were positive for rotavirus infection. One foal in group 4 was positive for Lawsonia intracellularis infection, and 1 foal in group 2 was positive for cryptosporidiosis.

Discussion

In the present study, horses in group 1 were clearly positive for CDAD, in that both toxin A and C difficile were identified in their feces, and horses in group 4 were clearly negative for CDAD, in that C difficile antigen and toxin A were not identified in their feces and results of microbial culture of fecal samples were negative. In contrast, the C difficile infection status of horses in groups 2 and 3 was less clear, as there currently is no consensus on how to define cases with evidence of C difficile infection (ie, horses with C difficile antigen in their feces or for which results of microbial culture are positive) that are negative for toxin A.⁹ It is possible that horses in groups 2 and 3 were colonized by C difficile strains in which toxin genes were silent (ie, nontranscribed) or in which toxin production was defective.^15,16 Alternatively, these horses may have been infected with strains negative for toxin A but positive for toxin B or with nontoxigenic strains of C difficile.^4,14,17 They may also have represented horses with levels of toxin A in their feces below the level of detection of the immunoassay (2 ng/mL). For these reasons, in some analyses in the present study, we elected to combine horses from groups 2 and 3 into a single group of horses (ie, horses with positive antigen assay or microbial culture results that were negative for toxin A). On the other hand, horses in group 2 differed from horses in group 3, in that C difficile antigen was detected in feces from group 2 horses, possibly indicating a greater extent of colonization. In group 3 horses, even though results of microbial culture were positive, results of the antigen assay were negative, possibly because the level of colonization was low enough that the concentration of C difficile antigen in feces was below the sensitivity of the assay. In support of this hypothesis is the fact that for 90% (26/29) of the horses in group 3, results of microbial culture were negative following direct plating but positive following broth enrichment.

Overall, only 33 horses in the present study were positive for toxin A in their feces, whereas C difficile was isolated from 85 horses following direct plating or broth enrichment. The low percentage of horses with positive culture results that also had positive toxin assay results (33/85 [39%]) was similar to that in previous studies^4,15 of horses. Arroyo et al,¹⁵ for example, found that only 44% of horses with diarrhea from which C difficile was isolated were positive for toxin A. It is unknown whether the discrepancy between microbial culture and toxin assay results represents insensitivity of the assay, the presence of toxigenic strains with silent toxin genes or defective toxin production, the presence of strains negative for toxin A but positive for toxin B, or the presence of nontoxigenic strains.⁴ An important limitation of the present study was that we were unable to test for toxin B in feces, which prevented us from detecting C difficile toxin production in horses colonized with strains that were only capable of producing toxin B. The ELISA used in the present study was selected because it was commonly used in the clinical microbiology laboratory at the University of California Veterinary Medical Teaching Hospital at the time of the study, provides rapid results (20 minutes), tests for both toxin A and C difficile antigen, and reportedly had higher sensitivity in the diagnosis of C difficile infection in humans and dogs than other commercial immunoassays available at that time.^10–12 The implications of finding evidence of C difficile infection (ie, C difficile antigen in feces or positive culture results) in horses with diarrhea that do not have detectable toxin A warrant further study, as does the possibility of infection with strains negative for toxin A but positive for toxin B and the potential role of other virulence factors in such cases.

Results of the present study provided some evidence that horses positive for toxin A (ie, group 1 horses) had more severe clinical disease than did horses negative for C difficile infection (ie, group 4 horses), in that the odds toxin A would be detected increased significantly as rectal temperature or band neutrophil count increased. In addition, higher heart rate and possibly the higher hospital costs also suggest that antigen or culture positive, but toxin negative, horses (groups 2 and 3) had more severe illness, compared with those that were negative for C difficile (group 4). The higher rectal temperatures and more severe band neutrophilia in horses positive for toxin A suggested that there was a more pronounced systemic inflammatory response in group 1 horses, in that rectal temperature and band neutrophil count are 2 of the 4 clinical criteria used to define systemic inflammatory response syndrome.¹⁸ Group 1 horses also had significantly higher Hct and hemoglobin concentration, compared with horses in group 4, although these factors were not retained in the final multivariate model, suggesting that these differences in the univariate analyses may have been the result of confounding factors. The higher mortality rate among horses in group 1, compared with horses in the other 3 groups combined, also supports the notion that horses positive for toxin A had more severe disease than did horses negative for toxin A.

Rectal temperature was higher and total hospitalization time was longer for horses positive for toxin A (ie, horses in group 1) in the present study than for horses with evidence of C difficile infection that were negative for toxin A (ie, horses in group 2 or 3), and heart rate and total hospitalization costs were higher for horses with evidence of C difficile infection that were negative for toxin A than for horses without evidence of C difficile infection (ie, horses in group 4). It is possible that there were other differences among groups that were not detected in the present study because the SD for many variables was quite large. In addition, although the overall number of horses was large, the number of horses positive for toxin A was not, further limiting our ability to detect significant differences among groups.

