Abstract
OBJECTIVE To develop a risk prediction model for factors associated with an SeM-specific antibody titer ≥ 3,200 in horses after naturally occurring outbreaks of Streptococcus equi subsp equi infection and to validate this model.
DESIGN Case-control study.
ANIMALS 245 horses: 57 horses involved in strangles outbreaks (case horses) and 188 healthy horses (control horses).
PROCEDURES Serum samples were obtained from the 57 cases over a 27.5-month period after the start of outbreaks; serum samples were obtained once from the 188 controls. A Bayesian mixed-effects logistic regression model was used to assess potential risk factors associated with an antibody titer ≥ 3,200 in the case horses. A cutoff probability for an SeM-specific titer ≥ 3,200 was determined, and the model was externally validated in the control horses. Only variables with a 95% credibility interval that did not overlap with a value of 1 were considered significant.
RESULTS 9 of 57 (6%) case horses had at least 1 titer ≥ 3,200, and 7 of 188 (3.7%) of control horses had a titer ≥ 3,200. The following variables were found to be significantly associated with a titer ≥ 3,200 in cases: farm size > 20 horses (OR, 0.11), history of clinically evident disease (OR, 7.92), and male sex (OR, 0.11). The model had 100% sensitivity but only 24% specificity when applied to the 188 control horses (area under the receiver operating characteristic curve = 0.62.)
CONCLUSIONS AND CLINICAL RELEVANCE Although the Bayesian mixed-effects logistic regression model developed in this study did not perform well, it may prove useful as an initial screening tool prior to vaccination. We suggest that SeM-specific antibody titer be measured prior to vaccination when our model predicts a titer ≥ 3,200.
Strangles is a highly contagious disease of the upper respiratory tract of horses caused by Streptococcus equi subsp equi, with high morbidity but generally low mortality rates. However, complications, including bastard strangles (dissemination of infection with abscess formation at distant sites) and purpura hemorrhagica (immune-mediated vasculitis involving a type III hypersensitivity reaction), can be life-threatening.1–5 After natural exposure to Sequi subsp equi during an outbreak, horses and other equids develop antibody titers against SeM, a fibrinogen-binding, protein of S equi.6,7 A commercial ELISA is available for measurement of SeM-specific antibody titers, with results reported as negative (ie, < 200) or positive doubling dilution values ranging from 200 to 12,800.a,b Anecdotal expert opinion has suggested that horses with titers ≥ 3,200 should not be vaccinated because of the risk for development of purpura hemorrhagica.5,8 These horses are considered hyperresponders, with their high antibody titers predisposing them to deposition of antigen-antibody complexes in blood vessel walls and, thus, purpura hemorrhagica.5 According to basic immunologic principles, the most likely time that a horse might have a titer ≥ 3,200 would be within the first 6 months after exposure to S equi. Vaccination during (and within a year after) a strangles outbreak is therefore not recommended because of this risk.5,9
We are not aware of any prior long-term studies evaluating how long horses have a high SeM-specific antibody titer following a strangles outbreak. As such, there is a lack of information regarding when affected horses can safely receive the strangles vaccine after an outbreak. Davidson et al6 reported that the log antibody titer against SeM decreased 6 months after naturally occurring strangles infection in S equi–naïve research ponies; however, no further samples were obtained. In a primary research study investigating immunologic aspects of strangles, Sheoran et al10 reported that SeM-specific opsonophagocytic IgG(b), which binds to SeM immune complexes, was the predominant antibody detected in serum samples from horses vaccinated IM or recently recovered from infection. Serum IgG(b) concentrations were highest approximately 5 weeks after experimental infection and remained high 7 months after infection.10 Determining the duration of high SeM-specific antibody titers in horses naturally exposed to strangles is important to inform vaccination recommendations.
