Association of admission L-lactate concentration in hospitalised equine neonates with presenting complaint, periparturient events, clinical diagnosis and outcome: A prospective multicentre study

Authors


  • Dr Borchers is currently affiliated with the University of California, Davis, California, USA. Dr Pantaleon is currently affiliated with Equine Medical Associates, Lexington, Kentucky, USA. Dr Levy is currently associated with the University of Calgary, Alberta, Canada. Dr Marsh is currently located in Lexington, Kentucky, USA.

Summary

Reasons for performing the study: Admission L-lactate concentration is a useful and commonly measured biomarker not previously prospectively evaluated in a large multicentre study of critically ill neonatal foals.

Objectives: To evaluate overall outcome and the association of survival and L-lactate concentration at admission ([LAC]ADMIT) by periparturient history, presenting complaint and clinicians' major diagnosis for ill neonatal foals.

Methods: Thirteen university and private equine referral hospitals enrolled 643 foals over the 2008 foaling season. Case details, historical, clinical and clinicopathological data were entered into standardised spreadsheets then unified for analysis.

Results: Overall survival was 79% (505/643). Risk of nonsurvival increased with each 1 mmol/l increase in [LAC]ADMIT (odds ratio 1.14, P<0.001). Mean arterial pressure had a small (r2= 19.1) but significant (P<0.001) association with [LAC]ADMIT. Foals experiencing known dystocia or premature placental separation had increased [LAC]ADMIT (P<0.001). Single umbilical problems (excluding uroperitoneum), meconium impaction only and failure of passive transfer of immunity only had 100% survival. Six clinicians' major diagnoses had increased odds of nonsurvival for each 1 mmol/l increase in [LAC]ADMIT: ‘sepsis’; ‘unspecified enterocolitis’; ‘unspecified colic’; ‘unspecified trauma’; ‘immune related (not failure of passive transfer of immunity)’ and ‘respiratory only’.

Conclusions and potential relevance: Survival of critically ill foals is good but varies with peripartum history, presenting complaint and clinicians' major diagnosis. L-lactate concentration at admission proves its utility as a valuable prognostic biomarker in neonatal foals and its utility appears to vary with peripartum history and clinicians' major diagnosis.

Introduction

Increased blood or plasma L-lactate concentration ([LAC]) is commonly associated with both disease severity and mortality in critically ill human adult and neonatal patients (Manikis et al. 1995; McNelis et al. 2001; Levraut et al. 2003; Nadeem et al. 2010; Srinivasjois et al. 2010) as well as in adult and neonatal equine emergencies (Corley et al. 2005; Johnston et al. 2007; Henderson et al. 2008; Wotman et al. 2009; Castagnetti et al. 2010; Tennent-Brown et al. 2010).

Increased disease severity and mortality of patients with hyperlactaemia remains a persistent feature regardless of whether [LAC] was evaluated as a single, one time measurement (Howell et al. 2007; Trzeciak et al. 2007; Mikkelsen et al. 2009) or sequentially over a period of time (Vincent et al. 1983; Nguyen et al. 2004; Wotman et al. 2009; Castagnetti et al. 2010; Tennent-Brown et al. 2010). One recent study (Nadeem et al. 2010) demonstrated that a serum lactate concentration of >5.8 mmol/l obtained within 24 h of birth in preterm human infants (<32 weeks of gestation) had a sensitivity of 100% and specificity of 85% forpredicting adverse outcomes, defined as death, severe intraventricular haemorrhage or leucoencephalomalacia.

Most importantly, the prognostic value of [LAC] in critically ill human patients may be unrelated to the underlying disease (Bakker 2001) or the presence of shock or organ failure (Mikkelsen et al. 2009). The role of [LAC] as a biomarker and its association with clinical disease is complex and many questions remain to be investigated.

Hyperlactaemia arises from an imbalance between oxygen delivery (DO2) and tissue demand (type A hyperlacataemia). However, hyperlactaemia under conditions of adequate DO2– as seen with impaired oxygen utilisation secondary to inflammatory responses and/or reduced clearance as in liver or renal failure – is also recognised, particularly among the critically ill (type B hyperlacataemia) (Cohen and Woods 1976; Luft et al. 1983; De Backer et al. 1995; Bakker and Jansen 2007). It is likely that some critically ill foals experience both type A and type B hyperlactaemia.

Several recent retrospective, and one prospective, smaller single centre studies of [LAC] in sick neonatal foals have consistently shown increased [LAC] at admission ([LAC]ADMIT) and at additional time periods during hospitalisation is associated with nonsurvival, suggesting that either impaired L-lactate removal from the circulation or increased lactate generation is present in nonsurvivors (Corley et al. 2005; Henderson et al. 2008; Wotman et al. 2009; Castagnetti et al. 2010). Prospective evaluation of [LAC] in adult equine emergencies showed comparable results (Tennent-Brown et al. 2010).

