Pharmacokinetics of Snake Venom

Understanding snake venom pharmacokinetics is essential for developing risk assessment strategies and determining the optimal dose and timing of antivenom required to bind all venom in snakebite patients. This review aims to explore the current knowledge of snake venom pharmacokinetics in animals and humans. Literature searches were conducted using EMBASE (1974–present) and Medline (1946–present). For animals, 12 out of 520 initially identified studies met the inclusion criteria. In general, the disposition of snake venom was described by a two-compartment model consisting of a rapid distribution phase and a slow elimination phase, with half-lives of 5 to 48 min and 0.8 to 28 h, respectively, following rapid intravenous injection of the venoms or toxins. When the venoms or toxins were administered intramuscularly or subcutaneously, an initial absorption phase and slow elimination phase were observed. The bioavailability of venoms or toxins ranged from 4 to 81.5% following intramuscular administration and 60% following subcutaneous administration. The volume of distribution and the clearance varied between snake species. For humans, 24 out of 666 initially identified publications contained sufficient information and timed venom concentrations in the absence of antivenom therapy for data extraction. The data were extracted and modelled in NONMEM. A one-compartment model provided the best fit, with an elimination half-life of 9.71 ± 1.29 h. It is intended that the quantitative information provided in this review will provide a useful basis for future studies that address the pharmacokinetics of snakebite in humans.


Introduction
In 2009, WHO listed snakebite as a neglected tropical disease, recognising its importance alongside many infectious diseases [1]. This is a particularly important public health issue in tropical and subtropical regions [2,3] mostly affecting those who live in rural areas, including the agricultural workforce in developing countries [3][4][5]. Snake envenomation causes significant morbidity and mortality usually requiring hospitalisation and occasionally causing permanent disabilities and in severe cases death [6]. Delayed access to appropriate medical facilities, lack of antivenom, and limited supportive treatments are considered the main contributors to the high morbidity and mortality [3,4].
Venomous snakes with medical importance are predominantly front-fanged, originating from three families: Atractaspididae, Elapidae, and Viperidae [7]. Snakes from the Viperidae family can be further divided into two subfamilies, Viperinae (true vipers) and Crotalinae (pit vipers). In all families, the venom Table 1. List of common venom components (enzymatic) found in snakes of the Elapidae and Viperidae families, their size and activities.

Enzymatic Components Approximate Molecular Mass (kDa) Mechanism of Action Examples of Biological Effects Snake Families
Phospholipase A 2 (PLA 2 ) 12-14 [37] Hydrolyses the ester bond at sn-2 position of phospholipids producing free fatty acids and lysophospholipid. Toxic effects can result from this enzymatic action or may be the results of non-enzymatic activity [38].
Toxins 2018, 10, 73 4 of 21 Table 2. List of common venom components (non-enzymatic) found in snakes of the Elapidae and Viperidae families, their size and activities.

of 21
The treatment of snake envenomation principally involves the administration of specific antivenoms for the snake species or type involved, and supportive care. Snake antivenoms are antibodies or antibody fragments which are derived from the plasma of animals (typically horses and sheep) that have been immunised with a snake venom [63]. The intravenously injected antibodies (IgG immunoglobulin) or antibody fragments [F(ab') 2 or Fab] bind and neutralise free venom in the patient's plasma, which aims to reverse or prevent further toxic effects. However, animal-derived antivenoms also carry a risk of hypersensitivity reactions because they are foreign proteins, which can result in cutaneous and multiorgan reactions that are potentially life-threatening with severe anaphylaxis [5,63]. Antivenoms remain expensive, and their supply is often limited in some regions, so more effective dosing and antivenom use may improve their cost-effectiveness.
The dose and timing of antivenom administration is still largely empirical and often based on in vitro neutralisation studies in animals, in which the toxic effects differ compared to humans [64]. In practice, the dose administered is determined by the treating clinicians on the basis of subjective symptoms or clinical or laboratory results. New approaches are required to improve antivenom dosing and determine the optimal dose and timing of antivenom administration. The knowledge of snake venom pharmacokinetics (the study of the time course of venom distribution in biological systems) will provide important information about the time course of the exposure to venom. This will allow for a better determination of a sufficient dose and timing of antivenom administration. In particular, it remains unclear how long post-bite the administration of antivenom remains an effective therapeutic intervention.
This paper aims to bring together the current knowledge of the pharmacokinetics of snake venom by reviewing: (1) laboratory studies performed in animals to investigate the time course of venom concentrations in plasma; and (2) reports describing the time course of venom concentrations in humans.

