Case-control study of factors associated with excessive proportions of ...

02-05-0235.qxd

2/6/2003

1:57 PM

Page 613

Case-control study of factors associated with excessive proportions of early fetal losses associated with mare reproductive loss syndrome in central Kentucky during 2001 Roberta M. Dwyer, DVM, MS, DACVPM; Lindsey P. Garber, DVM, MS; Josie L. Traub-Dargatz, DVM, MS, DACVIM; Barry J. Meade, MS, DVM; David Powell, BVSc; Michael P. Pavlick, DVM; Albert J. Kane, DVM, MPVM, PhD entral Kentucky is one of the largest Thoroughbred breeding epicenters in the world with more than 20,700 mares bred in 2001.1,2 The equine industry contributes $3.4 billion to the Kentucky economy3 and is one of the largest industries in the state. Thus, farm veterinarians and managers, diagnostic laboratory personnel, and extension veterinarians take care in monitoring for outbreaks of any type of disease, including outbreaks of abortion. Historically, most aborted fetuses in central Kentucky have been submitted to the University of Kentucky Livestock Disease Diagnostic Center (LDDC) in Lexington for complete diagnostic and pathologic examinations, as the LDDC does not charge a fee for livestock necropsy.a Because of the intensive breeding schedules in central Kentucky, especially on Thoroughbred and Standardbred farms, mares are palpated per rectum and have ultrasonographic examinations performed regularly for pregnancy determination and to monitor fetal viability and development. Most mares are palpated or have ultrasonographic examinations at 42 days of gestation for fetal insurance purposes. Several farms also have veterinarians perform ultrasonography on mares at 60 to 65 days of gestation for fetal sex determination. The occurrence of a markedly increased number of early fetal losses (EFLs) was brought to the attention of extension veterinarians at the University of Kentucky’s Maxwell H. Gluck Equine Research Center on May 1, 2001. Veterinary practitioners reported that an excessive number of previously pregnant mares were found to no longer be pregnant at 60 days of gestation.b,c Rectal palpation findings were normal; however, on ultrasonographic examination of affected mares, the fetus was either absent, present but dead, present but with slow heart rate and movements, or

From the Maxwell H. Gluck Equine Research Center, Department of Veterinary Science, University of Kentucky, Lexington, KY 40546-0099 (Dwyer, Powell); USDA:APHIS:VS Centers for Epidemiology and Animal Health, 2150 Centre Ave, Bldg B, Fort Collins, CO 80526 (Garber, Kane); the Department of Clinical Sciences, College of Veterinary Medicine and Biomedical Sciences, Colorado State University, Fort Collins CO 80523 (Traub-Dargatz); and USDA:APHIS:VS, PO Box 399, Frankfort, KY 40602 (Meade, Pavlick). Supported by the University of Kentucky College of Agriculture and the United States Department of Agriculture’s Veterinary Services as part of the Animal and Plant Health Inspection Service. Presented in part at the Society for Veterinary Epidemiology and Preventive Medicine Conference, Cambridge, England, April, 2002, and at the Annual Meeting of the American College of Veterinary Internal Medicine, Dallas, May, 2002. The authors thank Dr. Alfonso Torres (United States Department of Agriculture), Dr. Don Notter (Kentucky State Veterinarian), Marjorie Swanson and Mike Durham (Centers for Epidemiology and Animal Health), Diane Furry (Gluck Equine Research Center), and Linda Javid (Kentucky Thoroughbred Farm Managers Club) for assistance. Individuals who assisted in development of the questionnaire used in the study and in data collection are listed at the end of the article. Published as Kentucky Agricultural Experiment Station Article No. 02-14-126 with the approval of the Dean and Director, College of Agriculture and Kentucky Agricultural Experiment Station. Address correspondence to Dr. Dwyer. JAVMA, Vol 222, No. 5, March 1, 2003