Lactate concentration at admission has not been evaluated prospectively in a large multicenter study of critically ill equine neonates. Thus the main objective of this study was to investigate the association of [LAC]ADMIT with presenting complaint, periparturient events, outcome and clinicians' major diagnosis in a prospective multicentre approach. We hypothesised that increased [LAC]ADMIT is associated with mortality, but we also hypothesised that admission hyperlactaemia present with some clinicians' major diagnoses would be statistically associated with nonsurvival. The second aim of the study was to evaluate overall survival and outcome data, in addition to evaluating survival by periparturient history, presenting complaint and clinicians' major diagnosis category, in critically ill foals presenting for veterinary care at referral hospitals without regard to [LAC]ADMIT.

Materials and methods

Thirteen university and private equine referral hospitals worldwide were recruited for participation in this prospective study. All participating study sites routinely measure [LAC]ADMIT in admitted equine neonates, in addition to collecting and recording data associated with history, case details and other clinical and clinicopathological information. Neonatal foals aged 0–35 days were included during the 2008–2009 foaling season. Background, historical, clinical and clinicopathological data were entered into a standardised spreadsheet by a single veterinarian at each institution. Data from each participating institution were then unified into a single data set for analysis and included those foals where admission lactate concentration or other information was not obtained for a variety of reasons, including equipment malfunction or oversight. Data recorded, when known, included: gestational age, age at presentation, sex, weight, breed, information about parturition (dystocia, premature placental separation, other), presenting complaint, direct/indirect blood pressure, [LAC]ADMIT, blood culture results, clinicians' major as determined by the attending clinician, outcome and days hospitalised. Survival was defined as discharge from the hospital. The nonsurvival group included foals that died or that were subjected to euthanasia due to clinician determined poor prognosis or financial reasons. The clinicians' major diagnosis was determined by the recording veterinarian at each institution and was considered the single diagnosis most responsible for the foal's condition in the opinion of that veterinarian, even if other contributing diseases/problems were present. There were 18 clinicians' major diagnosis groups. Not all data, including [LAC]ADMIT, were available for all foals.

Arterial or venous blood samples were collected on admission using sampling techniques specific to each [LAC] analyser and analysed immediately for [LAC] with a blood gas/lactate analyser (Novaa, Accu-checkb, Accutrendb, Accu-check Aviva 400c, Lactate Scoutd, ABL 700e). No attempt was made to ensure that all sites used similar sampling techniques or similar lactate analysers. Similarly, mean arterial pressure (MAP) was determined by direct or indirect pressure measurement at some sites with no effort made to control techniques used.

Data analysis

Descriptive statistical analyses including mean, standard deviation, median and range were used to summarise the data. All analyses were performed by commercially available statistics software (Istatf, Stata 9.0g, Minitab version 13.0h, GraphPad instat 3.1ai).

Data were tested for normality using the Kolmogorov and Smirnov method. A nonparametric approach using Mann-Whitney was used to compare variables between groups, and Kruskal-Wallis, a nonparametric ANOVA test, was used to examine differences among different populations of foals (such as different diagnostic groups) followed by Dunn's multiple comparison tests. Specifically, the Kruskall-Wallis test was used to test the hypothesis that [LAC]ADMIT differed between foals of different clinicians' major diagnosis categories. Multinomial logistic regression models were used to examine the associations between [LAC]ADMIT and specific presenting complaints and to examine the associations between [LAC]ADMIT and survival within the population and within final diagnosis sub-groups ‘diarrhoea’ as the base (default) outcome due to it being the most frequent presenting complaint. The exposure examined was [LAC]ADMIT; interactions were not explored due to the limited numbers within various presenting complaints. As [LAC]ADMIT was not normally distributed, data were log transformed in order to achieve a normal distribution prior to performing logistic regression for determination of odds ratios (OR) for nonsurvival for each 1 mmol/l change in [LAC]ADMIT. Receiver operator curve (ROC) analysis was performed in an effort to determine an optimal [LAC]ADMIT cut-point, above which nonsurvival could be more accurately predicted. (Hanley and McNeil 1982; DeLong et al. 1988) A variety of cut points either side of 0.5 were explored and tested in regard to the Hosmer and Lemeshow Goodness of fit test (Hosmer et al. 1988). Significance was determined at P≤0.05.

Results

Of 643 foals enrolled in the study, 588 (91.4%) had [LAC]ADMIT recorded. There were more colts (n = 346) than fillies (n = 277) included in the study. The sex was not recorded in 20 foals. Horse breeds represented were 420 (65%) Thoroughbreds, 94 (14.6%) Standardbreds, 31 (4.8%) Warmbloods, 28 (4.4 %) Quarter Horses or Paints, 20 (3.1 %) Arabs, 15 (2.3%) Pony breeds, 14 (2.2%) Horses of ‘Other’ breeds, 8 (1.2 %) Mixed breed horses, 5 (0.7 %) Draft breeds and 2 (0.3 %) Donkeys. Breed was not recorded in 6 cases.

