Reasons for performing study: Epiploic foramen entrapment (EFE) has been associated with reduced post operative survival compared to other types of colic but specific factors associated with reduced long-term survival of these cases have not been evaluated in a large number of horses using survival analysis.
Objective: To describe post operative survival of EFE cases and to identify factors associated with long-term survival.
Methods: A prospective, multicentre, international study was conducted using clinical data and long-term follow-up information for 126 horses diagnosed with EFE during exploratory laparotomy at 15 clinics in the UK, Ireland and USA. Descriptive data were generated and survival analysis performed to identify factors associated with reduced post operative survival.
Results: For the EFE cohort that recovered following anaesthesia, survival to hospital discharge was 78.5%. Survival to 1 and 2 years post operatively was 50.6 and 34.3%, respectively. The median survival time of EFE cases undergoing surgery was 397 days. Increased packed cell volume (PCV) and increased length of small intestine (SI) resected were significantly associated with increased likelihood of mortality when multivariable analysis of pre- and intraoperative variables were analysed. When all pre-, intra- and post operative variables were analysed separately, only horses that developed post operative ileus (POI) were shown to be at increased likelihood of mortality.
Conclusions: Increased PCV, increased length of SI resected and POI are all associated with increased likelihood of mortality of EFE cases. This emphasises the importance of early diagnosis and treatment and the need for improved strategies in the management of POI in order to reduce post operative mortality in these cases.
Potential relevance: The present study provides evidence-based information to clinicians and owners of horses undergoing surgery for EFE about long-term survival. These results are applicable to university and large private clinics over a wide geographical area.
Knowledge of factors that are important determinants of horses' chances of survival following surgical treatment of colic provides the clinician and owner (or agent) with evidence-based information enabling informed decisions to be made about treatment and ongoing management of these cases. Information about patterns of survival following hospital discharge, including horses' chances of surviving to certain time points following surgery, provide the owner and horse's usual veterinary surgeon with important additional information.
Epiploic foramen entrapment (EFE) has been associated with reduced long-term survival following surgery compared to other small intestinal lesions (Proudman et al. 2002, 2005). Rates of overall short-term survival of 66–74% and short-term survival of 79–95% of EFE cases have been reported in the literature (Vachon and Fischer 1995; Archer et al. 2004b; Freeman and Schaeffer 2005). Post operative ileus (POI), repeat laparotomy (celiotomy) and elevated post operative heart rate (HR) have been identified as factors influencing short-term survival following small intestinal resection (Morton and Blikslager 2002). Long-term survival of horses recovering from surgery of the small intestine has been shown to be associated with preoperative total protein (TP) and packed cell volume (PCV), duration of surgery and repeat laparotomy (Proudman et al. 2005). Increased mortality of the 50 EFE cases reported in the latter study was in part due to lower TP and longer surgery times in these horses compared to other small intestinal lesions. However, specific determinants of survival within the post operative EFE population have not been evaluated in a large number of cases using survival analysis.
Several studies have reported percentage survival to hospital discharge and factors associated with reduced survival for surgical colic cases in general and for specific types of colic. However, relatively few have provided information about survival beyond hospital discharge. Various endpoints and definitions of long-term survival have also been used, making it difficult to perform comparisons between studies. In addition, most studies have provided information about survival of colic cases undergoing surgical treatment at only one clinic making extrapolation of results to other clinic populations potentially erroneous.
The aim of this study was to describe survival and factors associated with increased likelihood of mortality of EFE cases undergoing surgical treatment in a multicentre, international study. A priori hypotheses were that variables associated with altered survival of horses with other types of colic may be different to those for EFE. We also wished to explore whether crib-biting/windsucking behaviour was associated with altered post operative survival.