A previous study8 from our group evaluated SeM-specific antibody titers in 188 healthy client-owned horses at the time of annual vaccination. In that study, 21% (40) of the horses had a single titer ≥ 1,600, whereas 4% (7) had a single titer ≥ 3,200. Breeds other than warmbloods and Thoroughbreds, a history of previous vaccination, and older age were variables that were significantly associated with an SeM-specific antibody titer ≥ 1,600. Inclusion criteria limited that study to horses with no history of strangles for at least 1 year; only 8% (15) of the 188 horses had a history of clinically evident strangles from 1 to 8 years previously, and only 22% (42) had a confirmed history of strangles exposure.8 As such, there is a gap in the current knowledge base regarding the natural history of SeM-specific antibody titers in client-owned horses immediately following and for the 12-month period after an outbreak. Therefore, the objectives of the study reported here were, first, to report SeM-specific antibody titers in client-owned horses over a 27.5-month period following a naturally occurring strangles outbreak; second, to develop a model for predicting risk factors associated with an SeM-specific antibody titer ≥ 3,200 in client-owned horses over the 27.5 months after a naturally occurring strangles outbreak; and third, to externally validate the model in a population of control, healthy, client-owned horses residing in the same area. Our hypothesis was that a history of S equi disease would significantly increase the likelihood of an SeM-specific antibody titer ≥ 3,200.
Materials and Methods
Study protocol and design
This case-control study was reported in accordance with the STROBE (Strengthening The Reporting of Observational Studies in Epidemiology) Statement11 Veterinary Extension12 (Supplemental Table S1, available at http://avmajournals.avma.org/doi/suppl/10.2460/javma.250.12.1432). The study protocol was approved by the University of Pennsylvania Institutional Animal Care and Use Committee (Protocol #802555). Informed written consent was obtained from all owners prior to enrollment of each horse in the study (Supplemental Figure S1, available at http://avmajournals.avma.org/doi/suppl/10.2460/javma.250.12.1432).
Animals, setting, and participants
Case horses—For this study, we evaluated a convenience sample of 57 horses (case horses) residing within the service area of the Field Service of the New Bolton Center Hospital of the University of Pennsylvania during 3 different farm outbreaks of strangles. These cases included all horses available for blood sample collection that were residing on the farm during the outbreaks. Horses with clinical signs of infection, horses with known exposure (defined as nose-to-nose contact with or proximity to clinically affected animals, or both), and horses with no known exposure were included. The 3 outbreaks were designated A, B, and C; all 24 horses of outbreak A were available for sampling, 9 of 15 horses of outbreak B were Field Service patients and available for sampling; and 24 of 36 horses of outbreak C were available for sampling. Outbreak A had 4 (17%) horses with clinical signs, outbreak B had 11 (73%) horses with clinical signs, and outbreak C had 17 (47%) horses with clinical signs. Blood sample collection occurred from October 2007 to April 2012.
Control horses—For external validation of our model, we included samples obtained from 188 healthy horses (control horses) residing on 26 other farms also located within the New Bolton Center Field Service treatment area at the time of routine annual vaccination. These samples had been collected as part of a previous study8 between January 2006 and January 2007. Control horses were ≥ 1 year of age, with a complete medical history.
Sample collection and analysis
Blood samples were collected from the 57 case horses 1.5, 2.5, 3, 6, 8.8, 9.1, and 13 months after the identification of the index case for each respective outbreak. If any S equi SeM-specific antibody titers were ≥ 1,600 within this time frame and had not decreased by 13 months, collection of samples from the identified horses continued for an additional 7 to 14.5 months, at intervals of approximately 6 months. Not every horse was available at every sample collection time because of lack of access, having been sold, or having been moved out of the area. Samples were collected from control horses as previously described.8 All blood samples were collected in glass tubes. Serum was obtained within 16 hours after blood collection. Serum samples were stored at −18°C and analyzed in batches by a commercial laboratorya for SeM-specific antibody titers with an ELISA as previously described.8 Laboratory personnel were blinded to medical, exposure, and vaccination history of all horses.
Data analysis
For the 57 case horses, Bayesian implementation of mixed-effects logistic regression analysis was used to identify factors associated with having an SeM-specific antibody titer ≥ 3,200 after a naturally occurring S equi outbreak. This cutoff was derived from results of a previous study8 in which 21 horses with SeM-specific antibody titers of 1,600 received the intranasal S equi vaccine without adverse events, including purpura hemorrhagica, for up to 2 years after vaccination.
Risk factors examined in the logistic regression analysis included signalment (age as a continuous variable, sex as a categorical variable, and breed as a categorical variable [Thoroughbred or warmblood, Quarter Horse, or other breed]). Additional risk factors included whether the horse had been vaccinated with an attenuated-live intranasal S equi vaccine,c whether the horse had been treated with systemic antimicrobials, whether the horse had been exposed to horses with strangles, and whether the horse had ever had clinical signs of strangles and, if so, the severity of the clinical signs. Severity of clinical signs was represented categorically as absent, fever only, nasal discharge, draining external abscess formation, guttural pouch empyema, or draining abscess formation and auditory tube diverticulum (guttural pouch) empyema. Farm size was represented dichotomously as large (> 20 horses) versus small (≤ 20 horses). Time (days) after identification of the index case for each outbreak (continuous variable) was also examined but was dropped from the statistical analysis because time since the most recent outbreak was not known for some horses in the control cohort.