Overall survival rate was 79% (505/643). Nonsurvivors (n = 129) included foals subjected to euthanasia for poor prognosis determined by the clinician (n = 81) or financial reasons (n = 6). The foals subjectedto euthanasia for financial reasons had clinicians' major diagnoses of prematurity (n = 2), unspecified colic (n = 2), sepsis (n = 1) and perinatal asphyxia syndrome (PAS) (n = 1) and were included in summary data as they were fairly well distributed throughout the diagnostic categories. None were subjected to euthanasia at admission without attempting stabilisation and establishment of a diagnosis and severity of condition that carried a poor prognosis for life. These foals were not, however, included in ROC analysis. Foals without [LAC]ADMIT were not included in relative risk, logistic regression or ROC analysis. Forty-eight foals died.

There was no difference found between participating referral hospitals for either survival percentage (P = 0.100) or [LAC]ADMIT (P = 0.186) (Fig 1). Mean age at admission was 53 ± 16 h (range 0–840 h). Age at admission was not significantly different between survivors and nonsurvivors (55 ± 111, median 20 h, range 0–840 h vs. 47 ± 92 h, median 17 h, range 0–720 h, P = 0.419). Average gestational age was 338 ± 16 days (range 280–395 days). Gestational age varied significantly between survivors and nonsurvivors (339 ± 15 days, median 340 days, range 280–395 days vs. 331 ± 16 days, median 330 days, range 298–390 days, P<0.001). Foals born from known dystocia, even if not the clinicians' major diagnosis (16%, n = 95) had significantly higher [LAC]ADMIT than foals without known dystocia (n = 493) (6.9 ± 4.4 mmol/l, median 6.1 mmol/l, range 6.1–20.3 mmol/l vs. 4.9 ± 4.4 mmol/l, median 3.6 mmol/l, range 0.3–36.6 mmol/l, P<0.001). The same was true for foals born from premature placental separation (11%, n = 65) compared with those born without premature placental separation (n = 516) (7.8 ± 5.8 mmol/l, median 6.7 mmol/l, range 0.6–31.7 mmol/l vs. 4.9 ± 4.1 mmol/l, median 3.6 mmol/l, range 0.3–36.6 P<0.001). Median duration of hospitalisation for survivors and nonsurvivors was 6 days (0.25–60 days) and 2 days (0–23 days), respectively (P<0.001).

Figure 1.

Top panel depicts total number of foals enrolled by site and number of survivors/nonsurvivors by site. There were no differences found in survival percentage by site. Lower panel depicts median L-lactate concentration at admission ([LAC]ADMIT) by site. Again, no differences were found in median [LAC]ADMIT by site. Site C contributed the most foals (n = 146) while Site H contributed the fewest (n = 4). The largest mean [LAC]ADMIT was from Site D (6.2 mmol/l) while the lowest was from Site L (3.6 mmol/l).

The most common presenting complaint (Table 1) was diarrhoea, followed by ‘dummy’ foal, recumbency, caesarean section/dystocia/high risk pregnancy, colic, lethargy, not nursing, premature, orthopaedic problems, suspected sepsis, colostral related, respiratory problems, umbilical related problems, trauma, failure of passive transfer/requiring blood, rejected by mare, meconium impaction, neurological/seizures and dysphagia. L-lactate concentration at admission was different between presenting complaints (P<0.001). Mean arterial pressure on presentation was 69.0 ± 13.3 mmHg (n = 158) (Fig 2). Average MAP was significantly different between surviving and nonsurviving foals (71 ± 13 mmHg, median 71 mmHg, range 37–101 mmHg vs. 64 ± 14 mmHg, median 66 mmHg, range 34–91 mmHg, P = 0.001). There was a small (r2= 19.1) but significant (P<0.001) association of MAP with [LAC]ADMIT.

Table 1. L-lactate concentration at admission ([LAC]ADMIT) mmol/l, per cent survival to discharge and relative risk rate (RRR) compared with ‘Diarrhoea’ by presenting complaint category
Presenting complaintN [LAC]admit / N total[LAC]admit mmol/l median (range)Relative risk rate (95% CI)Survival alive/total (%)
  1. The RRR is per 1 mmol/l change in [LAC]ADMIT compared with that of ‘Diarrhoea’. Significant differences were present by presenting complaint(Kruskal-Wallis, P<0.001; Bonferroni test: medians significantly different if z-value >3.6491). Neuro/seizure and Dysphagia categories were not evaluated statistically due to low N. N = number of samples; FPTI = failure of passive transfer of maternal immunity; C-section = caesarean section; HRP = high risk pregnancy; Neuro = neurological disease. a = different from recumbency; b = different from colic; c = different from diarrhoea; d = different from uroperitoneum; e = different from orthopaedic; f = different from sepsis; g = different from C-section/dystocia/HRP. *= significant(P<0.05) increased or decreased relative risk of nonsurvival compared with diarrhoea.