Materials and methods
Data were collected from horses presented to clinics collaborating in a prospective, international, multicentre study (Archer et al. 2008) in which a diagnosis of EFE was made at exploratory laparotomy over a 30 month period (January 2004–July 2006) and for whom clinical data could be collected. Clinical data were recorded on a standardised form and included collection of preoperative variables: age, gender, breed, actual or estimated weight and height, heart rate (HR), PCV and TP on presentation. Intra- and post operative data recorded included: length of entrapped small intestine (SI), whether this was considered ischaemic and required resection, length of SI resected, resection type (jejunojejunostomy/jejunoileostomy or jejunocaecostomy), duration of surgery (defined as time from the start of making the incision until closure), and whether the horse developed POI (defined as >2 l of net reflux obtained through a nasogastric tube on at least one occasion and not associated with a known mechanical obstruction) or underwent repeat laparotomy (defined as repeat surgery performed within 7 days of the first surgery) during hospitalisation. Because crib-biting or windsucking (CBWS) behaviour has been shown to place horses at increased risk of EFE (Archer et al. 2008), these data were also recorded to determine whether horses known to display this form of behaviour were at altered likelihood of surviving long-term. Following discharge from the clinic, horses' progress was monitored by periodic (approximately 6 monthly) telephone or email contact with the owner/agent. Horses lost to follow-up (due to being sold or being unable to contact the owner) were censored. Reasons for death or euthanasia and the date of this were also recorded.
Descriptive data were generated for each variable and for causes of death or euthanasia. The percentage of all horses that underwent general anaesthesia that walked out of the anaesthetic recovery room (surgical recovery) and that survived until discharge home (overall short-term survival), to one year (overall one year survival) and to 2 years were calculated (overall 2 year survival). Of those horses that stood following general anaesthesia, the percentage that survived until discharge home (short-term survival), to one year (one year survival) and to 2 years (2 year survival) were also calculated. Horses whose status (alive/dead) was unknown at one or 2 years following surgery were excluded from the relevant calculations.
Survival time was measured as a continuous variable starting from the date of induction of anaesthesia until death or censoring. Categorical variables were examined graphically using Kaplan-Meier plots (Kaplan and Meier 1958). Differences in mortality rate were assessed with univariable Cox proportional hazards models where a likelihood ratio test statistic (LRTS) with a P<0.05 was considered significant. The relationship between continuous variables and survival time was modelled using penalised Cox regression (Dohoo et al. 2003). Exploration of possible nonlinear relationships between these variables and outcome was performed by categorising them into quartiles or other biologically plausible categories, evaluating quadratic fits (Parkin et al. 2005) and by generating smoothing splines for each continuous variable (Therneau and Grambsch 2000). Where variables were shown to be highly correlated (Pearson correlation coefficient >0.9) the most statistically significant variable was selected. Continuous and categorical variables with a univariable P value of <0.2 were used to build a multivariable model. A backward elimination process was used removing the least-significant variable until all remaining variables were statistically significant with a Wald test P<0.05. All remaining variables were then forced back into the model to ensure that no significant or confounding variables had been excluded. To explore the influence of individual clinic on survival, this variable was included as a random (frailty) effect term in the final model. To assess the assumption of proportional hazards, assessment of Schoenfeld residuals was performed (Dohoo et al. 2003). The statistical packages Stata1 and S-plus2 were used for data analysis.
A total of 126 horses that were diagnosed with EFE at exploratory laparotomy were recruited onto this part of the study from 15 clinics in the UK (9 clinics, n = 94 cases), Ireland (2 clinics, n = 9 cases) and the USA (4 clinics, n = 23 cases). Fifteen horses were lost to follow-up (due to sale of the horse or not being able to contact the owner) between the start of the study and the final point of contact (between January 2007–November 2008). Reasons for mortality and the numbers of each are detailed in Table 1. The proportion of horses that survived to different endpoints following surgery is given in Table 2. Of horses that developed POI (n = 36) and underwent repeat laparotomy (n = 16), 58.3 and 50.0% were discharged home, respectively. For horses that were discharged from the hospital and whose status was known at one and 2 years post operatively, 70.7 and 53.7% were still alive one and 2 years post operatively, respectively.