The fact that C difficile was isolated only after broth enrichment in some horses in the present study (4/33 [12%] group 1 horses, 12/38 [32%] group 2 horses, and 26/29 [90%] group 3 horses) suggested that concentration of C difficile in these horses was low. According to the manufacturer,^b the ELISA used in the present study requires an antigen concentration of 2 ng/mL in feces for a positive result. Taken together, these findings indicate that horses negative for antigen for which results of microbial culture following broth enrichment are positive likely have low concentrations of C difficile in their feces, compared with horses with positive microbial culture results following direct plating, and confirms that broth enrichment is a more sensitive method than direct plating, as has been reported previously.^19–21 We propose that horses with positive microbial culture results following broth enrichment but negative toxin assay results be tested for toxin A again, as these findings may represent early infection. Alternatively, growth of C difficile following broth enrichment may only be a marker of disruption of the gastrointestinal tract microflora, and strains that are isolated may not be the cause or primary agent of diarrhea in these horses. In a study²² evaluating fecal shedding of C difficile in horses treated with penicillin, horses were colonized with the organism after experimental exposure, yet feces remained negative for toxins. In addition, horses in that study remained subclinically infected, as they did not develop diarrhea or colic. These findings support the hypothesis that some horses in the present study, particularly those negative for C difficile antigen and toxin A, were secondarily colonized with C difficile as a result of concurrent disease, antimicrobial use, or some other factor. These findings raise questions as to what criteria should be used to make a diagnosis of CDAD in horses and whether the diagnosis should rely on results of microbial culture following direct plating or broth enrichment. The clinical implications of obtaining positive culture results only after broth enrichment require further study. In addition, more information is needed on whether horses with positive C difficile antigen or microbial culture results but negative toxin A assay results should be specifically treated for C difficile infection.

Sensitivity of the ELISA used in the present study for detection of the C difficile antigen was 93% (40/43) when assay results were compared with results of microbial culture following direct plating. Similarly, sensitivity of this assay for detection of C difficile antigen in human feces was reported to be 93% when assay results were compared with results of microbial culture following direct plating.²³ On the other hand, sensitivity of the assay in the present study was only 66% (56/85) when assay results were compared with results of microbial culture following broth enrichment. This lower sensitivity was likely attributable to the limit of quantification of the assay (2 ng/mL). The N statistic indicated that there was good agreement between results of the antigen assay and results of microbial culture following direct plating or broth enrichment. The percentage of concordant results when assay results were compared with results of microbial culture following direct plating (258/292 [88%]) was similar to the percentage when assay results were compared with results of microbial culture following broth enrichment (248/292 [85%]). However, when assays results were compared with results of microbial culture following direct plating, 31 of the 34 (91%) discordant results were attributable to positive immunoassay results and negative culture results and only 3 (9%) were attributable to negative assay results and positive culture results. In contrast, when immunoassay results were compared with results of microbial culture following broth enrichment, 15 of the 44 (34%) discordant results were attributable to positive assay results and negative culture results and 29 (66%) were attributable to negative assay results and positive culture results.

All horses in group 1 had been treated with antimicrobials prior to the onset of diarrhea, except 1 neonatal foal. Foals can develop diarrhea following infection with C difficile without prior antimicrobial administration, suggesting that C difficile may be a primary pathogen in this age group.²⁴ In adult people, antimicrobial administration is the most important but not the only risk factor for CDAD, and prior antimicrobial administration has previously been associated with C difficile infection in horses.^25,26 In contrast, Weese et al²⁷ found that only 26% of horses with CDAD had a history of previous antimicrobial administration. It is possible that differences in strains, the gastrointestinal tract microbial population in horses, and other environmental and management factors account for the difference in the percentage of horses with CDAD that had a history of antimicrobial administration between the present study and the study by Weese et al.²⁷ In addition, Weese et al²⁷ used a different assay for detection of toxin A in feces, and differences in sensitivity and specificity between assays may also account for the discrepancy in results. Because a number of horses in the present study with evidence of C difficile infection that were negative for toxin A (ie, horses in groups 2 and 3) had not received antimicrobials prior to development of diarrhea, other factors, including stress, hospitalization, transportation, withholding of feed, and surgery, likely played a role in colonization of the gastrointestinal tract with C difficile.

Mortality rate for horses in group 1 (8/33 [24%]) in the present study was significantly (P = 0.04) higher than mortality rate for horses in the other 3 groups combined (ie, horses negative for toxin A; 26/252 [10%]), which was similar to results of previous studies.^4,28 Another study,²⁷ however, did not find any significant difference in mortality rate between horses positive and negative for toxin A. The reasons for this difference are unclear, but may represent differences in strain virulence or host susceptibility. When comparing the mortality rate among the 4 groups in our study, there was not a statistical difference; this may reflect the relatively small number of horses in groups 1, 2, and 3 resulting in inadequate statistical power.

A potential drawback of the present study was the heterogeneity of the study population, which represented a wide range of ages, underlying conditions, and disease severity. We elected to include horses with a wide spectrum of disease severity to increase the number of horses with CDAD. Age distribution and distribution of fecal consistency did not differ significantly among study groups, making comparisons of clinical findings possible.

In conclusion, results of the present study provided some evidence that horses positive for toxin A had more severe clinical disease than did horses with evidence of C difficile infection that were negative for toxin A and horses with diarrhea without evidence of C difficile infection, highlighting the importance of toxin identification in horses with diarrhea. The finding that a large number of horses had evidence of C difficile infection but were negative for toxin A demonstrates the need for better defining the pathogenesis of CDAD in horses. Given the high sensitivity of the C difficile antigen assay, compared with results of microbial culture following direct plating, use of this assay may replace microbial culture as a more rapid screening test for C difficile infection in horses with diarrhea, although the assay cannot replace microbial culture following broth enrichment. If C difficile antigen is detected but not toxin A, additional or repeated testing for C difficile toxins may be warranted.