Umbilical 9/144.1(1.4–36.6)1.29(1.14–1.47)*13/14(93)
Recumbency 69/705.7(0.6–34.7)abe1.26(1.15–1.38)*56/70(80)
Dummy 67/735.8(1.6–19.2)bcdf1.23(1.12–1.35)*56/73(77)
C-section/dystocia/HRP 47/565.4(1.7–18.2)cefg1.22(1.11–1.35)*43/56(77)
Colostral related 14/163.8(1.1–13.8)1.17(1.01–1.36)*15/16(94)
Lethargy 46/474.1(0.9–20.8)1.16(1.04–1.28)*29/47(62)
Not nursing 46/463.6(0.8–22.0)1.12(1.01–1.26)*21/46(89)
Premature 37/414.8(0.9–20.4)1.14(1.02–1.28)*22/41(54)
Neuro/seizures 1/25.11.13(0.66–1.94)2/2(100)
Trauma 9/114.3(2.0–10.9)1.09(0.89–1.35)8/11(73)
Respiratory 13/143.4(1.1–9.9)1.05(0.86–1.28)13/14(93)
Rejected 8/104.2(3.4–5.7)1.04(0.81–1.34)10/10(100)
Diarrhoea 103/1153.2(0.3–20.0)ac1.00105/115(91)
FPTI/plasma transfusion 10/112.5(1.1–13.2)1.01(0.79–1.29)10/11(91)
Colic 53/564.0(0.7–10.0)bg0.88(0.75–1.02)48/56(86)
Orthopaedic problem 18/222.4(1.3–9.6)e0.89(0.70–1.12)17/22(77)
Sepsis(suspected) 18/202.0(0.9–13.2)ag0.86(0.67–1.10)12/20(60)
Meconium impaction 7/72.8(1.4–5.3)0.79(0.51–1.23)7/7(100)
Uroperitoneum 9/91.4(0.9–6.6)ad0.55(0.31–0.96)*9/9(100)
Dysphagia 1/20.90.03(0.00–5.26)2/2(100)
Figure 2.

Fitted line regression curve for L-lactate concentration at admission ([LAC]ADMIT) vs. mean arterial pressure. P<0.001.

Admission blood culture results were available for 466 (72.5%) foals, with 346/466 (74.2%) reported as negative and 120/466 (25.8%) reported as positive. Fifty-three (44.2%) isolates were Gram positive, whereas 73 (60.8%) were Gram-negative isolates. Isolated bacteria include E. coli (32), Streptococcus spp. (15), Enterococcus spp. (15), Staphylococcus (12), Pantoea agglomerans (9), Enterobacter spp. (7), Bacillus spp. (7), Clostridium spp. (6), Klebsiella (5), Actinobacillus spp. (5), Pasteurella spp. (3), Diphteroids (3), Pseudomonas spp. (2), Coryneform spp., Salmonella spp., Aeromonas hydrophilus, Arcanobacterium pyogenes, Aerococcus viridans (one each). Positive blood culture was significantly associated with nonsurvival (P<0.001); however, [LAC]ADMIT was not significantly different in blood culture positive vs. negative foals when considered over the entire study population (5.3 ± 4.6 mmol/l, median 4.4 mmol/l, range 0.8–36.6 mmol/l vs. 5.0 ± 3.9 mmol/l, median 3.7 mmol/l, range 0.3–20.8 mmol/l).

After admission and diagnostic testing, the foals were placed in one of 18 clinicians' major diagnosis groups based on which diagnostic category most represented the primary problem as determined by the attending clinician (Table 2). Perinatal asphyxia syndrome, unspecified enterocolitis and sepsis were the most frequent clinicians' major diagnoses while muscle disease and umbilical (not uroperitoneum) were the least common. Median [LAC]ADMIT was significantly (P<0.001) increased in nonsurvivors compared with survivors in those foals where it was measured (n = 588/643; 91%) (Table 2). L-lactate concentration at admission was statistically associated with nonsurvival in some specific clinicians' major diagnoses but not all (Table 2). Sepsis, unspecified enterocolitis, unspecified colic, trauma, immune related (not failure of passive transfer of immunity [FPTI]) and respiratory diagnoses each carried an increased risk of nonsurvival for each 1 mmol/l increase in [LAC]ADMIT.

Table 2. L-lactate concentration at admission ([LAC]ADMIT) mmol/l and per cent survival to discharge by final diagnosis category
Final diagnosisN[LAC]admit mmol/l median (range)[LAC]admit mmol/l mean ± s.d.Survival(%)Odds ratio OR (95% CI)OR P value
  1. Not all foals had both [LAC]ADMIT(N = 586) and outcome(survival vs. nonsurvival)(N = 643) recorded and this table represents only those foals with both recorded. ‘N’ for each diagnosis is the total number of foals with both [LAC]ADMIT recorded and outcome that received that final diagnosis from the attending clinician. Median perinatal asphyxia syndrome(PAS) [LAC]ADMIT was different from that of both unspecified enterocolitis(P<0.001) and meconium impaction only(P<0.01). Overall, an increase of 1 mmol/l in [LAC]ADMIT carried a 14% increased risk of nonsurvival(OR = 1.14). Sepsis, unspecified enterocolitis, unspecified colic, trauma, immune related(not FPTI) and respiratory diagnoses each carried a significant increased risk of nonsurvival for each 1 mmol/l increase in [LAC]ADMIT ranging from 21–282%. N = number of samples; N# =[LAC]ADMIT >2.2 mmol/l perfectly predicted outcome; NA = not applicable, in the case of OR due to either small N or lack of nonsurvivors; superscript a or b= indicates significant difference between diagnoses groups. *Significantly increased risk of nonsurvival per 1.0 mmol/l increase in [LAC]ADMIT.

