The study was approved by the Ethical Review Committee, Faculty of Medicine, University of Colombo and was registered with the Sri Lanka Clinical Trials Registry, SLCTR/2007/005. All patients gave written and informed consent to the study. For patients between 14–18 years, written informed consent was obtained from the patient's parents/guardians as well as from the patients themselves.

Patients were recruited from a secondary referral hospital in Chilaw, Sri Lanka between 21^st January 2007 and 31^st July 2009. Patients ≥14 years requiring treatment with snake antivenom were eligible for inclusion. Patients administered antivenom prior to recruitment to the study were excluded. After commencing antivenom (10 vials of polyvalent antivenom (Bharat Serums and Vaccines Limited, India) in 500 ml normal saline), patients were observed for four hours for evidence of a systemic hypersensitivity reaction. Blood was collected into EDTA tubes prior to antivenom administration and then 15 minutes, 2 hours, 6 hours, 12 hours and 24 hours after antivenom administration. Samples were immediately centrifuged and the plasma aliquoted and frozen.

Demographic information, details of the bite, information on snake identification, clinical features of envenoming, treatments and outcomes were recorded. Baseline observations were recorded prior to antivenom administration. The clinical features of early reactions to antivenom were recorded along with blood pressure, heart rate, respiratory rate and pulse oximetry readings.

Clinical envenoming syndrome(s) were defined in each patient as venom induced consumption coagulopathy (INR>1.5), neurotoxicity (ptosis, diplopia, bulbar paralysis, limb weakness and/or respiratory muscle paralysis) or non-specific systemic symptoms (nausea, vomiting, headache, abdominal pain, diarrhoea). The type of snake was based on identification of the snake by the patient or clinician, and confirmed by enzyme immunoassay (EIA) for Russell's viper cases [26].

Adverse reactions to antivenom were classified as hypersensitivity reactions as we observed “objectively reproducible symptoms or signs, initiated by exposure to a defined stimulus at a dose tolerated by normal subjects” [27]. The use of the term hypersensitivity does not imply previous exposure to antivenom. Systemic hypersensitivity reactions were defined by typical skin manifestations (urticaria/erythema) or National Institute of Allergy and Infectious Diseases/Food Allergy and Anaphylaxis Network (NIAID-FAAN) diagnostic criteria for anaphylaxis [28], which recognise that some reactions may occur in the absence of typical skin features. Our adaptation of this system is presented in Table 1. Because gastrointestinal symptoms were a common effect of envenoming, only new and persistent gastrointestinal symptoms were included in the NIAID-FAAN Definition 2 for the purposes of this study. We also added a further classification of respiratory reactions (Definition 4) to account for the severe respiratory reactions sometimes noted to occur without typical skin features. These reactions were distinguished from those arising from respiratory paralysis because they were not preceded by a descending paralysis and result in primary hypoxia rather than hypoxia secondary to ventilator failure. Reactions were graded as skin only (which does not correspond with a diagnosis of anaphylaxis), moderate anaphylaxis or severe anaphylaxis if hypotension or hypoxia was observed during the course of a hypersensitivity reaction [29]. Pyrogenic reactions (rigors and/or fever) were coded separately and not included in the hypersensitivity reaction grading system.

Treatment of anaphylactic reactions was directed by the attending physician and most commonly involved a combination of parenteral adrenaline (0.25 to 1 mg), corticosteroids (hydrocortisone) and H1-antagonists.

Plasma concentrations of MCT, histamine, IL-6, IL-10, sTNFRI and TNFα, were measured as described previously [25]. MCT concentrations in plasma were analyzed using the Phadia ImmunoCAP system. MCT results greater than 11.4 ng/ml are considered positive by this system, however we also assessed change in MCT (ΔMCT) with a positive MCT result regarded as a deviation of 2.0 µg/L (ng/ml) between high and low values for each case as measured over time [30]. Histamine concentrations were determined using IBL Histamine ELISA Kits according to the manufacturer's instructions (IBL, Hamburg, Germany). Plasma samples were analyzed for IL-6, IL-10, sTNFRI and TNFα using Flex Sets (BD Biosciences, San Jose, USA). Plasma levels of C3a, C4a and C5a were determined using the Anaphylatoxin CBA Kit from BD Biosciences. We defined a positive result (>upper limit) as a concentration greater than the 99^th centile of healthy control samples (n = 34) in our laboratory.