Scientific Reports: Original Study

613

EQUINE

Objective—To identify factors associated with excessive proportions of early fetal losses associated with mare reproductive loss syndrome in central Kentucky during 2001. Design—Case-control study. Procedure—Questionnaires were used to collect information on farm-, pasture-, and individual animallevel factors purportedly associated with mare reproductive loss syndrome. Data were collected for 133 farms (97 with excessive proportions of early fetal losses and 36 control farms) representing 6,576 mares. Results—Factors significantly associated with an increased risk of excessive early fetal losses were exposure to moderate to high concentrations of Eastern tent caterpillars, exposure to cherry trees, farm size ≥ 50 broodmares, being bred during February 2001, and frequent exposure to waterfowl. Feeding hay to mares outside was associated with a decreased risk of excessive proportions of early fetal losses. Pasture composition and management factors were not significantly different between affected and control pastures. Individual animal-level factors were investigated on 6 farms representing 340 mares, and age, parity, and pre- and postbreeding treatments were not significantly associated with risk of early fetal loss. Conclusions and Clinical Relevance—Results suggest that limiting exposure to Eastern tent caterpillars and cherry trees and feeding hay to mares outside may help decrease the risk of excessive proportions of early fetal losses associated with mare reproductive loss syndrome. (J Am Vet Med Assoc 2003;222: 613–619)

C

EQUINE

02-05-0235.qxd

2/6/2003

1:57 PM

Page 614

present but surrounded by abnormal echogenic fluid (cloudy and flocculent). Fetuses with abnormal heart rates and movements later died. Some fetuses that were surrounded by abnormal echogenic fluid did survive. Concurrently, a dramatically increased number of late-term aborted fetuses were submitted to the LDDC for necropsy. Historical findings for mares with lateterm abortions (LTAs) included dystocia, foaling while standing, explosive parturition, premature placental separation, production of a stillborn foal, and agalactia. Many foals were born weak and survived for only about 4 days, overwhelming the local equine veterinary hospitals. No infectious or contagious agents could be identified as the primary causes of these EFLs and LTAs,4 and additional extensive pasture sampling and testing revealed no endophyte-infected pastures or known mycotoxins.5 A preliminary survey to determine the extent of EFLs was faxed to > 280 Thoroughbred farms in central Kentucky on May 7, 2001, and 159 farm managers responded. Of mares that were pregnant at 42 days of gestation, only 79% were still pregnant as of May 7, 2001, reflecting a 21% loss.6 Pregnancy loss rates for Thoroughbreds have been studied, and from 40 days of gestation to parturition, approximately 12% fetal loss is typical.d Similarly, the number of late-term aborted fetuses examined by the LDDC was high with 318 aborted fetuses received between April 28 and May 8, 2001.7 By the end of June, > 550 late-term aborted fetuses had been submitted. Fetuses represented 18 breeds and were submitted from multiple Kentucky counties. Despite extensive bacteriologic, virologic, and pathologic examination of the submitted fetuses, no infectious cause for the abortions was determined. The simultaneous occurrence of EFLs and LTAs was subsequently named mare reproductive loss syndrome (MRLS). Environmental factors were unusual during April and May of 2001 in Kentucky. The weather during April included warm periods with highs of 26.7oC (80oF), followed by 2 days of frost on April 17 and 18, then warm weather again. Additionally, it was the driest April in the past 104 years of Kentucky history.8 Another unusual environmental factor was the emergence of multitudes of Eastern tent caterpillars (Malacosoma americanum) living primarily in black cherry trees (Prunus serotina) that are abundant in Kentucky and other fruit trees (eg, apple, crabapple, and pear). In areas of high concentrations, these caterpillars covered fence lines, filled feed buckets, blanketed driveways, and were found on the walls of barns, stalls, and homes. As events unfolded, team leaders were formally assigned areas of organization and investigation including disease pathology, toxicologic and experimental studies (including mycotoxins, ergot alkaloids, phytoestrogens, and other toxins), and field epidemiology (the subject of the current report). The purpose of the study reported here was to identify factors associated with MRLS and areas for further research and investigation. Materials and Methods Study design—A case-control study was conducted to identify factors related to excessive fetal losses on horse 614