Bias
Potential sources of bias addressed in the logistic regression analysis included age, which was confirmed by examination of the teeth; history of clinical signs of disease as determined from the medical records, from diagnostic testing for S equi infection, or from 1 of 2 authors' (AGB and MAS) or their colleagues' personal clinical experiences with the case; and the fact that the number of horses could be physically counted to determine the size of the farm. The variables severity of clinical signs and whether the horse had been exposed to horses with strangles were dropped from the analysis because of a high degree of colinearity with the presence of clinical signs of disease. An a priori power calculation was not performed.
Model development and validation
Variables to be included in further analyses were selected by means of univariate logistic regression. Only variables with values of P ≤ 0.2 were included unless otherwise stated. The final multivariate regression model was determined with a combination of purposeful and automatic backward model selection processes. Missing data were handled on the basis of a missing-at-random assumption. The final regression model included 2 variables that were missing 1 observation (sex and age) and 1 variable missing 2 observations (whether disease was present). Overall, 3 horses were excluded because of missing data. All analyses were performed with standard software.d Values of P ≤ 0.05 and 95% confidence and credibility intervals that did not include a value of 1 were considered significant.
Model performance was evaluated in the validation data set of 188 control horses, with non-parametric area under the calculated receiver operating characteristic curve as a summary statistic for goodness of fit.13 Details of the Bayesian mixed-effects logistic regression model are provided (Supplemental Appendix S1, available at http://avmajournals.avma.org/doi/suppl/10.2460/javma.250.12.1432).
Results
Participants
A total of 245 horses were included in the study. The recruitment and flow of participants is presented (Figure 1).
Flow diagram illustrating recruitment and flow of participants in a case-control study for development of a model for predicting risk factors associated with an SeM-specific antibody titer ≥ 3,200 in client-owned horses after strangles outbreaks.
Citation: Journal of the American Veterinary Medical Association 250, 12; 10.2460/javma.250.12.1432
Descriptive data
Age of 1 case horse was not recorded. For the remaining 56 case horses, median age was 6.5 years (interquartile range [25th to 75th percentile], 4 to 12.5 years). Thirty-three (58%) case horses were Quarter Horses, 11 (19%) were Thoroughbreds, 4 (7%) were warmbloods, and 9 (16%) represented other breeds. Five (9%) horses were stallions, 23 (40%) were geldings, and 29 (51%) were mares. Of the 57 horses, 24 (42%) had clinical signs of strangles, 22 (39%) were known to have only been exposed to horses with strangles (defined as nose-to-nose contact with or proximity to clinically affected animals, or both), and 11 (19%) had no known exposure. Of the 24 clinically affected horses, 17 (71%) had nasal discharge, external abscess formation, guttural pouch empyema, or both abscesses and empyema; the remaining 7 (29%) clinically affected horses had fever only. Outbreak A lasted 21 days, outbreak B lasted 90 days, and outbreak C lasted 60 days. Median duration of clinical disease for 23 clinically affected horses was 10 days (interquartile range, 7 to 21 days; duration of clinical disease was not available for 1 horse). Seven of 57 (12%) horses had previously been vaccinated against S equi.
Median age of the 188 control horses was 11 years (interquartile range, 7 to 16 years). Of the 188 horses, 105 (56%) were geldings, 80 (43%) were mares, and 2 (1%) were stallions (sex of 1 horse was not recorded). Fifty-six (30%) of the horses were Thoroughbreds, 56 (30%) were Quarter Horses and related breeds, and 20 (11%) were warmbloods. The remaining 56 (30%) horses represented 8 other breeds or were of unknown breeding. Fifteen (8%) horses had a history of clinically evident strangles, 27 (14%) had only been exposed to horses with strangles, and 146 (78%) had no known exposure to horses with strangles. Data on clinical signs and duration of disease were not available for these horses. Median time since the last strangles outbreak was 5 years (interquartile range, 4 to 5 years). Eighty-nine (47%) horses had previously been vaccinated against S equi.