Data are presented as median (interquartile range; IQR). Because of the skewed distributions of mediator concentrations, non-parametric tests were applied to non-transformed (raw) data. The measured baseline and peak mediator concentrations were compared across all groups using the Kruskal-Wallis test and between two groups using the Wilcoxon rank-sum (Mann-Whitney) test. Matched data pairs (i.e. baseline versus peak) were compared using the Wilcoxon signed-rank test. Correlations between baseline and peak mediator concentrations were assessed with Spearman rank correlation and Bonferroni adjustments were applied for multiple simultaneous testing. The semi-parametric lowess smoothing procedure was used to illustrate the relationship between time from antivenom and the concentration of each mediator. Statistical analysis was performed with Stata version 10.1 (StataCorp, College Station, Texas).

Of 1004 patients presenting with suspected snakebites during the study period, 247 patients received antivenom. One hundred and ninety eight were randomised to receive antivenom by their allocated infusion rate, and of these, 120 patients had blood samples collected for this study. Baseline characteristics for these patients are presented in Table 2. INR results were available for 112/120 patients, with 111/112 patients presenting with coagulopathy (INR>1.5). The median value for the maximum INR measured was 7.0 (IQR 3.7–13). Signs of neurotoxicity were present in 72/120 (60%) patients and non-specific systemic symptoms were observed in 91/120 (76%) patients. Eighty-six patients bitten by Russell's viper had pre-antivenom venom concentrations measured, with a median of 171(IQR 85–399) ng/ml. Thirty-three patients received a second dose of antivenom a median of 8 (IQR 7–13) hours after the first dose, and six of these patients received a third dose of antivenom.

Hypersensitivity reactions after the initial dose of antivenom occurred in 77/120 (64%) patients, including 20/120 (17%) with skin-only reactions and 57/120 (48%) with anaphylaxis. All reactions occurred within the first four hours of starting antivenom treatment. Cases of anaphylaxis satisfied Definition 1 in 32/57 (56%), Definition 2 in 11/57 (19%), Definition 3 in 6/57 (11%) and Definition 4 in 8/57 (14%). Severe anaphylaxis was observed in 51/57 (89%) anaphylaxis cases, of which 21/51 (41%) were hypotensive, 18/51 (35%) were hypoxemic and 12/51 (24%) were both hypotensive and hypoxaemic. As reported previously, there was no difference in hypersensitivity reaction rates between the two study arms (i.e. 20 minute vs 2 hour antivenom infusion rate) [5]. Pyrogenic reactions were observed in 32/120 patients (27%), 10/32 (31%) reactions were pyrogenic alone and 22/32 (69%) had a pyrogenic reaction in addition to a hypersensitivity reaction (1 skin-only, 21 anaphylaxis). Pyrogenic reactions were seen more often in patients with anaphylaxis (21/57, 41%) compared to those without (11/63, 17%).

A comparison between pre-antivenom mediator concentrations in envenomed patients and healthy controls is shown in Table 3. Almost all patients with snake envenoming had evidence of complement activation before receiving antivenom, with plasma concentrations of C3a elevated above the 99^th centile of normal in 91% of patients. Plasma concentrations of IL-6, IL-10, C4a and C5a were also significantly higher in envenomed patients before they received antivenom. Although plasma concentrations of MCT were greater than the 99^th centile of our healthy control group in 31% of envenomed patients, only 13/120 (11%) patients had MCT concentrations above 11.4 ng/ml, the upper limit of normal defined by the assay (median 15 (IQR 13.9–28.6). Pre-antivenom concentrations of histamine and sTNFRI were not significantly elevated in envenomed patients compared to healthy controls. The median and interquartile ranges for TNFα were zero for both groups.