All foals 5883.9 (0.3–36.6)5.2 ± 4.3791.14 (108–1.20)*<0.001*
Survivors 4683.6 (0.3–36.6)4.6 ± 3.7NANANA
Nonsurvivors 1205.5 (0.9–31.7)*7.4 ± 5.6NANANA
Prematurity/dysmaturity 424.6 (0.9–16.9)5.2 ± 3.3531.07 (0.90–1.28)0.439
Sepsis 1004.3 (0.6–31.7)5.4 ± 4.9591.21 (1.08–1.35)<0.001*
Perinatal asphyxia syndrome 1445 (0.9–20.4)a6.3 ± 4.2881.09 (0.99–1.19)0.074
Failure of passive transfer ONLY 203.9 (1.1–6.6)3.7 ± 1.6100NANA
Uroperitoneum 112.6 (0.9–13.9)3.9 ± 3.9731.10 (0.78–1.34)0.591
Unspecified enterocolitis 1113.1 (0.6–20)ab4.0 ± 3.4931.28 (1.12–1.46)(<0.001)*
Meconium impaction ONLY 263 (0.9–5.3)ab3.1 ± 1.0100NANA
Unspecified colic 182.8 (0.7–20.8)4.6 ± 5.1751.21 (1.02–1.43)0.028*
Trauma 183.9 (0.9–20.3)6.0 ± 5.4721.45 (1.10–1.91)(0.008)*
Orthopaedic 154.3 (1.4–18.2)6.3 ± 4.8680.97 (0.77–1.21)0.771
Immune related (not FPT) 164.6 (1.1–13.8)5.9 ± 3.6941.54 (1.04–2.28)0.032*
Respiratory 114.4 (1.4–11.3)5.6 ± 3.6422.82 (1.48–5.38)0.002*
Muscle disease 31.8 (1.6–4.6)2.7 ± 1.7100NANA
Unspecified congenital deformity 67.7 (2.2–18.3)8.6 ± 6.117NA#NA#
High risk pregnancy, difficult delivery (c-section; dystocia) 125.3 (1.9–13.2)5.7 ± 2.9100NANA
Umbilical (not uroperitoneum) 32.9 (1.4–9.6)4.6 ± 4.4100NANA
Dysphagia/aspiration 74.9 (0.9–14.5)5.5 ± 4.3780.68 (0.33–1.40)0.294
Other 262.5 (0.3–36.6)5.1 ± 7.4691.03 (0.91–1.17)0.599

Foals born following a known dystocia most commonly received a clinician's diagnosis of perinatal asphxial syndrome (n = 33), followed by orthopaedic (n = 9), and sepsis and trauma (8 each). Six were categorised by the clinician as high risk pregnancy (HRP)/difficult delivery. Dysphagia, unspecified enterocolitis and FPTI only had 5 foals each assigned, while 4 foals were categorised as premature, 3 foals each as respiratory only and other, 2 foals each as unspecified congenital deformity and uroperitoneum, and one foal each for meconium impaction and patent urachus. Foals born following a known premature placental separation (‘red bag’) delivery most commonly received a clinician's diagnosis of PAS (n = 27), followed by sepsis (n = 17) and prematurity (n = 10). Two foals were categorised by the clinician as trauma while uroperitoneum, respiratory only, orthopaedic, unspecified colic and unspecified enterocolitis had one foal each.

Overall, a cut point of 4.43 mmol/l was found to maximise both sensitivity (0.63) and specificity (0.63) to distinguish survivors from nonsurvivors (Fig 3). When [LAC]ADMIT was dichotomised into either above or below 4.43 mmol/l [LAC]ADMIT correctly classified 80% of foals as either survivors or nonsurvivors.

Figure 3.

Receiver operator curve (ROC) for L-lactate concentration at admission ([LAC]ADMIT) as a predictor of nonsurvival. A cut point of 4.4 mmol/l was determined to maximise both sensitivity (63%) and specificity (63%) and correctly classified 80% of foals enrolled in the study.

Discussion

Perhaps the most important finding of this study is the overall large survival rate of 79% for foals presenting to neonatal intensive care units throughout the world, in addition to the uniformity of that survival rate across multiple centres providing such care. One single centre study evaluating long- and short-term outcome in 131 ill foals that were aged <7 days when admitted to the University of Florida Veterinary Medical Teaching Hospital between 1981 and 1983 (Baker et al. 1986) had 54% survival to discharge. Another single centre study evaluating prognostic variables reported an overall survival of 66% in a group of 56 foals in 1992 while an additional single centre study with the purpose of retrospectively determining an equation for future prognosis evaluation from a series of 99 foals appeared in 1997 and had a 68% overall survival rate (Hoffman et al. 1992; Furr et al. 1997). In 2005 Corley et al. reported 67% survival in a study of 67 critically ill foals, Castagnetti et al. (2010) reported 81% survival in a 2010 study of 81 critically ill foals and Wotman et al. (2009) reported 70% survival to discharge in a 2009 study of 225 foals. The largest outcome study to date was performed in the USA as a single site study (Axon et al. 1999) that focused exclusively on racing breeds and reported an 81% short-term survival rate in 289 studied foals. As the largest multicentre prospective report of outcomes for ill foals to date, this current study result is certainly a positive finding and should encourage those engaged in the practice of neonatal intensive care, both practitioners and owners, to pursue treatment when financially acceptable. It is also apparent that outcome depends on the clinician's diagnosis, with some carrying a better prognosis than others (Table 2).