The SeM-specific antibody titers for 200 serum samples obtained from the 57 case horses over the 27.5 months following outbreaks ranged from negative (< 200) to 12,800 (Table 1). Nine of the 57 (6%) case horses had a titer ≥ 3,200 for a total of 15 observations from 6 to 20 months after an outbreak. Seven of the 188 (3.7%) healthy control horses had titers ≥ 3,200.
SeM-specific antibody titers for horses (cases, n = 57) in a case-control study designed to develop and validate a model for predicting risk factors associated with an SeM-specific antibody titer ≥ 3,200 in client-owned horses after strangles outbreaks.
| SeM-specific antibody titer | Months after beginning of outbreak | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 1.5 | 2.5 | 3 | 6 | 8.8 | 9.1 | 13 | 20 | 27.5 | ||
| < 200 | 0 | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 1 | 3 |
| 200 | 15 | 1 | 14 | 15 | 12 | 2 | 16 | 1 | 2 | 78 |
| 400 | 11 | 5 | 8 | 5 | 6 | 8 | 7 | 3 | 0 | 53 |
| 800 | 6 | 5 | 2 | 6 | 3 | 10 | 0 | 1 | 0 | 33 |
| 1,600 | 5 | 0 | 3 | 1 | 4 | 2 | 0 | 2 | 1 | 18 |
| 3,200 | 4 | 0 | 0 | 2 | 0 | 0 | 0 | 1 | 0 | 7 |
| 6,400 | 4 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 6 |
| 12,800 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 2 |
| Total | 45 | 11 | 30 | 30 | 27 | 22 | 23 | 8 | 4 | 200 |
Data represent number of horses.
Bayesian implementation of the mixed-effects logistic regression analysis
Results of univariate logistic regression analysis of potential associations between study variables and a serum SeM-specific antibody titer ≥ 3,200 in the 57 case horses were summarized (Table 2). Breed, history of strangles vaccination, and treatment with systemic antimicrobials were not significantly associated with a serum SeM-specific antibody titer ≥ 3,200 and were excluded from further analyses. The odds of having a titer ≥ 3,200 decreased at a rate of 1%/d after outbreaks; however, this variable was not included in the final multivariate regression model because data on time since the last strangles outbreak were not available for all horses in the control cohort.
Results of univariate logistic regression analysis of clinically relevant independent variables potentially associated with an SeM-specific antibody titer ≥ 3,200 for the horses in Table 1.
| Variable | |||
|---|---|---|---|
| Farm size | |||
| > 20 horses | |||
| ≤ 20 horses | |||
| History of clinically evident disease | |||
| Yes | |||
| No | |||
| Days after outbreak start* | |||
| Age† | |||
| Breed | |||
| Quarter Horse and all other breeds | |||
| Thoroughbred and warmblood | |||
| Sex | |||
| Female | |||
| Male | |||
| Treatment with systemic antimicrobials | |||
| Yes | |||
| No | |||
| History of S equi vaccination | |||
| Yes | |||
| No | |||
Time (days after outbreak) was a continuous variable; the OR represents a decrease in odds associated with each 1-day increase after the start of the outbreak.
Age was evaluated as a continuous variable; the OR represents the increase in odds associated with each 1-year increase in age.
CI = Confidence interval.
The final statistical model included factors for history of clinical disease, age, sex, and farm size (Table 3). After an outbreak, a horse with a history of clinical disease was 7.92 times as likely to have a titer ≥ 3,200 as was a horse with no history of clinically evident strangles. With each 1-year increase in age, a horse was more likely to have a titer ≥ 3,200; however, the effect of age was not significant (95% credibility interval, 0.93 to 1.35). Nevertheless, there was substantial evidence that inclusion of age improved the predictive value of the statistical model. Stallions were substantially less likely (OR, 0.11) than geldings to have a titer ≥ 3,200, and horses on large farms (> 20 horses) were substantially less likely (OR, 0.11) than horses on small farms to have an antibody titer ≥ 3,200.
Final Bayesian implementation of a mixed-effects logistic regression model demonstrating the potential association between clinically relevant variables and an SeM-specific antibody titer ≥ 3,200 for the horses in Table 1 (198 observations).
| Variable | ||
|---|---|---|
| Farm size | ||
| > 20 horses | ||
| ≤ 20 horses | ||
| History of clinical disease | ||
| Yes | ||
| No | ||
| Age* | ||
| Sex | ||
| Stallion | ||
| Mare | ||
| Gelding | ||
Age was evaluated as a continuous variable; the OR represents the increase in odds associated with each 1-year increase in age.