Table 1. Reasons for mortality or euthanasia in 63 horses among 126 horses that underwent exploratory laparotomy for treatment of epiploic foramen entrapment (EFE) in 15 clinics in the UK, Ireland and USA
Reason for death/euthanasia
Number of horses (%)
Euthanasia - poor prognosis/financial reasons
Euthanasia - hopeless prognosis
Death - haemorrhagic shock
Death - endotoxic shock
During recovery from anaesthesia
Euthanasia - fractured leg
Euthanasia/death - unable to stand due to endotoxic shock
During the hospitalisation period
Post operative ileus
Hepatic failure/hepatic encephalopathy
Following hospital discharge
Laminitis associated with colic
Unrelated accident or illness
Table 2. Proportion of horses that survived following surgery for treatment of EFE (see text for definitions)
Definition of survival
Percentage survival (%)
For all horses undergoing surgery (n = 126)
Overall short-term survival
Overall one year survival
Overall 2 year survival
For all horses that walked out of the anaesthetic recovery box (n = 109)
Survival to one year
Survival to 2 years
Figure 1 shows a Kaplan-Meier plot of cumulative probability of survival. The median survival time (i.e. the time at which 50% of horses were still alive) for all horses with EFE was 397 days. Kaplan-Meier plots of categorical variables with a univariable P value of P<0.2 are shown in Figure 2. Those with strangulated SI requiring resection (median survival 125 days), that developed POI (median survival 33 days) and underwent repeat laparotomy (median survival 15 days) had reduced survival compared to horses with nonstrangulated SI, that did not develop POI or undergo repeat laparotomy (median survival >1400 days for each of these categories). Jejunocaecostomy was also associated with reduced survival (median survival 337 days) compared to jejunojejunostomy/jejunoileostomy (median survival 1003 days) and both had reduced survival compared to horses in which no resection was required (median survival >1400 days). CBWS behaviour was not associated with likelihood of survival. The functional form of the relationship between continuous variables with a univariable P value of <0.05 and survival is shown in Figure 3. These demonstrate increased likelihood of mortality with increasing PCV, HR, length of entrapped SI, length of resected SI and increased duration of surgery. These variables were shown to fit each model best in a linear form. Univariable associations between mortality and categorical or continuous variables are reported in Table 3.
Table 3. Univariable associations between potential explanatory variables and likelihood of mortality following surgery for epiploic foramen entrapment (EFE)
LRTS P value
LRTS: likelihood ratio test statistic; TB: Thoroughbred; WB: Warmblood.
Heart rate (beats/min)
Packed cell volume (%)
Total protein (g/l)
Length of small intestine entrapped (cm)
Length of small intestine resected (cm)
Duration of surgery (min)
Small intestine strangulated
Post operative ileus
Increasing PCV and length of SI resection were significantly associated with increased likelihood of mortality when multivariable analysis of pre- and intraoperative variables were analysed (Model 1). When post operative variables were included in the multivariable model, only the development of POI was significantly associated with increased likelihood of mortality (Model 2). To illustrate the results of Model 1, a horse with a preoperative PCV of 30% and no small intestinal resection would have an expected median survival time of >1000 days and a probability of 0.76 and 0.74, respectively, for survival to one and 2 years post operatively. Alternatively a horse with a PCV of 40% and 3 m of small intestine resected would have an expected median survival time of 735 days and probability of survival to one and 2 years of 0.55 and 0.5, respectively. For a horse with a PCV of 48% and resection of 6 m, median survival time would be 35 days and probability of survival to one and 2 years post operatively would be 0.33 and 0.24, respectively.
Because the 4 horses that were subjected to euthanasia on the operating table due to a poor prognosis and/or for financial reasons could potentially have survived had surgical treatment been completed, the models were re-run excluding these cases. This had no significant effect on either model. Inclusion of clinic identification number as a random effect term had no significant effect on Model 1 (P = 0.24) but could not be fitted to Model 2. Graphical and statistical evaluation of Schoenfeld residuals confirmed the assumption of proportional hazards to be valid for the variables PCV and POI. There was some evidence of nonproportionality of the variable length resected, with a suggestion of reduced risk over time. A time x length resected interaction term was tested and although significant, the effect was very small.