Regarding the other direct intent of this study, this prospective multicentre study confirms the findings of several smaller equine studies (Corley et al. 2005; Johnston et al. 2007; Henderson et al. 2008; Wotman et al. 2009; Castagnetti et al. 2010; Tennent-Brown et al. 2010) and several human retrospective and prospective studies (Manikis et al. 1995; Bakker et al. 1996; Lavery et al. 2000; McNelis et al. 2001; Levraut et al. 2003; Nguyen et al. 2004; Shapiro et al. 2005; Howell et al. 2007; Rivers et al. 2007; Trzeciak et al. 2007; Jansen et al. 2009; Mikkelsen et al. 2009) that [LAC]ADMIT, is associated with outcome and is, therefore, not surprising. Overall, there was an increased risk of nonsurvival in this population of foals, where [LAC]ADMIT was measured, for each 1 mmol/l increase in [LAC]ADMIT of 14% (OR = 1.14; P<0.001). It is important that readers remain aware that the clinicians involved in this study were not blinded as to results of any [LAC] testing and there may be bias in the study associated with this lack of blinding, a sort of ‘self-fulfilling prophesy’ phenomenon. Uniquely and perhaps more importantly, this multicentre prospective study suggests that the utility of [LAC]ADMIT as a prognostic indicator may be different and varies by clinicians' major diagnosis category in the sick equine neonate. While Castagnetti et al. (2010) investigated the association of [LAC]ADMIT and clinical diagnosis in their single site study population, they used a slightly different classification system and did not investigate OR related to nonsurvival, [LAC]ADMIT and clinical diagnosis. In the current study the clinicians' major diagnoses of ‘sepsis’, ‘unspecified enterocolitis’, ‘unspecified colic’, ‘trauma’, ‘immune related (not FPTI)’ and ‘respiratory’ diagnoses each carried an increased risk of nonsurvival for each 1 mmol/l increase in [LAC]ADMIT, despite having very similar [LAC]ADMIT to other clinician's major diagnosis. In fact, the similarity of [LAC]ADMIT across the clinicians' major diagnoses was of interest, particularly when combined with the observation that the diagnostic categories were only, but clearly, finally distinguished related to [LAC]ADMIT by the OR analysis. This is in alignment with the clinical observation of the authors that sick foals consistently present with increased lactate concentrations. It is of interest, and somewhat unanticipated, that the clinicians' major diagnosis categories of ‘prematurity/dysmaturity’ and ‘PAS’ did not carry an increased risk of nonsurvival based on [LAC]ADMIT, although ‘PAS’ was close. On reflection, there is an age-dependent decrease in [LAC] over the first few days of life and we would speculate that premature foals are more obviously abnormal closer to parturition and may be referred while still in the very early stages of age-related decreases in [LAC]. In fact, although not specifically examined in this study, a post hoc examination of age at admission revealed that both ‘PAS’ and ‘prematurity/dysmaturity’ cases presented at around age 6 h, with only HRP/difficult delivery foals presenting sooner. It may also be that foals in these clinician-determined diagnostic categories suffered an acute insult or had adapted to a more chronic insult and had recoverable problems, negating the ability of [LAC]ADMIT to predict outcome.

The clinicians' major diagnoses that carried an increased risk of nonsurvival based on [LAC]ADMIT make sense as they are those we traditionally associate with problems related to the cardiovascular system (hypovolaemic shock associated with sepsis, enterocolitis or colic, trauma) or with oxygen delivery to tissues such as neonatal isoerythrolysis (incorporated into the immune related not FPTI category), trauma (due to blood loss) and respiratory compromise. In some clinicians' major diagnosis categories either all foals with that diagnosis survived (FPTI, meconium impaction, high risk pregnancy/difficult delivery), had [LAC]ADMIT above a specific concentration that perfectly predicted nonsurvival (unspecified congenital deformity) or there were insufficient numbers to justify statistical evaluation (muscle disease and umbilical not uroperitoneum). A word of caution is important relating to both PAS and high risk pregnancy/difficult delivery categories as there is instinctually a considerable overlap in these clinicians' major diagnoses. We would urge caution in over-interpretation of the finding in ‘HRP/difficult delivery’ of no increased risk of nonsurvival associated with increased [LAC] as this clinicians' major diagnosis category was probably used in cases where the foal was born following an adverse event but did not develop the expected complications of PAS. This is further clarified by the observation that one-third of known dystocia cases and almost half of cases where premature placental separation was observed had a clinicians' major diagnosis of PAS. Combined with common sense and clinical experience, these observations would lead us to the conclusion that many foals in the PAS category also resulted from adverse peripartum events that either were not observed or not considered the clinicians' major diagnosis by the attending clinician; rather, PAS was.