Pre-antivenom mediator concentrations did not correlate with maximum INR or pre-antivenom serum venom concentrations (p>0.5 for all comparisons, Spearman's). Patients with signs of neurotoxicity had similar pre-antivenom plasma concentrations of all mediators when compared to patients without signs of neurotoxicity (p>0.2 for all comparisons, Mann Whitney). Patients with non-specific systemic symptoms of envenoming had significantly higher pre-antivenom levels of IL-6, IL-10 and sTNFRI compared to those without (median serum concentrations of IL-6 [12.6 (IQR 0–31.9) versus 6.2 (IQR 0–19.5), p = 0.044], IL-10 [29.6 (IQR 10.7–68.7) versus 7.6 (IQR 0–16.5), p<0.001] and sTNFRI [2043 (IQR 1131–3257) versus 962 (IQR 603–1546), p<0.001]). There was no significant difference in pre-antivenom mediator levels between patients subsequently having no reaction to antivenom and patients experiencing pyrogenic reactions and/or anaphylaxis.

Changes in serum concentrations of all immune mediators measured before and after treatment with antivenom are illustrated in Figure 1 (MCT and histamine), Figure 2 (IL-6, IL-10, TNFα and sTNFRI) and Figure 3 (C3a, C4a and C5a). As all adverse reactions to the initial dose of antivenom occurred during the first four hours of antivenom treatment, and second and third doses of antivenom were administered at least four hours after the initial dose for all patients, only patients with samples available from the first four hours of antivenom treatment were used for comparisons of baseline (pre-antivenom) and peak mediator concentrations (post-antivenom) by Wilcoxon Signed Rank Test (n = 98). There were significant increases (baseline versus peak) in MCT in all patient groups, whereas histamine only increased significantly during skin-only and anaphylactic reactions (Figure 1). Following antivenom administration, peak levels of MCT and histamine were significantly higher in anaphylaxis cases than patients with no reaction and patients with skin-only reactions (Table 4). ΔMCT was positive in 77% of anaphylaxis cases, 56% of skin-only reactions and 31% of patients with no reaction. The proportion of anaphylaxis patients with a positive ΔMCT was significantly higher than patients with no reaction (p<0.001, Fisher's Exact Test). In patients who had anaphylaxis, concentrations of histamine returned to baseline levels between 5–10 hours after the initial dose of antivenom whereas concentrations of MCT remained elevated in some patients for up to 24 hours (Figure 1).

Plasma concentrations of IL-6, IL-10, TNFα and sTNFRI all increased significantly after antivenom compared with baseline in all patient groups with the exception of TNFα and sTNFRI in skin-only reactions (Figure 2). However, peak mediator concentrations were not different between patients with no reaction, skin-only reactions and anaphylaxis (Table 4). In patients who had anaphylaxis, concentrations of IL-6 and IL-10 remained elevated for up to 24 hours after the initial dose of antivenom, whereas concentrations of TNFα and sTNFRI returned to baseline levels between 10–15 hours after the initial dose of antivenom (Figure 2).

Figure 3 demonstrates that plasma concentrations of C3a, C4a and C5a were significantly elevated above healthy control levels prior to antivenom, but did not increase significantly in any patient group after administration of antivenom. There were no significant differences in peak concentrations of C3a, C4a and C5a between groups (Table 4).

In agreement with our finding that there was no difference in hypersensitivity reaction rates between the two study arms (i.e. 20 minute vs 2 hour antivenom infusion rate), there was also no significant difference in serum concentrations for any immune mediator measured between the different infusion rates.

When all patients were considered together, patients who had pyrogenic reactions to the antivenom had significantly higher peak concentrations of IL-6, IL-10, TNFα and sTNFRI than patients without pyrogenic reactions (data not shown). A confounding factor was that patients with pyrogenic reactions were also more likely to have anaphylaxis. Patients with pyrogenic reactions alone had significantly higher peak levels of IL-6, IL-10 and sTNFRI compared to patients with no reaction (pyrogenic or hypersensitivity) (Table S1). Similarly, patients with both pyrogenic reactions and anaphylaxis had significantly higher peak concentrations of IL-6, IL-10, TNFα and sTNFRI compared to patients with severe anaphylaxis alone.