farms in central Kentucky during 2001. Epidemiologists discussed the fetal loss problem with veterinarians, basic scientists, and farm managers to gain an appreciation for the magnitude of the problem. A questionnaire was developed by a group of epidemiologists, animal scientists, veterinarians, reproductive physiologists, and an economist. Additional input on questionnaire design and administration was solicited from local veterinarians, farm managers, members of the equine industry, and other researchers. All questionnaires were administered on the farm by 23 volunteers, each of whom visited 6 to 7 farms. A list of 156 Thoroughbred farms willing to participate in further MRLS studies was available from a previous survey conducted through a local farm managers club on May 7, 2001.6 Since most of these farms had experienced problems with MRLS, additional Thoroughbred farms with few or no fetal losses attributable to MRLS were needed for comparison. These were identified through telephone calls to veterinarians and farm managers. Farms with Standardbred mares or equids of other breeds were also solicited to participate in the study by telephoning several equine practices in the central Kentucky area. To be eligible to participate in the study, farms had to have ≥ 20 broodmares, and veterinarians had to have performed ultrasonographic examination on broodmares prior to 45 days of gestation and again at 60 days of gestation or later. A list of 150 farms that fit these study criteria was created. It became apparent early in the course of the study that the principle manifestation of MRLS on the recruited farms was EFLs. Therefore, although the occurrence of LTAs was recorded, the focus of analyses of data obtained through the case-control study was EFLs. A 3-tiered approach to data collection and analysis was used. A general questionnaire was designed to collect farmand pasture-level data. Farm-level data related to farm management practices and demographic information. Case farms included those on which owners and managers perceived EFLs to be in excess of what they considered normal for their farms in comparison with the previous few years. Control farms were those for which owners and managers reported losses typical for their farms. For pasture-level data, the 1 pasture on each affected farm with the highest number of EFLs was selected as a case pasture. Control pastures consisted of the 1 pasture on each control farm with the greatest number of at-risk mares but no excess EFLs. This portion of the questionnaire included questions regarding pasture treatments, grass types, types of trees present, and other factors related to pasture management.9 An additional questionnaire was administered on 6 case farms with ≥ 50 mares at risk. This questionnaire was designed to collect data (including age, parity, breeding status, and any medications that had been administered and may have been associated with EFL) on individual animals. All horses on these farms that were bred between February 10 and April 1, 2001 and confirmed to be pregnant by ultrasonographic examination prior to 45 days of gestation were included. Case horses were horses confirmed by means of ultrasonographic examination at 60 days of gestation or later to no longer be pregnant. Control horses were horses confirmed to still be pregnant after 60 days of gestation. Study implementation—A training session was held for all study personnel during which questionnaires were reviewed and terminology used in the document was discussed. Confidentiality agreements for the farms were attached to the questionnaires, and farms were identified by farm number only. Data collectors were the only personnel who received both the names and identification codes of the farms for which they were responsible, since confidentiality was an important concern for the participating farms. The JAVMA, Vol 222, No. 5, March 1, 2003

02-05-0235.qxd

2/6/2003

1:57 PM

Page 615

Statistical analyses—A single investigator reviewed completed questionnaires, and questionable responses were validated with telephone calls to the data collector and farm representative as needed. Farm- and pasture-level data were entered into a databasee and exported for analysis with commercially available software.f Individual horse-level data were entered into a spreadsheetg and then similarly exported for analysis. Because of the low number of farms that only had LTAs (n = 4), LTA was not analyzed separately as an outcome. Farm size was dichotomized (< 50 mares and ≥ 50 mares). Proportions of case and control farms exposed to each factor of interest were calculated. Potential associations between farm EFL status (case or control) and individual farm-level variables were further evaluated with logistic regression.h On the basis of a priori hypotheses suggested by the unusual environmental conditions that summer, the presence of cherry trees, caterpillar concentration, and farm size were included as covariates in each model to simultaneously adjust for the potential effects of these variables while evaluating the other variables of interest. Case and control pastures were cross-classified by exposure to pasture-level factors of interest. When proportions of exposed pastures differed by > 10% between affected and control pastures, these variables were considered eligible for multivariable logistic regression modeling with backward elimination, in which pasture status was the response variable.h The Wald test and z-statistic for 1 model parameter at a time was used to select variables for the final model with a P value ≤ 0.05 required to remain in the final model. The proportion of mares that experienced EFLs was calculated for each level of the horse-level factors of interest. Odds ratios were calculated, and significance was determined with the χ2 test. Variables with an odds ratio ≥ 2 or ≤ 0.5 and a P value ≤ 0.20 were considered eligible for multivariable logistic regression modeling with EFL status as the response variable.h The Wald test and z-statistic for 1 model parameter at a time was used to select variables for the final model with a P value ≤ 0.05 required to remain in the final model. The clustered nature of the horse-level data (ie, horses within farm) was accounted for by use of the Taylor series linearization method. Goodness-of-fit statistics are not available with this survey analysis procedure. Therefore, final models selected for the pasture- and horse-level analyses were also analyzed using standard logistical regression techniques,i and goodness of fit was evaluated with the HosmerLemeshow (HL) statistic without accounting for clustering in the horse-level data. JAVMA, Vol 222, No. 5, March 1, 2003