Blood lactate concentration found its role as a biomarker to accurately predict morbidity and mortality many years ago and the predictive power seems fairly nonspecific regardless of whether [LAC] was evaluated as a single measurement (Howell et al. 2007; Trzeciak et al. 2007; Mikkelsen et al. 2009; Castagnetti et al. 2010) or sequentially (Vincent et al. 1983; Nguyen et al. 2004; Wotman et al. 2009; Castagnetti et al. 2010; Tennent-Brown et al. 2010). This holds true for our prospective multicentre study, as [LAC]ADMIT was significantly higher in nonsurvivors on initial admission. These findings are in agreement with previous recent studies in mature and neonatal equine patients (Corley et al. 2005; Johnston et al. 2007; Henderson et al. 2008; Wotman et al. 2009; Castagnetti et al. 2010; Tennent-Brown et al. 2010).

Regarding the pathophysiology of increased [LAC], hyperlactaemia can be the result of the imbalance of tissue DO2 and metabolic demand, leading to tissue dysoxia, anaerobic glycolysis and increased lactate production (type A lactic acidosis). Type B lactic acidosis refers to increased [LAC] in the face of adequate DO2, perhaps due to increased tissue oxygen demand or decreased ability of the tissue to utilise oxygen, possibly resulting from decreased clearance of this metabolite. Type B lactic acidosis is thought to occur most commonly in systemic inflammatory response syndrome (SIRS), sepsis, renal or hepatic failure, neoplasia, diabetes mellitus, mitochondrial dysfunction or in association with administration of certain drugs (James et al. 1999; Fall and Szerlip 2005).

When evaluating both reported historical adverse peripartum events and clinicians' major diagnosis groups and their potential association with [LAC]ADMIT, foals with reported adverse intrapartum events such as dystocia and premature placental separation had significantly higher [LAC]ADMIT than foals with reported uneventful parturition. Foals from the PAS clinicians' major diagnosis group had significantly higher [LAC]ADMIT compared with the unspecified enterocolitis and meconium impaction only groups. The unspecified congenital deformity group had the largest apparent median and mean [LAC]ADMIT compared with all other clinicians' major diagnosis groups by inspection but, due to the small number of foals in that group, this difference was not significant. Because each foal was allowed only one final diagnosis determined by the attending clinician, many foals in the PAS and unspecified congenital deformity clinicians' major diagnosis groups may also have had a history of adverse peripartum event; perhaps this explains the large survival rate and small numbers present in the final diagnosis category of HRP/difficult delivery (caesarean-section; dystocia). This is a weakness throughout the study as no strict definition for any diagnostic category was provided to the participating institutions and some foals may have been placed in different clinicians' major diagnosis categories if treated by a different clinician.

Interestingly, Corley et al. (2005) observed that foals with PAS had high [LAC]ADMIT and speculated that foals with PAS may have suffered from severe cardiovascular disturbances. However, a proportion of PAS foals are also thought to be exposed to inefficient transplacental transfer of gases, nutrients and metabolites, in addition to being potentially exposed to ante partum intrauterine cytokinaemia resulting in ‘fetal inflammatory response syndrome’ (FIRS, the fetal response to chorioaminionities), confounding a theory that cardiovascular disturbances are singly responsible for hyperlactaemia in these cases (Gotsch et al. 2007; Gantert et al. 2010). In addition, tissue hypoxia/tissue injury may induce an inflammatory response on their own, resulting in SIRS. Both hypoxia ischaemia and inflammation are supported in our study as causes of hyperlactaemia in PAS by the documentation of large percentages of foals with PAS experiencing either dystocia or premature placenta separation in their pre-admission history. However, as it stands, our study is unable to clarify the pathogenesis of hyperlactaemia at admission in PAS cases and this question deserves further investigation.

We found that a positive blood culture on admission is associated with nonsurvival, yet [LAC]ADMIT is not significantly different between blood culture positive and negative foals. Wotman et al. (2009) also found no association between [LAC]ADMIT and bacteraemia in a study of 225 foals, as did Castagnetti et al. (2010) in their study of 88 foals. Henderson et al. (2008) found that blood culture positive foals actually had lower [LAC]ADMIT, but those results may be spurious due to the low number of positive blood cultures (n = 13) in that particular study. In contrast, Corley et al. (2005) found that blood culture positive foals had significantly increased [LAC]ADMIT (P = 0.049), although there were again few blood culture positive foals (26) in the entire study population of 72 cases. In both the Corley et al. (2005) and Castagnetti et al. (2010) studies the presence of SIRS, rather than bacteraemia, was more closely associated with [LAC]ADMIT. The lack of association of blood culture status with outcome has already been questioned in 133 neonatal foals with diarrhoea (Hollis et al. 2008), a primary presenting complaint in 115 of the foals of the current study, where 66 foals were bacteraemic at admission and no association of bacteraemia with survival was found. One recent study (Hackett et al. 2010) reported that normal foals repeatedly sampled over the first 72 h of life were commonly blood culture positive on at least one sample; all positive cultures were obtained under age 12 h and no foal was clinically ill at any point during or following the study. Although only a small number of foals (7) were studied, the results do call into question the common use of blood culture as a method of determination of ‘septic vs. not septic’ in equine neonates, as bacteraemia may be relatively common in the first few days of life in apparently clinically normal foals. Although we did not evaluate specifically for the presence of SIRS in the current study, once a strict definition of SIRS is developed in equine neonates and uniformly accepted, it may be a useful inclusion criterion for sepsis, severe sepsis and septic shock in our species of interest and have more utility than blood culture in this regard. In fact, our results may be compromised by the use of blood culture status as a marker for sepsis.