The present study provides data about survival following surgery for a very specific type of colic (EFE) in a multicentre, international study. To our knowledge, this is the first study investigating post operative survival following colic surgery in multiple clinics and on more than one continent. The results of this study can be used by clinicians to provide owners with information about likely survival of individual cases of EFE and assist informed decision making about the treatment of these cases.
When pre- and intraoperative variables were evaluated in the present study, increased mortality was associated with increasing PCV and increasing length of intestine resected. Preoperative PCV has been shown to be associated with reduced long-term survival for other colic types (Proudman et al. 2006). Increasing duration of surgery was also identified as a risk factor for long-term survival of horses recovering from surgery of the small intestine by Proudman et al. (2005). This variable was also significantly associated with survival in the present study; however duration of surgery was highly correlated with length of intestine resected and the latter was chosen for incorporation into the multivariable model as a better fit (Table 4). It is likely that PCV and length of intestinal resection were not retained in Model 2 for several reasons. Horses that died intraoperatively or in the first 24 h post operatively would not have had the chance to develop POI and were therefore classified as unknown for this variable. Also, preoperative PCV was not obtained for 18% of horses resulting in different numbers of horses being included in each model. In addition, increased PCV, increased duration of surgery/anaesthesia and intestinal resection have been identified as risk factors for POI in several studies (Blikslager et al. 1994; Roussel et al. 2001; Cohen et al. 2004). POI has also been identified as a risk factor for post operative mortality in horses that have undergone small intestinal resection (Morton and Blikslager 2002). Increased risk of mortality of EFE cases compared to other forms of small intestinal colic has been associated with lower preoperative total protein values (Proudman et al. 2005). This variable was not shown to be associated with survival in the present study.
Table 4. Multivariable Cox proportional hazards models for likelihood of mortality following surgery for epiploic foramen entrapment
Wald test P value
Model 1 incorporates pre- and intraoperative variables only, and Model 2 all variables.
Multivariable Model 1 (n = 98)
Length resected (cm)
Multivariable Model 2 (n = 103)
Post operative ileus
The present study re-emphasises the importance of early diagnosis and referral of cases of EFE prior to the development of marked ischaemic change of entrapped bowel minimising the subsequent increase in PCV, potential need for resection (and possible increased length of resection required) and consequent increased risk of POI. It also highlights the need for improved strategies in the management of POI in order to reduce post operative mortality in EFE cases and other forms of small intestinal strangulation. The definition of POI used in the present study is similar to that used in other studies and was chosen to minimise the chance of misclassifying cases of post operative mechanical obstruction as POI. This issue was discussed by Freeman (2008) and Merrit and Blikslager (2008), where criteria for POI were given. These criteria should be used in future studies investigating POI to provide greater consistency between studies. In addition, there is a need for prospective, well designed and rigorously conducted multicentre studies to determine the effectiveness of current and new therapies for POI.
The median survival time of horses that had a jejunocaecostomy in the present study was much shorter compared to the jejunojejunostomy/jejunoileostomy group and horses in which no resection was performed. However, resection type was not significantly associated with likelihood of survival nor were horses that had a jejunocaecostomy performed more likely to develop POI compared to the jejunojejunostomy/jejunoileostomy group. Whilst horses that are known to display CBWS behaviour are at increased risk of EFE (Archer et al. 2004a,b, 2008), this behaviour was not associated with altered risk of post operative survival.
Survival time was measured in the present study from the time of induction of general anaesthesia, thereby incorporating horses that died during, or failed to recover following, anaesthesia. This provides useful information for clinicians and owners in addition to that provided for horses that walked out of the anaesthetic recovery box. Mortality in the present study, as in most colic survival studies, was usually due to euthanasia rather than death of the horse. The most common reasons for euthanasia in the present study were colic that was nonresponsive to analgesia and where further surgical intervention was declined by the owner and in horses in which POI failed to resolve.