Results Of the 150 farms eligible for inclusion in the study, 133 (88%) in 10 counties participated. Thoroughbred, Standardbred, American Saddlebred, and Morgan Horse farms were represented. Reasons cited for nonparticipation were lack of time because of being too busy with breeding stallions, illness of the farm manager or family members, or the farm manager being absent from the farm. The number of broodmares per farm at the time of data collection ranged from 11 to > 300 (mean, 126; median, 93). Of the 133 farms that participated in the survey, 97 were classified as having excessive EFLs (ie, owners and managers perceived EFLs to be in excess of what they considered normal for their farms in comparison with the previous few years). Of these, 66 had excessive EFLs but normal proportions of LTAs, 30 had excessive EFLs and LTAs, and 1 had excessive EFLs but was not at risk for LTAs because no mares were in late pregnancy at the time of the survey. The remaining 36 farms were classified as control farms (proportion of EFLs was not considered excessive, compared with previous years). Four of these farms reported having excessive LTAs but were included in the control group as they did not have excessive EFLs. Farm-level results—On EFL-affected farms, 79.7% (n = 2,233) of mares that were pregnant at 45 days of gestation underwent ultrasonographic examinations at ≥ 60 days of gestation. Of these, 58.4% (1,304) of mares were still pregnant and had normal ultrasonographic examination results, 6.0% (133) were pregnant but had abnormal uterine ultrasonographic examination results, and 35.6% (795) were not pregnant. On control farms, 89.7% (n = 501) of mares that were pregnant at 45 days of gestation underwent ultrasonographic examinations at ≥ 60 days of gestation. Of these, 86.6% (n = 434) were still pregnant and had normal ultrasonographic examination results, 3.6% (18) were pregnant but had abnormal uterine ultrasonographic examination results, and 9.8% (49) were not pregnant. Mean percentages of mares no longer pregnant at the time of ultrasonographic examination at ≥ 60 days of gestation were 35.9% for case farms (median, 37.9%) and 8.9% for control farms (median, 7.3%). Thirteen of 929 (1.4%) mares on case farms and 1 of 67 (1.5%) mares on control farms had signs of illness prior to an abnormal ultrasonographic finding. Factors associated with farms having excessive proportions of EFLs included having ≥ 50 mares on the farm (P = 0.01; Table 1), the presence of a high concentration of caterpillars (P < 0.001), and the frequent presence of waterfowl (P < 0.001). The only factor associated with a lower risk of EFLs was feeding hay outside, whether on the ground or in a manger or other device (P = 0.006). Factors not associated with increased EFLs were water source; method of delivering drinking water; ground moisture level; time on pasture during the day or night; percentage of total ration made up by pasture; being held off pasture from April 15 through 25; mowing practices before, during, and after the frosts of April 17 and 18; spreading manure on pastures; fertilization during fall 2000 or spring 2001 Scientific Reports: Original Study

615

EQUINE

project coordinator retained the key to all farm names and identification codes. Data collection was started in May 2001, 3 weeks after the initial idea for the study was conceived, and data collection was completed in 19 days because of the urgency of the situation. Data collectors contacted farm managers or owners for an appointment and completed the questionnaire with the farm manager or another representative from the farm. Photographs of affected and control pastures were taken to document pasture conditions, surface water, and tree composition. Data definitions provided for data collectors and farm representatives included high caterpillar concentration (ie, blankets of caterpillars on fences, waterers, or other surfaces), moderate caterpillar concentration (ie, many caterpillars in trees with some in pastures or barns), and low caterpillar concentration (ie, few caterpillars observed). Other terms that were defined included “feeding hay outside” (ie, horses fed on the ground or in feeders) and “frequent presence of waterfowl” (ie, waterfowl often seen or evidence thereof seen on the horse premises or in the immediate vicinity). A typical farm interview lasted 2 to 3 hours in addition to travel time.

02-05-0235.qxd

2/6/2003

1:57 PM

Page 616

Table 1—Farm-level factors associated with excessive proportions of early fetal losses on horse farms in central Kentucky during 2001

Factor

EQUINE

Farm size ⱖ 50 mares High caterpillar concentration Frequent presence of waterfowl Feeding hay outside

Case farms (%)

Control farms (%)

95% CI

Factor

66.0

38.9

3.2

1.3–7.7

67.0

22.2

6.4

2.5–16.9

Moderate or high caterpillar concentration Pasture primarily contains maiden or barren mares

60.8 82.1

25.0 91.4

5.6 0.19

2.2–14.3 0.06–0.61

Adjusted OR*

*Each factor was examined individually using logistic regression with farm size, presence of cherry trees, and caterpillar concentrations included as covariates when these factors were not the factor of interest. OR = Odds ratio. CI = Confidence interval for OR.