Literature Search
The initial search generated 493 studies for screening, after duplicates were removed. Of these, 481 studies were excluded on the basis of the inclusion and exclusion criteria (see Figure 1). This left 12 studies available for data extraction and to investigate the characteristics and pharmacokinetic parameters of snake venom in animals.

Demographics of Snake Venom Pharmacokinetic Studies in Animals
Of the 12 studies included, eight studies investigated the pharmacokinetics of whole venom, three investigated the pharmacokinetics of isolated snake toxins, and one looked at both venom and purified toxins. Eight of the 12 studies were performed in rabbits. The pharmacokinetic parameters reported in these studies were estimated from the plasma concentration time course following the administration of snake venoms or toxins via the intravenous (IV), intramuscular (IM), or subcutaneous (SC) routes. A summary of the pharmacokinetic parameters reported in these 12 studies are shown in Tables 3-5.

Pharmacokinetic Parameters of Snake Venoms and Toxins Following Intravenous Administration
In most studies, the investigators identified a two-compartment output model as the best fit for their data. Overall, regardless of the type of snake, the concentration initially decreased rapidly with a half-life ranging from, for whole venoms, 15 (Cryptelytrops purpureomaculatus) to 48 min (Naja sumatrana and Hypnale hypnale), for toxins, 5 (habutobin of Trimeresurus flavoviridis) to 42 min (PLA2 of N. sumatrana). In the terminal phase, the concentrations decreased more slowly with longer halflives ranging from, for whole venoms, 12 (Vipera aspis and Bothrops alternatus) to 28 h (C. purpureomaculatus), for toxins, 0.8 (habutobin of T. flavoviridis) to 12 h (PLA2 of N. sumatrana). It is assumed that the initial phase is the distribution and the terminal phase is the elimination, but this cannot be determined from the data presented. Habutobin, a toxin purified from the venom of T. flavoviridis, had the shortest distribution and elimination half-lives compared to the rest of the snake venoms and toxins. Also of note, N. naja atra cytotoxin had a shorter half-life than the venom and toxins from two other types of cobra studied (N. sumatrana and N. sputatrix).  [67] reported a three-compartment model fit for their data, with three half-lives (24 min, 48 min, and 19 h, respectively). The half-lives were similar to the two-compartment models.
The reported volume of distribution at steady state (Vss) for snake venoms and toxins ranged widely. The Vss of snake venoms ranged from as low as 0.12 L·kg −1 in M. fulvius venom to 1.2 L·kg −1 in V. aspis venom. The Vss was notably higher for the thrombin-like enzyme (TLE) from A. halys ussuriensis Emelianov venom (1.8 L·kg −1 ).
The systemic clearance (CL) of snake venoms and toxins reported also varied from species to species. The CL of snake venom ranged from 0.007 L·h −1 ·kg −1 for H. hypnale venom to 0.093 L·h −1 ·kg −1 for M. fulvius venom, and from 0.048 to as high as 0.324 L·h −1 ·kg −1 for snake toxins (PLA2 in N. sumatrana and TLE from A. halys ussuriensis Emelianov venom, respectively). It appears that the reported values of CL of snake toxins are generally larger than those reported for whole snake venoms.