Use of standardised terms relating to survival of horses undergoing surgical treatment of colic is important, particularly when making comparisons between multiple studies. Whilst short-term survival following colic surgery is commonly reported as survival to hospital discharge, it is important to report whether this includes all horses undergoing surgical treatment (overall short-term survival) or only those that recovered following surgery (short-term survival) as the percentage of each can differ greatly. The short-term survival rates in the present study are similar to those from other large EFE case series (Vachon and Fischer 1995; Archer et al. 2004a,b). However, long-term survival is less frequently reported for colic cases, including cases of EFE, and its definition varies between studies making comparisons difficult. This issue has also been identified as problematic in human survival studies, particularly those investigating relatively rare conditions in which such data from multiple centres are valuable in assessing whether particular treatments may be associated with altered survival (Stephens et al. 1996). Survival analysis also enables standardised data to be reported; the median survival time of EFE cases in the present study is similar to that reported from a smaller cohort of horses by Proudman et al. (2002).
Multicentre, prospective studies provide data that are applicable to a broader population (rather than specific to one clinic or a few surgeons) and enable specific variables of interest to be collected more reliably, compared to retrospective studies, from a larger number of cases over a shorter time period. This is particularly useful in less common diseases, such as EFE, which accounts for a relatively small proportion of colic cases in most clinics and where treatment of cases (e.g. anaesthetic and post operative treatment regimes) can change over time. The issue of clustering of data (e.g. influence of clinic on outcome) can also be evaluated in studies of this type; in the present study, clinic had no effect on outcome. Collection of other variables that have been associated with altered post operative survival, such as blood or peritoneal fluid lactate (which was measured by only some clinics), was not undertaken nor were other post operative variables recorded as they could not be reliably or easily obtained by all clinics and may have introduced issues relating to interclinic variability. Multicentre studies are time consuming and require sufficient funding to enable proper coordination with clinic personnel and contact with owners. In addition, it is important to be aware of issues relating to clinic and client confidentiality. However, collection of systematic data from contributing centres in multiple countries has important potential benefits to equine surgeons, their patients and the owners or trainers of these horses (Mair and White 2005).
The present study provides data about survival following surgery for EFE that are applicable to university and large private clinics over a wide geographical area. Increasing PCV and increasing length of small intestine resected were significantly associated with increased likelihood of mortality in a multivariable model incorporating pre- and intraoperative variables. When all pre-, intra- and post operative variables were incorporated, POI was the single variable significantly associated with reduced survival. This, together with probability of survival to certain time points and median survival times, provides useful evidence-based information for clinicians and owners.
Authors' declaration of interests
No conflicts of interest have been declared.
Source of funding
We are very grateful to the Horserace Betting Levy Board who funded Debra Archer's Research Training Scholarship.
We are extremely grateful to all the clinicians and support staff at all the collaborating clinics that participated in this colic epidemiology study, including those who were involved in this part of the study. In particular, we thank our colleagues at Liverpool University, Louise Southwood and Cindy Brockett (New Bolton Centre), Tim Mair, David Sinclair and Kelly Russell (Bell Equine Clinic), Richard Payne and the surgery team (Beaufort Cottage Equine Hospital), Marie Harty, Mary Wade and Catherine McAvinney (Anglesey Lodge Equine Hospital), Henry Tremaine, Nigel Woodford, Wendy Atkinson and Sue Thomas (Bristol University), Jorge Nieto (Davis University), Bruce Blaydon and Sam Oates (Donnington Grove Veterinary Surgery), Lance Voute and team (Glasgow University), Justin Perkins, Mike Archer and Matt Smith (Royal Veterinary College), John Walmsley, Belinda May, Tim Phillips, Jane Boswell, David Lloyd and Giles Summerhays (The Liphook Equine Hospital), David Rendle (Liphook Equine Hospital and Glasgow University), Anthony Blikslager and Matt Gerard (North Carolina State University), Jennifer Brown, Nathaniel White and colleagues (Marion duPont Scott Equine Medical Center), Mark Hillyer (The Newmarket Equine Hospital) and Warren Schofield, Juan Perez Olmos, Hugh Dillon and Geraldine Carty (Troytown Equine Clinic). We are also very grateful to the owners/carers of horses for all their help with the study.