with ammonium nitrate, urea, or other compounds; pasture treatment with lime in 1999, 2000, or 2001; treatment of pastures with herbicides or pesticides during spring 2001; forage type (fescue, orchard grass, bluegrass, clover, or other grasses); percentage (10, 20, or > 40%) of clover in pastures combined with mowing practices during April; ragweed in pastures; percentage of edible vegetation in broodmare pastures; source of hay; hay from a 2000 or 2001 crop; grain fed to mares inside or outside; type of grain concentrate (unpelleted sweet feed, unpelleted grain, complete feed pellets or cubes, or grain or sweet feed mixed with pellets); grain contamination by wild or domestic animals’ feces; source of grain (bag, bulk, or home grown); type of bedding used; exposure to wildlife including rodents, raccoons or skunks, coyotes or foxes, deer or elk, and bats or opossums; wild animal exposure to stored grain or hay; domestic animal exposure to broodmares; and insect control measures (on horses and in barns, pastures, trees, and feed). Pasture-level results—Information from 95 affected pastures and 47 control pastures was analyzed. Two pasture-level factors were significantly associated with excessive EFLs: the presence of moderate or high caterpillar concentrations (P < 0.001; Table 2) and pastures containing barren or maiden mares bred in 2001 (P = 0.004). This model exhibited good fit with the data (P = 0.95). Factors not associated with the number of EFLs were pasture type including > 40% of any type of grass; presence of ragweed; quality of forage; manure (composted or not composted) spread on fields; chain harrowing of pastures; supplemental hay fed; pasture mowing before, during, or after the frosts of April 17 and 18; fertilization in fall 2000 or spring 2001 with any type fertilizer; pasture treatment with lime in 1999, 2000, or 2001; presence of fruit trees (besides cherry), evergreens, or deciduous trees other than fruit trees in or around pastures; and surface water in pasture. Additional analyses combining factors for percentage of clover in pastures (10, 20, or > 40%) with mowing practices during April did not indicate any association with increased percentages of EFLs. The number of herbicide applications to pastures was very low, and no pastures were sprayed with pesticides. Mare-level results—Mares on the 6 farms for which individual horse data were collected ranged from 2 to 22 years old (mean, 7.8 years; median, 7.0 616

Table 2—Pasture-level factors associated with excessive proportions of early fetal losses on horse farms in central Kentucky during 2001


Case pastures (%)

Control pastures Adjusted (%) OR

95% CI

87.4

40.5

7.1

2.6–19.8

67.4

27.7

6.5

1.8–23.4

Factors were analyzed by means of multivariable logistic regression with backward elimination. See Table 1 for key.

Table 3—Individual horse-level factors associated with excessive proportions of early fetal losses on horse farms in central Kentucky during 2001

Factor Bred in February Exposed to cherry trees in or around pasture

Case mares (%)

Control mares (%)

Adjusted OR

95% CI

50.0

10.5

5.4

2.5–12.0

96.1

74.2

7.6

1.1–51.4

Factors were analyzed by means of multivariable logistic regression with backward elimination. See Table 1 for key.

years). Parity ranged from 0 to 14 (mean, 2.5; median, 2.0). Of the 340 mares included in this portion of the study, 91 were maiden, 82 were barren, 162 had foaled in 2001 and were bred in 2001, and 5 had an unknown breeding status. Mare-level factors associated with excessive proportions of EFLs included being bred in February (P = 0.003; Table 3) and having been exposed to cherry trees (P = 0.04). This model exhibited acceptable fit with the data (P = 0.35). Mare-level factors that were not associated with EFLs included the mare’s state of origin (Kentucky vs another state), age, prebreeding uterine infusion, postbreeding uterine infusion, status at 2001 breeding (maiden, barren, or foaled in 2001), parity, medications administered (hormonal treatments, antimicrobials, nonsteroidal antiinflammatory drugs, or other drugs), use of probiotics or mycotoxin binders in feed, administration of rotavirus or herpesvirus vaccine (killed or modifiedlive virus vaccine), number of hours on pasture per day (day or night), any illness prior to breeding or during pregnancy, and caterpillar concentration in the mare’s environment. Discussion The goal of this study was to identify factors associated with MRLS in the hope that these factors could be used to direct future research activities toward finding a cause and preventing future occurrences of this new disease. Because of the low numbers of farms that only had LTAs, conclusions from these data are best limited to farms and pastures housing horses with EFLs or both EFLs and LTAs and to individual mares with EFLs. Results of the analyses performed in the present study indicated 5 factors associated with an increased risk of having excessive proportions of EFLs: farms having ≥ 50 mares, farms with frequent exposure to waterfowl, moderate to high concentration of caterpillars on farms and in pastures, mares with breeding JAVMA, Vol 222, No. 5, March 1, 2003