Pharmacokinetic Parameters of Snake Venoms and Toxins Following Intravenous Administration
In most studies, the investigators identified a two-compartment output model as the best fit for their data. Overall, regardless of the type of snake, the concentration initially decreased rapidly with a half-life ranging from, for whole venoms, 15 (Cryptelytrops purpureomaculatus) to 48 min (Naja sumatrana and Hypnale hypnale), for toxins, 5 (habutobin of Trimeresurus flavoviridis) to 42 min (PLA 2 of N. sumatrana). In the terminal phase, the concentrations decreased more slowly with longer half-lives ranging from, for whole venoms, 12 (Vipera aspis and Bothrops alternatus) to 28 h (C. purpureomaculatus), for toxins, 0.8 (habutobin of T. flavoviridis) to 12 h (PLA 2 of N. sumatrana). It is assumed that the initial phase is the distribution and the terminal phase is the elimination, but this cannot be determined from the data presented. Habutobin, a toxin purified from the venom of T. flavoviridis, had the shortest distribution and elimination half-lives compared to the rest of the snake venoms and toxins. Also of note, N. naja atra cytotoxin had a shorter half-life than the venom and toxins from two other types of cobra studied (N. sumatrana and N. sputatrix).  [67] reported a three-compartment model fit for their data, with three half-lives (24 min, 48 min, and 19 h, respectively). The half-lives were similar to the two-compartment models.
The reported volume of distribution at steady state (V ss ) for snake venoms and toxins ranged widely. The V ss of snake venoms ranged from as low as 0.12 L·kg −1 in M. fulvius venom to 1.2 L·kg −1 in V. aspis venom. The V ss was notably higher for the thrombin-like enzyme (TLE) from A. halys ussuriensis Emelianov venom (1.8 L·kg −1 ).
The systemic clearance (CL) of snake venoms and toxins reported also varied from species to species. The CL of snake venom ranged from 0.007 L·h −1 ·kg −1 for H. hypnale venom to 0.093 L·h −1 ·kg −1 for M. fulvius venom, and from 0.048 to as high as 0.324 L·h −1 ·kg −1 for snake toxins (PLA 2 in N. sumatrana and TLE from A. halys ussuriensis Emelianov venom, respectively). It appears that the reported values of CL of snake toxins are generally larger than those reported for whole snake venoms. Table 3. Pharmacokinetic parameters of snake venom and toxins following intravenous (IV) injection expressed as mean (±SD).

Snake Species Animal Model
No.

Agkistrodon halys ussuriensis Emelianov
Sprague  * SD calculated from standard errors reported in the original papers using the following equation: SD = SE × √ n ; ** Assumed to be reported as SD; *** SD not reported; a = Apparent volume of distribution; b = Unit by weight (·kg −1 ) calculated on the basis of the mean of the reported animal weight; c = Apparent volume of central compartment by weight (V c ); d = Apparent volume of peripheral compartment by weight (V p ); Abbreviations: t 1/2α , half-life of the distribution phase; t 1/2β , half-life of the elimination phase; V ss , apparent volume of distribution at steady state; CL, systemic clearance. Table 4. Pharmacokinetic parameters of snake venom and toxins following intramuscular (IM) injection expressed as mean (±SD).

Snake Species Animal Model
No. Dose (mcg·kg −1 ) * SD calculated from standard errors reported in the original papers using the following equation: SD = SE × √ n ; ** Assumed to be reported as SD; a = Apparent volume of distribution; b = Unit by weight (·kg −1 ) calculated on the basis of the mean of the reported animal weight; Abbreviations: t 1/2ka , half-life of the absorption phase; F, bioavailability; t 1/2α , half-life of the distribution phase; t 1/2β , half-life of the elimination phase; V ss , apparent volume of distribution at steady state; CL, systemic clearance. Table 5. Pharmacokinetic parameters of snake venom and toxins following subcutaneous (SC) injection expressed as mean (±SD).

Snake Species Animal Model
No. . In these studies, the distribution half-life, which is expected to be related to the ongoing absorption that occurs during the initial distribution phase, was not commonly seen. The studies by Zhao et al., 2001 [68] and Guo et al., 1993 [27] are the only two studies that report the half-life of the absorption phase as well as two phases of disposition half-lives. In these two studies, the half-lives following SC administration were reported to be 2.5, 4.8, and 125 h, respectively [68], and those following IM administration were 0.077, 0.37, and 5.9 h, respectively [27]. Guo et al., 1993 [27], Audebert et al., 1994 [35], and Yap et al., 2013 [73] found that the terminal half-life following IM injection was longer than that following IV administration. However, in the studies by Sim [66] found the terminal half-life to be longer following SC administration than after IV injection. In these cases, it is expected that the prolonged duration of the half-life would be a function of prolonged absorption since the disposition processes should be unaffected by the input processes.