In human medicine, Jansen et al. (2009) and Mikkelsen et al. (2009) found that the prognostic value of hyperlactaemia was similar in haemodynamically stable and unstable adult patients. In human emergency departments, [LAC]ADMIT has been shown to be a better predictor of mortality in adults than physiological triage criteria (heart rate, respiration rate, blood pressure, Glasgow coma scale) (Lavery et al. 2000). In our study of equine neonates we found a small but significant association with [LAC]ADMIT and MAP, comparable with results of Corley et al. (2005), who found an association between [LAC]ADMIT and MAP in a single centre selected population of equine neonates. Castagnetti et al. (2010) reported a correlation coefficient between [LAC]ADMIT and MAP of 0.44 in their single site report of 88 selected foals. Haemodynamic instability was defined as a MAP of <60 mmHg or the requirement of catecholamine administration to maintain adequate blood pressure in the Jansen et al. (2009) study. Interestingly, the study by Wotman et al. (2009) also evaluated the association between mean blood pressure and [LAC]ADMIT and found no association between the 2 parameters until they removed from analysis all foals with low pressure (<60 mmHg) that also had [LAC] <4.5 mmol/l, a commonly used cut point for [LAC]ADMIT as a predictor of survival. While a MAP of <60 mmHg was considered as hypotension in our study, in addition to those of Corley et al. (2005), Wotman et al. (2009) and Castagnetti et al. (2010), the need for inopressor support was not considered, possibly contributing to our different results. Probably most importantly, our population consisted of only neonates, including premature neonates known to have lower blood pressures, while the human study included only adult patients, probably explaining differences between man and foals in findings regarding pressure. Further study of these findings is necessary.

One weakness of this study might be the utilisation of either arterial or venous blood for [LAC]ADMIT measurement from either central or peripheral sampling sites. Several studies in human medicine have shown only small, insignificant differences in [LAC] in either venous or arterial blood from either peripheral or central sampling sites (Lavery et al. 2000; Bakker and Jansen 2007). The overall cut point of 4.4 mmol/l at admission to predict nonsurvival determined in this study was lower than most of those described in previous studies by Wotman et al. (2009) (5.5 mmol/l; sensitivity 0.79, specificity 0.70 correctly classifying 79% of the cases), Henderson et al. (2008) (6.9 mmol/l; correctly classifying 89% of cases) and Castagnetti et al. (2010) (5.0 mmol/l; sensitivity 0.76, specificity 0.63 with 7.94 mmol/l applied to provide a larger OR of nonsurvival and maximise sensitivity and specificity; sensitivity 0.76 and specificity 0.83) but not of that by Corley et al. (2005) (2.5 mmol/l; sensitivity 0.95, specificity 0.37). These differences may be due to choices to maximise both sensitivity and sensitivity by some investigators or to variation in age at presentation and disease severity/prevalence within different referral populations. Clearly, based on our results, the variability in OR related to diagnostic category plays a role in the imperfect predictive ability of [LAC]ADMIT cut points when applied to a general population of sick foals and should caution the clinician that the utility of [LAC]ADMIT as a prognostic indicator may depend on the clinicians' major diagnosis. In conclusion, [LAC] proves its utility as a valuable biomarker for morbidity and mortality in equine neonates as [LAC]ADMIT was significantly increased in nonsurvivors compared with survivors.

There are clear differences in the interpretation of [LAC]ADMIT based on the clinician determination of major clinically important diagnoses that should be considered when using [LAC]ADMIT as a prognostic indicator.

Authors' declaration of interests

No conflicts of interest have been declared.

Source of funding

None.

Acknowledgements

The authors would like to acknowledge Colin Schwarzwald, University of Zurich, for his critical review of this manuscript and mentorship at his institution. Portions of these data were presented in abstract at the 2010 American College of Veterinary Internal Medicine Forum and the 2010 International Veterinary Emergency and Critical Care Symposium.

Manufacturers' addresses

a Nova Biomedical, Waltham, Massachusetts, USA.

b Roche Diagnostics, Mannheim, Germany.

c Roche Diagnostic Corps, Indianapolis, Indiana, USA.

d SensLab GmbH, Leipzig, Germany.

e Radiometer, Copenhagen, Denmark.

f Abbott Laboratories, Inc., North Chicago, Illinois, USA.

g Stata Corp, College Station, Texas, USA.

h Minitab, State College, Pennsylvania, USA.

i GraphPad, La Jolla, California, USA.

Ancillary