02-05-0235.qxd

2/6/2003

1:57 PM

Page 617

JAVMA, Vol 222, No. 5, March 1, 2003

consistent from year to year. Most farms maintain extensive production records for their mares as a matter of economic necessity, and the farm managers’ estimations of whether EFLs were suffered amounted to more than speculative recollections. Post-survey data verification showed that on farms where managers or owners identified a problem with EFLs, excessive EFLs occurred regardless of farm size. The association between frequent exposure of mares to waterfowl and excessive EFLs was an unusual and unexpected finding. Answers to this question were categorized as frequent, occasional, and never, with the interpretation of the latter 2 left to the farm manager. Follow-up is necessary to determine whether this finding is spurious or if waterfowl could be an environmental marker for another cause of MRLS. Feeding hay to mares outdoors was significantly associated with a decreased risk of excessive EFLs in the farm-level analysis but was not significant in the pasture-level analyses. A possible explanation is that on the farm-level portion of the survey, the question asked whether hay was fed outside, whereas on the pasture-level portion of the survey, the question asked only whether mares on that particular pasture were fed supplemental hay (nearly all were). The pasture-level portion of the questionnaire did not specifically ask whether the supplemental hay was fed indoors or outdoors. The farm-level results indicated that feeding hay outside was protective, while feeding practices in the barn were not significant. It is possible that mares eating hay outside were eating less pasture forage in areas with high concentrations of caterpillars. The feeding of hay outside was independent of how it was fed, whether on the ground or in racks or other aboveground devices. Many farm managers commented that even though pasture forage was plentiful in April and May, mares ate hay provided to them outside. This was unusual in the managers’ experience. Many managers and veterinarians indicated in May that mares that had an EFL also had signs of sickness (eg, colic, founder, or vaginal discharge) prior to aborting. However, in the present study, these findings were not widespread among the 133 farms investigated. Farm managers and veterinarians voiced concerns about the amount of clover in broodmare pastures that could have been damaged during the mid-April frost; some types of clover contain cyanogenic compounds,12 and concentrations of these can be higher than normal in plants under environmental stress, such as drought or temperature extremes. The percentage of clover in pastures and the percentage of clover combined with mowing practices before, during, and after the April frosts were scrutinized in the analyses, but there were no significant differences between affected and control pastures. The list of management factors not associated with excessive EFLs was of great value to farm managers who were at a loss as to how to manage pastures and grazing times for their horses. This was why many factors of concern brought forward from the industry at multiple informational meetings during May 2001 were included in the questionnaire. One strength of this study was the rapidity with which it was designed, implemented, analyzed, and Scientific Reports: Original Study

617

EQUINE

dates in February 2001, and exposure of mares to cherry trees. The only factor that was significantly associated with a decreased risk of EFLs was feeding hay to mares outside. Of the factors associated with an increased risk of excessive EFLs, exposure to moderate or high caterpillar concentrations was distinctive. This was associated with excessive EFLs at both the farm and pasture levels. It also had 1 of the strongest associations with EFL, as evidenced by the odds ratios. The explosion of Eastern tent caterpillars in Kentucky during April and May 2001 was unusual and may have been a result of the unique weather conditions. Black cherry trees have been grown in Kentucky for hundreds of years, and they are the preferred food source for the caterpillars, although caterpillars also nest in crabapple, apple, pear, and peach trees. Caterpillars often defoliate infested trees, although leaves generally grow back later in the summer.10 For 9 farms where managers indicated they had EFLs or LTAs but no cherry trees on the farm, followup visits revealed infested cherry trees in neighboring pastures, across a road, or a short distance away from mares’ pastures. Additionally, crabapple trees infested with caterpillars were found within 50 feet of the pastures on 1 farm. Caterpillars can travel several hundred feet from their home tree in pursuit of other food sources or during their wandering phase prior to spinning a cocoon, which occurs in late May and early June.10 The strong association between EFLs and cherry trees at the horse level may simply be a reflection that caterpillars were associated with the presence of cherry trees during this time in central Kentucky. The association of farm size with excessive EFLs was likely attributable to the presence of more barren and maiden mares on these farms. Consequently, a larger percentage of mares from these farms were bred during the high-risk time period in February, 2001. A breeding date in February was highly associated with excessive EFLs, and maiden and barren mares are the animals most likely to be bred in this month. Thoroughbred breeding sheds open in midFebruary, because for registration purposes, all foals share a January 1 birth date. Therefore, nonpregnant mares are often bred in February with the expectation that they will foal early during the subsequent year after a 345-day gestation. Mares bred in February would have been approximately 60 days pregnant during what was considered the time of highest risk for exposure to factors leading to EFLs and LTAs (April 18 to 28). At approximately 60 days of gestation, the nutritional source of the fetus is changing from the choriovitelline placenta to the chorioallantoic placenta, a very delicate stage of fetal development.11 This could possibly help explain why mares bred in February were at a much higher risk of EFL than mares bred in late March or early April, which would have been 28 to 45 days of gestation. It seems unlikely that larger farms were associated with higher odds of having excessive numbers of EFLs simply because more mares were at risk in comparison to the previous few years. Broodmare populations on central Kentucky Thoroughbred farms are relatively