Literature Search
The initial search generated 576 studies for screening, after duplicates were removed. A total of 552 studies were further removed on the basis of the inclusion and exclusion criteria (Figure 2), leaving 26 studies for data extraction and ultimately post-hoc (model-based) analysis. . In these studies, the distribution half-life, which is expected to be related to the ongoing absorption that occurs during the initial distribution phase, was not commonly seen. The studies by Zhao et al., 2001 [68] and Guo et al., 1993 [27] are the only two studies that report the half-life of the absorption phase as well as two phases of disposition half-lives. In these two studies, the half-lives following SC administration were reported to be 2.5, 4.8, and 125 h, respectively [68], and those following IM administration were 0.077, 0.37, and 5.9 h, respectively [27]. Guo et al., 1993 [27], Audebert et al., 1994 [35], and Yap et al., 2013 [73] found that the terminal half-life following IM injection was longer than that following IV administration. However, in the studies by Sim [66] found the terminal half-life to be longer following SC administration than after IV injection. In these cases, it is expected that the prolonged duration of the half-life would be a function of prolonged absorption since the disposition processes should be unaffected by the input processes.

Literature Search
The initial search generated 576 studies for screening, after duplicates were removed. A total of 552 studies were further removed on the basis of the inclusion and exclusion criteria (Figure 2), leaving 26 studies for data extraction and ultimately post-hoc (model-based) analysis.

Data Extraction
Timed venom concentration data were extracted from 24 (Table 6) of the 26 studies included from the literature review. One study was not included for further analysis, as the serum antigen concentrations appeared to have increased over time post-bite for all patients over a period of 24 h, which was not consistent with data reported in the other studies. Another study was excluded from data extraction because of the difficulty in distinguishing individual data points from the overlapping concentration-time profiles of the 37 patients, and, hence, we were unable to digitise the data from the figure. Nine studies reported elapid envenomation, and sixteen studies reported viper envenomation. From the 24 included studies, we were able to retrieve data for 145 individuals. A total of 218 timed concentration data were obtained from the text of the case reports. For studies in which the results were reported graphically, we successfully recreated similar plots by digitising the data points from the figures using the WebPlotDigitizer software.

Data Analysis
A plot of the extracted data is shown in Figures 3 and 4. Most data were collected within the first few hours since snakebite, and most individuals provided only a single timed venom concentration. This is because most studies primarily reported a clinical presentation of snakebite, and only one sample was taken prior to the administration of the antivenom. Only five studies reported serial venom concentrations over time in the absence of antivenom. The dataset obtained was sparse in terms of observations per individual, and there were large magnitudes of difference between the snake venom concentrations. In some cases, snake venom concentrations could still be detected 24 h after the bite. Figure 4 compares the concentration-time profiles of patients bitten by snakes from the Elapidae family ( Figure 4a) and the Viperidae family (Figure 4b). It can be seen that the venom concentrations detected in patients bitten by vipers are typically higher than those in patients bitten by elapids.
The extracted data were modelled using NONMEM. The concentration-time data were best described by a one-compartmental model with zero-order input and first-order elimination. A prediction-corrected visual predictive check is presented in Figure 5. The observed percentiles are mostly within the 95% confidence interval of simulated percentiles, suggesting that the model provides a good description of the data. The half-life calculated from the estimated values of clearance and volume of distribution was 9.71 ± 1.29 h (see Table A1 of Appendix A for parameter estimates from the model). The model also accounted for the relative exposure (i.e., "dose") of patients, resulting from bites of snakes from the Elapidae and Viperidae family. It was apparent from these data that vipers inject approximately 75% more venom than elapids, but it should be noted the between-subject variability in the exposure was large, corresponding to 275%, meaning that the variability between bites within a family is at least as significant as the differences between families of snakes. It is possible that some of this variability could be further explained by the snake species, but these data were not available. concentrations detected in patients bitten by vipers are typically higher than those in patients bitten by elapids. The extracted data were modelled using NONMEM. The concentration-time data were best described by a one-compartmental model with zero-order input and first-order elimination. A prediction-corrected visual predictive check is presented in Figure 5. The observed percentiles are mostly within the 95% confidence interval of simulated percentiles, suggesting that the model provides a good description of the data. The half-life calculated from the estimated values of clearance and volume of distribution was 9.71 ± 1.29 h (see Table A1 of Appendix A for parameter estimates from the model). The model also accounted for the relative exposure (i.e., "dose") of patients, resulting from bites of snakes from the Elapidae and Viperidae family. It was apparent from these data that vipers inject approximately 75% more venom than elapids, but it should be noted the between-subject variability in the exposure was large, corresponding to 275%, meaning that the variability between bites within a family is at least as significant as the differences between families of snakes. It is possible that some of this variability could be further explained by the snake species, but these data were not available. concentrations detected in patients bitten by vipers are typically higher than those in patients bitten by elapids. The extracted data were modelled using NONMEM. The concentration-time data were best described by a one-compartmental model with zero-order input and first-order elimination. A prediction-corrected visual predictive check is presented in Figure 5. The observed percentiles are mostly within the 95% confidence interval of simulated percentiles, suggesting that the model provides a good description of the data. The half-life calculated from the estimated values of clearance and volume of distribution was 9.71 ± 1.29 h (see Table A1 of Appendix A for parameter estimates from the model). The model also accounted for the relative exposure (i.e., "dose") of patients, resulting from bites of snakes from the Elapidae and Viperidae family. It was apparent from these data that vipers inject approximately 75% more venom than elapids, but it should be noted the between-subject variability in the exposure was large, corresponding to 275%, meaning that the variability between bites within a family is at least as significant as the differences between families of snakes. It is possible that some of this variability could be further explained by the snake species, but these data were not available.