Details of the Bayesian mixed-effects logistic regression model are provided (Supplemental Appendix S1, available at http://avmajournals.avma.org/doi/suppl/10.2460/javma.250.12.1432).
External validation and model application
The final Bayesian implementation of the mixed-effects logistic regression model (including variables for history of clinical disease, age, sex, and farm size) was applied to the healthy control cohort of 188 horses, with a probability ≥ 0.002 used as the cutoff to predict an SeM-specific antibody ≥ 3,200. The model had a sensitivity of 100% (95% confidence interval, 0.59 to 1.00) and specificity of 24% (95% confidence interval, 0.18 to 0.31) with area under the receiver operating characteristic curve of 0.62 (95% confidence interval, 0.59 to 0.65; Figure 2).
Receiver operating characteristic curve for a Bayesian mixed-effects logistic regression model (with variables for age, sex, history of clinically evident disease, and farm size) used to predict whether horses had an SeM-specific antibody titer ≥ 3,200 (the cutoff probability for an SeM-specific antibody titer ≥ 3,200 was ≥ 0.002). The model was applied to data for serum samples obtained from a control cohort of 188 healthy horses at the time of vaccination. The model had a sensitivity of 1.0 (95% confidence interval, 0.59 to 1.00) and specificity of 0.24 (95% confidence interval, 0.18 to 0.31); area under the curve was 0.62 (95% confidence interval, 0.59 to 0.65). Details of the model are provided (Supplemental Appendix S1, available at http://avmajournals.avma.org/doi/suppl/10.2460/javma.250.12.1432).
Citation: Journal of the American Veterinary Medical Association 250, 12; 10.2460/javma.250.12.1432
Discussion
In the present study, we developed a Bayesian implementation of a mixed-effects logistic regression model to assess the risk for an SeM-specific antibody ≥ 3,200 over a 27.5-month period after naturally occurring strangles outbreaks in horses. We then validated this model, applying it to a cohort of healthy horses from the same practice area in an effort to predict which horses would have an SeM-specific antibody titer ≥ 3200.
For case horses in the present study, the final statistical model included 4 factors—a history of strangles, age, sex, and farm size—that were significantly associated with an SeM-specific antibody titer ≥ 3,200 after an outbreak. Although detectable SeM-specific antibody titers may not protect against future exposure to S equi, they are indicative of past exposure.10 Disease exposure and severity of clinical signs of disease could not be evaluated separately in the present study because these 2 variables reflected a history of strangles. Interestingly, in a previous study,8 we found that a history of clinical signs of disease was not a significant risk factor for a titer ≥ 1,600 when only healthy horses (ie, horses with no history of clinical signs of S equi disease for at least 12 months) were examined. Because the present study evaluated both healthy and affected (ie, involved in a recent outbreak) horses, we suggest that the results more accurately assessed the risk for high (ie, ≥ 3,200) titers.
Interestingly, living on a large farm (ie, > 20 horses) was protective against an SeM-specific antibody titer ≥ 3,200 following an outbreak in the present study. This was unexpected because large farms typically have more movement of horses on and off the farm, often resulting in more exposure to different horses. However, there was potential bias in the present study because study horses were from only 3 different farms after outbreaks, and results may have reflected individual farm and disease management factors, biosecurity measures, housing density, or some combination of these factors; rather than the size of the farm. One of the farms with > 20 horses had only 4 horses with clinical signs. This farm was very successful at controlling spread of the disease with biosecurity measures (eg, isolation of affected animals, personnel barrier precautions, and designated personnel for managing sick animals), such that many of the study horses on this farm had low SeM-specific antibody titers. The other 2 farms had higher overall percentages of horses with a history of clinical signs of disease. Most case horses that had SeM-specific antibody titers ≥ 3,200 in the present study were from a single farm with < 20 horses. Interestingly, these horses had been moved after the outbreak, and the SeM-specific antibody titers were measured when the animals were on different farms; as such, we suggest that it is possible that S equi strains that were not present during that outbreak contributed to the SeM-specific antibody titers ≥ 3,200.