EQUINE

02-05-0235.qxd

2/6/2003

1:57 PM

Page 618

reported. Few other large equine emergency epidemiology studies have been reported.13,14 In the face of an emerging threat to the horse industry, > 120 farm visits were conducted by the investigators and a team of volunteers in just 1 week. Data collection during June 2001 also helped minimize memory bias, as data were collected within 1 month of the peak number of EFLs. Less than 2 months after initiation of the survey, a summary of preliminary results was released. Final results were summarized for the industry approximately 2 months later. All of the work reported in the present study was performed concurrently with other activities investigating the cause of MRLS throughout the region. Onsite administration of the questionnaire was used to ensure timely return of the data, allow explanation of questions with farm managers, and clarify responses for unusual circumstances. This also allowed data collectors to observe the farm and evaluate some variables with farm managers as data were collected. For example, the presence of wild cherry trees, which not all managers were able to identify, was often evaluated by data collectors working with the farm managers. Photographs were also taken by the data collectors to maintain a visual record of the pastures in case they were needed later. Some of the limitations of this study were related to the urgency of finding a solution to this outbreak. The choice of using farm managers’ perceptions that EFLs were in excess of those during the past few years as a qualitative outcome was made to allow more farms to be included in the study and to expedite data collection. More quantitative approaches, such as using a cutoff for the number of EFLs that would be considered excessive (eg, a > 20% or > 30% increase compared with previous years), were investigated and not found to have an advantage, considering the wide range in farm size. In retrospect, the approach reported here seems to have worked well. Farms identified by managers as having excessive numbers of EFLs had 36% of mares previously confirmed to be pregnant that were no longer pregnant after ≥ 60 days of gestation, compared with 9% of mares on control farms. Given the large amount of publicity surrounding this outbreak, the many theories on possible causes of the problem, and the work being done to address the problem, it is possible that some of the more subjective observations in the data were susceptible to bias. Information bias caused by differential recall is difficult (if not impossible) to eliminate in this type of retrospective study. However, because there were many theories circulating at the time, it seems unlikely that the net effect of this bias for any given variable would have resulted in odds ratios of the magnitude reported here. Also, recall bias has less potential effect on easier to recall observations (eg, massive numbers or blankets of caterpillars covering fences and water buckets) than on harder to remember observations (eg, how much clover was present in a pasture several weeks ago). Areas for further investigation identified in the present study include determining whether the pathogenesis of MRLS is related to caterpillars, cherry trees, feeding hay, or the presence of waterfowl as causal factors or 618


whether these are markers for other causal factors. Because of the geographic distribution and multiple breeds involved, the inability to identify a primary infectious agent after exhaustive efforts, and the unusual environmental factors that were present in April 2001, researchers recognize that a complex disease led to the reproductive losses associated with MRLS. Additional experimental and observational studies will be needed to clearly identify the causes of MRLS. In an economic impact survey completed at the University of Louisville in September 2001, the cost of MRLS to the Kentucky equine industry from 2000 to 2003 was estimated to be $336 million.15 With this tremendous impact on the horse industry in Kentucky alone, investigations and prospective studies that are underway will continue until the etiology of this costly and complex disease is understood. a