Pharmacokinetic Studies in Animals
To date, there is only a handful of experimental studies on the pharmacokinetics of snake venom in the absence of antivenom. Most of these studies were performed in large animals, commonly rabbits, as they allowed for the analysis of multiple samples over a longer period of time, consistent with the duration of envenomation. Since a snake venom is comprised of a cocktail of protein components, it is common that snake venoms are investigated as whole venom rather than as individual toxins. Some investigators took the additional step of isolating the major toxins of interest in order to study their profiles [68][69][70]. For the purpose of this review, we only assessed the pharmacokinetic parameters that arose from studies carried out using ELISA, although there were some other studies performed using RIA methods [94][95][96].

Pharmacokinetic Studies in Animals
To date, there is only a handful of experimental studies on the pharmacokinetics of snake venom in the absence of antivenom. Most of these studies were performed in large animals, commonly rabbits, as they allowed for the analysis of multiple samples over a longer period of time, consistent with the duration of envenomation. Since a snake venom is comprised of a cocktail of protein components, it is common that snake venoms are investigated as whole venom rather than as individual toxins. Some investigators took the additional step of isolating the major toxins of interest in order to study their profiles [68][69][70]. For the purpose of this review, we only assessed the pharmacokinetic parameters that arose from studies carried out using ELISA, although there were some other studies performed using RIA methods [94][95][96].