Being a stallion was also found to be protective against an SeM-specific antibody titer ≥ 3,200 for case horses in the present study. However, this finding may have been a result of confounding associated with different management practices related to competition, travel, and housing facilities to prevent unscheduled breeding for stallions versus geldings or females. In addition, the small sample size limits the generalizability of this finding, and further investigation is required.
The passage of time was also protective against an SeM-specific antibody titer ≥ 3,200 after an outbreak. The odds of having a titer ≥ 3,200 decreased at a rate of 1%/d after the start of an outbreak, such that > 100 days after the start of the outbreak, the predictive value of time at the start of the outbreak was minimal. An understanding of immunology indicates that SeM-specific antibody titers will slowly decrease with time, resulting in the assumption that, 1 year after the beginning of an outbreak, most affected horses would have an SeM-specific antibody titer < 3,200. Sheoran et al10 evaluated IgG and IgA subisotypes in 6 S equi–naïve research horses over a 7-month period after an experimentally induced strangles outbreak. Results indicated that IgG(b) was the predominant subisotype, with titers highest 5 to 6 weeks after initial infection and remaining high at 7 months, although titers had declined 4-fold compared with the first values measured at week 3. The SeM ELISA binds to SeM immune complexes with protein G, which bonds most strongly with IgG(b) in horses.14 In the present study, the first samples collected would correspond with those obtained at 6 weeks in the study by Sheoran et al,10 and earlier samples were unfortunately not collected.
In a previous study by Davidson et al,6 SeM-specific antibody titers obtained following naturally occurring strangles outbreaks were assessed for 6 months in a group of 30 previously S equi–naïve 4- to 6-month-old research ponies with clinically evident disease. The authors reported a 2-fold decrease in mean log SeM-specific antibody titers by 6 months after infection; however, the animals were not followed further. In that study,6 animals were in a controlled research setting, and titers were measured with a validated FgBP1 ELISA available for research purposes.
Unlike the populations in previous studies involving young naïve research horses10 and ponies,6 our study population was varied with respect to immunologic status and history (eg, vaccination, previous S equi exposure, and age). Being a population of client-owned animals, the patients were housed in a typical farm or boarding stable setting with varied implementation of biosecurity procedures. The management goals for these outbreaks were to minimize morbidity rate, develop lasting immunity, and minimize facility quarantine periods. The horses were treated with a variety of systemic and local (guttural pouch) antimicrobials that we suspect, but could not statistically prove, decreased the immunologic response. Two horses in the present study had aberrant titers with values ≥ 3,200 late in the first year and into the second year after the outbreak. Additional (ie, new) exposures could have been the cause for the late increases in SeM-specific antibody titers in these horses, although both of these horses had left the practice area and been moved to a new state in a private trailer and were housed in a private barn on the owner's property. Neither horse had clinical signs of purpura hemorrhagica or distant abscesses associated with these high titers. Unfortunately, titers were not obtained from sentinel horses concurrently with these horses to identify whether there were any additional S equi exposures.
Financial constraints and field and opportunistic design were some of the limitations of the present study. Without the ability to anticipate natural outbreaks, no baseline SeM antibody titers could be obtained, and the decision to perform the study was made near the resolution of the first outbreak. Therefore, no titers were measured when horses had clinical signs. The number of horses with titers ≥ 3,200 was small. This may have been affected by the use of antimicrobials in 12 of 23 horses with clinical signs in the case cohort, but this could not statistically be proven. Streptococcus equi strains or management of outbreaks may also have affected our results. Furthermore, in this case-control study, we were unable to collect time-matched samples from the case horses for model development and collected samples from the control horses only at the time of vaccination. However, we attempted to account for this in the model by removing time after outbreaks as a variable. All control horses at the time of vaccination had a history of at least 1 year since the last strangles outbreak, and most had a history of 3 years. One year was beyond the time during which titers were increasing in the case horses.