Harrison L, Livestock Disease Diagnostic Laboratory, University of Kentucky, Lexington, Ky: Personal communication, 2001. b Brown S, Hagyard-Davidson-McGee Associates PSC, Lexington, Ky: Personal communication, 2001. c Riddle T, Rood and Riddle Equine Hospital, Lexington, Ky: Personal communication, 2001. d McDowell K, Gluck Equine Research Center, University of Kentucky, Lexington, Ky: Personal communication, 2001. e Epi-Info version 6.0, Centers for Disease Control and Prevention, Atlanta, Ga. f PROC FREQ, SAS/STAT version 8.2, SAS Institute Inc, Cary, NC. g Microsoft Excel 97, Microsoft Corp, Redmond, Wash. h PROC RLOGIST, SUDAAN release 5.50, Research Triangle Institute, Research Triangle Park, NC. i PROC LOGISTIC, SAS/STAT version 8.2, SAS Institute Inc, Cary, NC. The following individuals assisted in the development of the questionnaire used in this study: Drs. Bob Coleman, Barry Fitzgerald, Ralph d’Arge, Barry Meade, Michael Pavlick, and Karen McDowell. Data collectors from the Kentucky USDA offices included Amy Applegate, Richard Broadwater, Garry Cracraft, Grace Halmhuber, Roger Murphy, Mitch Netherly, and Dr. Michael Pavlick. Data collectors from the Kentucky State Veterinarian’s Office included Ron Craig, Tammy Cobb, Dr. Ed Hall, Dennis Hasty, Shane Mitchell, Mike Tobin, and Dr. Chris Young. Volunteer data collectors included Robert Holley, Catherine Wagner, and Drs. Atwood Asbury, James Bowen, Jessica Hoane, David Powell, DL Proctor Sr, H Sutton Sr, and Melissa Newman.

References 1. The Jockey Club 2002 fact book statistical data. Available at: home.jockeyclub.com/FACTBOOK/index.html. Accessed Feb 27, 2002. 2. The Blood-Horse Stallion Register for 2001. Lexington, Ky: The Blood-Horse Inc, 2001. 3. Barents Group LLC. The economic impact of the horse industry in the United States. Vol 2: estimates for eleven focus states. Washington, DC: American Horse Council, 1996. 4. Mare reproductive loss syndrome. University of Kentucky Livestock Disease Diagnostic Center Web site. Available at: www.uky.edu/Ag/VetScience/mrls/harrison.htm. Accessed Feb 27, 2002. 5. Mare reproductive loss syndrome. Sample testing summary results and conclusions. Available at: www.uky.edu/Agriculture/ VetScience/mrls/henning.htm. Accessed Feb 27, 2002. 6. Dwyer RM, Garber L, Traub-Dargatz J, et al. An epidemiological investigation of mare reproductive loss syndrome: breaking ground on a new disease, in Proceedings. Soc Vet Epi Prev Med 2002;44–47. 7. Mare reproductive loss syndrome. University of Kentucky Livestock Disease Diagnostic Center Web site. Available at: www.uky.edu/Ag/VetScience/mrls/lddcdataupdate.htm. Accessed Feb 27, 2002. JAVMA, Vol 222, No. 5, March 1, 2003

02-05-0235.qxd

2/6/2003

1:57 PM

Page 619

8. Mare reproductive loss syndrome. Weather data. Available at: www.uky.edu/Ag/VetScience/mrls/weatherdata.htm. Accessed Feb 27, 2002. 9. Mare reproductive loss syndrome questionnaire. Available at: www.uky.edu/Agriculture/VetScience/mrls/questionnaire.htm. Accessed Feb 27, 2002. 10. Fitzgerald TD. The tent caterpillars. Ithaca, NY: Cornell University Press, 1995;112. 11. Zavy MT, Geisert RD. Embryonic mortality in domestic species. London: CRC Press, 1994;141–152. 12. Pederson GA, Fairbrother TE, Greene SL. Cyanogenesis

and climatic relationships in US white clover germplasm collection and core subset. Crop Sci 1996;36:427–433. 13. McCluskey BJ, Hurd HS, Mumford EL. Review of the 1997 outbreak of vesicular stomatitis in the western United states. J Am Vet Med Assoc 1999;215:1259–1262. 14. Bridges VE, McCluskey BJ, Salman MD, et al. Review of the 1995 vesicular stomatitis outbreak in the western United States. J Am Vet Med Assoc 1997;211:556–560. 15. Thalheimer R, Lawrence RG. The economic loss to the Kentucky equine breeding industry from mare reproductive loss syndrome (MRLS) of 2001. Louisville, Ky: University of Louisville, 2001.

EQUINE

JAVMA, Vol 222, No. 5, March 1, 2003


619