Pharmacokinetic Studies in Animals
To date, there is only a handful of experimental studies on the pharmacokinetics of snake venom in the absence of antivenom. Most of these studies were performed in large animals, commonly rabbits, as they allowed for the analysis of multiple samples over a longer period of time, consistent with the duration of envenomation. Since a snake venom is comprised of a cocktail of protein components, it is common that snake venoms are investigated as whole venom rather than as individual toxins. Some investigators took the additional step of isolating the major toxins of interest in order to study their profiles [68][69][70]. For the purpose of this review, we only assessed the pharmacokinetic parameters that arose from studies carried out using ELISA, although there were some other studies performed using RIA methods [94][95][96].
The pharmacokinetic profiles of snake venoms and isolated toxins were investigated in animals following intravenous injection. This avoids issues related to bioavailability and provides profiles that are dependent only on distribution and clearance. Most studies demonstrated that the concentration-time profiles of snake venoms and toxins from both snake families were best described by a two-compartment model. The concentrations reportedly declined rapidly during the initial phase. This suggests that the venoms (or toxins) were rapidly distributed out of the central compartment to peripheral compartments. The initial half-life values reported for the whole venoms and the isolated toxins ranged from 5 min to 48 min, which is significantly faster than the terminal half-lives which could be as long as 28 h. Given the low CL values and the relatively high V ss values of the venoms, we suspect that the terminal phase half-life from the two-compartment model, may be predominantly contributed to by elimination. This does not rule out further distribution components. The volume of distribution of snake venom in animals was relatively large when compared to their blood volume (approximately 0.054-0.070 L·kg −1 for rats, 0.057-0.065 L·kg −1 for rabbits, and 0.058-0.064 L·kg −1 for sheep), which supports the significant distribution of the venom outside of blood. This means that an antivenom will be much less effective (potentially ineffective) in binding the venom once the venom leaves the circulation. Understanding the distribution of a venom between the central compartment and the peripheral compartments and, importantly, the rate at which a venom moves out of the central compartment is essential to determining the effective timing of antivenom administration.
Pharmacokinetic studies in animals were also performed following intramuscular or subcutaneous injection in order to mimic the real-life nature of snakebites. The initial phase of the concentration-time profile was governed by the absorption and simultaneous distribution of the snake venom (and toxins). Some studies observed multiple absorption phases which may be due to the difference in the absorption rate of venom components with a wide range of molecular weights (Tables 1 and 2). In addition, the lymphatic absorption from the injection site to the blood compartment may be associated with a delayed increase in venom concentration observed in these studies. The absorption of venom appeared to be relatively slow, on the basis of the prolonged elimination half-lives reported, especially following SC administration. This indicates that the absorption is likely to be the rate-limiting step for snake venom disposition, and that the 'flip-flop' phenomenon may govern the pharmacokinetics of snake venom. In contrast, some studies reported a similar elimination half-life following IV and IM injection, which indicates that the slow elimination is likely to be the rate-limiting factor of the terminal phase of snake venom time course.
Not all snake venoms are absorbed from the injection site, with their bioavailability being as low as 4% in some cases and as high as 81.5% in other cases. The unabsorbed components of snake venom retained at the injection site may possibly be responsible for local tissue damage. Some confounding estimates of V ss characterized by large values were observed in some studies (Guo et al., 1993 [27], Zhao et al., 2001 [68] and Paniagua et al., 2012 [66]), compared with the V ss following IV injection.
It is important to keep in mind that snake venom is a mixture of large and small molecular weight molecules and that there are limitations in using ELISA to measure such a mixture, because multiple different antibodies are used to detect multiple different antigens. Therefore, the venom concentrations measured and the pharmacokinetic parameters obtained in these studies are an averaged representation of the venom profile and may not capture the true characteristics or variability of the various snake toxins in the biological system. High-performance liquid chromatography with mass spectrometry would provide a more accurate measure of each toxin type, but such an approach to venom measurement has not been undertaken to date.

Pharmacokinetic Studies in Humans
In this meta-analysis, we were able to model literature-derived data to explore the pharmacokinetic parameters of snake venom for the first time. Timed venom concentration data from the literature were obtained from pre-antivenom samples described mainly in studies investigating patients who were intended to be administered an antivenom. The data extracted from the literature and modelled were best described by a one-compartment model. It is likely that venom may well display multicompartment pharmacokinetics in humans, but the available data that could be extracted were too sparse to support more than a one-compartment model. In this work, we also identified a bite-dependent exposure related to snakes of different families, indicating that the Viperidae family injects approximately a 75% higher dose. In the model, we noted a large between-subject variability in the relative dose injected (indicated as the between-subject variability in F1 from Table A1 of Appendix A). The variability between subjects in the remaining parameters, e.g., CL and V, was not wholly different from that of therapeutic drugs for which the normal variability in CL is approximately 50% in the case of small drug molecules [97]. In the model, the elimination half-life was calculated to be approximately 10 h. Of note, Audebert et al., 1994 also reported similar findings, showing a mono-exponential decline of snake venom concentration, with elimination half-lives ranging from 5 h to 12 h [75]. However, our results are in contrast to those of most animal studies in which two-compartment model fits are reported and hence the terminal half-life is often prolonged more than what observed here. Larger and richer venom concentration datasets in human envenoming are required to support multicompartment models and explore between-subject and between-snake variability, as well as other potential pharmacokinetic covariates.
When comparing our results to an allometrically scaled human half-life from animal data following IV administration of venom, a significant interspecies variability is observed. The allometric scaling of half-life is calculated as: where WT is the weight of the human (assumed 70 kg) or animal (assumed 0.31 kg for rats and 2.3 kg for rabbits) examined. Considering a terminal half-life of 5.3 h in rats (derived as the arithmetic average from the data in Table 3) and 16 h in rabbits (derived as the arithmetic average from the data in Table 3), the allometrically scaled half-life in humans was estimated to be 21 h and 39 h. The terminal phase half-lives calculated from rat and rabbit data were significantly longer compared to those in our study. The discrepancy in these results may be due to the sparse human data only supporting a one-compartment model compared to the two-compartment models for the rats and rabbits, considering that the half-life from the one-compartment model will be shorter than the terminal phase half-life from the two-compartment model. It is possible, therefore, that the terminal half-life in humans could be longer than one day. Further analysis needs to be performed to account for these differences and to identify outstanding covariates that may inform the terminal phase of exposure.