Despite the small number of horses with titers ≥ 3,200 in both the control (7/188), and case (9/57) horses, findings of the present study suggested that, given the high morbidity and potential mortality rates associated with purpura hemorrhagica, clinicians should evaluate SeM-specific antibody titers prior to vaccinating horses involved in a strangles outbreak within the past year. Although the Bayesian mixed-effects logistic regression model we developed in this study did not perform well, as demonstrated by the area under the receiver operating characteristic curve and the low specificity, it may prove useful as an initial screening tool to identify which horses should be tested prior to future vaccination against S equi. All the information in the model is easily obtainable. The cost of possible overtesting needs to be weighed with the risk that a horse will develop purpura hemorrhagica. Currently, we recommend that SeM-specific antibody titers be measured prior to vaccination for S equi when the model reported here predicts an SeM-specific antibody titer ≥ 3,200. Horses identified by the model with a titer < 3,200 are considered safe to vaccinate for S equi without prior testing. Furthermore, on the basis of results of the present study, we suggest that a smartphone application could be developed to enable field veterinarians to enter the values for the 4 variables used in our model (ie, age, sex, history of clinically evident disease, and farm size) to easily and rapidly determine whether the SeM-specific antibody titer should be measured. Future research, including with a larger number of affected horses and broader geographic distribution, will improve our model and increase the generalizability of these findings.
Acknowledgments
Supported by the Frances Cheney Glover Endowment Fund, Section of Field Service, Department of Clinical Studies—New Bolton Center, School of Veterinary Medicine, University of Pennsylvania.
The authors declare that there were no conflicts of interest. All titer measurements were performed as fee-for-service.
Presented in part as a poster at the 9th International Conference on Equine Infectious Diseases, Lexington, Ky, October 2012.
Footnotes
IDEXX, Sacramento, Calif.
Equine Diagnostic Solutions LLC, Lexington, Ky.
Pinnacle, Fort Dodge Animal Health, Fort Dodge, Iowa.
STATA 14, Statacorp, College Station, Tex.
References
1 Pusterla N, Watson JL, Affolter VK, et al. Purpura haemorrhagica in 53 horses. Vet Rec 2003; 153: 118–121.
2 Galan JE, Timoney JF. Immune complexes in purpura hemorrhagica of the horse contain IgA and M antigen of Streptococcus equi. J Immunol 1985; 135: 3134–3137.
3 Carlson G. Diseases of the hematopoietic and hemolymphatic systems. In: Smith B, ed. Large animal internal medicine. 3rd ed. St Louis: CV Mosby, 2002; 1043.
4 Radostits O, Gay C, Hinchcliff K W, et al. Disease associated with allergy: purpura hemorrhagica. In: Radostits O, Gay C, Hinchcliff KW, et al, eds. Veterinary medicine. A textbook of the diseases of cattle, sheep, goats, pigs and horses. 10th ed. Philadelphia: Saunders Elsevier, 2007; 1925–1926.
5 Sweeney CR, Timoney JF, Newton JR, et al. Streptococcus equi infections in horses: guidelines for treatment, control, and prevention of strangles. J Vet Intern Med 2005; 19: 123–134.
6 Davidson A, Traub-Dargatz JL, Magnuson R, et al. Lack of correlation between antibody titers to fibrinogen-binding protein of Streptococcus equi and persistent carriers of strangles. J Vet Diagn Invest 2008; 20: 457–462.
7 Harrington DJ, Sutcliffe IC, Chanter N. The molecular basis of Streptococcus equi infection and disease. Microbes Infect 2002; 4: 501–510.
8 Boyle AG, Sweeney CR, Kristula M, et al. Factors associated with likelihood of horses having a high serum Streptococcus equi SeM-specific antibody titer. J Am Vet Med Assoc 2009; 235: 973–977.
9 Wilson WD. Vaccination of mares, foals and weanlings. In: McKinnon AO, Squires EL, Vaala W, eds. Equine reproduction. 2nd ed. Ames, Iowa: Wiley-Blackwell, 2011; 302–327.
10 Sheoran AS, Sponseller BT, Holmes MA, et al. Serum and mucosal antibody isotype responses to M-like protein (SeM) of Streptococcus equi in convalescent and vaccinated horses. Vet Immunol Immunopathol 1997; 59: 239–251.
11 von Elm E, Atman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 2008; 61: 344–349.
12 O'Connor AM, Sargeant JM, Dohoo IR, et al. Explanation and elaboration document for the STROBE-Vet Statement: Strengthening the Reporting of Observational Studies in Epidemiology—veterinary extension. J Vet Intern Med 2016; 30: 1896–1928.
13 Hosmer DW, Lemeshow S. Assessing the fit of the model. In: Hosmer DW, Lemeshow S, eds. Applied logistic regression. 2nd ed. New York: John Wiley & Sons, 2000; 143–202.
14 Lewis MJ, Wagner B, Woof JM. The different effector function capabilities of the seven equine IgG subclasses have implications for vaccine strategies. Mol Immunol 2008; 45: 818–827.