Conclusions
While there is a limited number of studies that investigated the pharmacokinetics of snake venom in animals and reported venom concentrations in humans, this review yields important quantitative information which can be used as a basis for the future development of snake venom pharmacokinetic models. A meta-analysis of snake venom pharmacokinetics in humans has led to the development of a preliminary model, which was able to describe literature-derived data. Further exploration of the relationship between parameters and covariates is necessary to examine the factors that may influence the pharmacokinetic profiles of snake venom and construct a substantial model from data in a larger series of snake envenomation cases. It is surprising that a relatively meaningful analysis was possible, despite the limitations of ELISA, the fact that venom is a mixture of small and large molecules, and the differences in snakes. The development of improved methods to measure venom is required, as well as studies focusing on single snake species or on similar types of snakes.

Pharmacokinetic Studies in Animals
We reviewed pharmacokinetic studies of the administration of snake venom to animals and summarised relevant data.

Literature Search Strategy
EMBASE (1974-present) and Medline (1946-present) were used to identify relevant articles. Keywords and text terms included "exp envenomation", "exp snakebite", "exp snake venom", "exp pharmacokinetics", "exp toxicokinetics". The search was limited to the English language and animals.

Inclusion and Exclusion Criteria
The publications were only included if they identified the venom or toxins administered, provided timed venom and toxin concentrations (in the absence of antivenom) or a compartmental analysis of their data measured by ELISA, and conducted a pharmacokinetic analysis. Studies that measured venom concentration using RIA were not included because of the possibility of an inaccurate determination of the pharmacokinetic parameters as a result of the metabolism or of the degradation of the radio-labelled proteins [98]. Review articles, commentaries, editorial papers, and conference abstracts were excluded.

Data Extraction and Creation of a Summary Table
The studies meeting all the inclusion criteria were grouped according to animal species. The pharmacokinetic parameters extracted from these publications included: half-life, bioavailability, clearance, and volume of distribution. A summary table was created showing the demographics of the animal studies, snake identification or venom or toxin type, type and number of animals, route of administration, dose administered, and pharmacokinetic parameters. When different units of doses or parameters were reported in the studies, they were converted to mcg·kg −1 for the dose administered, hours for the half-lives, L·kg −1 for the volume of distribution at steady state (V ss ), and L·h −1 ·kg −1 for the systemic clearance (CL).

Pharmacokinetic Studies in Humans
We reviewed all pharmacokinetic studies of snake venom in humans and extracted relevant data for further analysis.
The publications were only included for data extraction if they reported the identity of the snake, timed (post-bite) venom concentrations (in the absence of antivenom), and used ELISA to measure the snake venom concentrations. The review was limited to English language and humans. Review articles, commentaries, editorial papers, and conference abstracts were excluded.

Data Extraction and Synthesis
The snake venom concentration-time data post-envenomation were extracted by (1) obtaining the reported concentrations from the written text, or (2) digitising relevant data points from the concentration-time graphs published in the paper. The graphical data were digitised using WebPlotDigitizer v3.12 software (available at: http://arohatgi.info/WebPlotDigitizer/). All data points considered to have occurred after the administration of an antivenom were excluded.

Data Analysis
The extracted data were transferred to a Microsoft Excel spreadsheet and replotted as a concentration-time profile of different snake venoms from different studies. Since the data included repeated measures on many individuals, the data were modelled within a nonlinear mixed-effects modelling framework using NONMEM 7.2 (ICON Development Solutions, Ellicott City, MD, USA), and standard pharmacokinetic models were considered. The best model was determined using the likelihood ratio test, with degrees of freedom equal to the number of parameters different in successive models. The final model was evaluated using a prediction-corrected visual predictive check.

Conflicts of Interest:
The authors declare no conflict of interest.