Perioperative pharmacokinetics and pharmacodynamics of meloxicam in emus ( Dromaius novaehollandiae) of different age groups using nonlinear mixed effect modelling

Meloxicam is a widely used nonsteroidal anti-inflammatory drug in avian species. However, variability in pharmacokinetic (PK) and pharmacodynamic (PD) parameters in birds warrants species-specific studies for dose and dosing interval optimization. We performed a perioperative PK study of meloxicam (0.5 mg/kg, intravenously) on emus of three different age groups: 3 chicks (5 weeks old, 3.5 kg), 4 juveniles (26 weeks old, 18.8 kg) and 6 adults (66 weeks old, 38.8 kg). A two-compartment population PK model including weight as a significant covariate on clearance and central volume of distribution (V1) best fitted the data. The typical values (20 kg bird) for clearance and V1 were 0.54 L/kg/h and 0.095 L/kg. Both parameters significantly decreased with increasing weight/age. Meloxicam potency and selectivity for COX-1 and COX-2 were measured in whole blood assays (TxB 2 production endpoint). Meloxicam was partially selective in emus (IC 50 COX-1:COX-2 = 9.1:1). At the current empirical dose (0.5 mg/kg/24 hr), plasma meloxicam concentration is above IC 50 of COX-2 for only 2 hr. PK/PD predicted dose required for 80% COX-2 inhibition over 24 hr were 3.4, 1.4 and 0.95 L/kg/day in chicks, juveniles and adult emus, respec-tively. The safety, therapeutic efficacy and practicality of modifying the daily dose or dose interval should be considered for dose recommendations in emus.

including gastroprotection, renal protection and haemostasis.
Inhibition of this isoform by NSAIDs relates to their adverse effects of gastric ulceration and renal toxicity (Herschman, 1996;Warner et al., 1999). COX-2 is an inducible isoform expressed in response to injury or inflammation and is associated with the therapeutic benefits of NSAIDs. However, COX-2 has physiological effects on salt and water excretion and therefore COX-2 selectivity does not signify renal safety (Horl, 2010;Weir, 2002). Increased understanding of the differing roles of COX-1 and COX-2 has prompted a need for research into the selectivity of existing NSAIDs and the development of COX-2 specific NSAIDs.
There are marked species differences in the expression and activity of COX, conferring variation in selectivity, potency and safety of NSAIDs (Brideau et al., 2001). This can result in sub-optimal dose and dose frequency when choosing a dose empirically in veterinary species. For effective NSAID therapy, pharmacokinetic (PK) and pharmacodynamic (PD) studies performed on the target species with specific NSAIDs are required to ensure adequate analgesia with minimal side effects (Lees et al., 2004). There is minimal research into the PD and PK of NSAIDs in birds (Goessens et al., 2016;Houben et al., 2016), and none published with emus as the target species. A study by Baert and De Backer (2003) investigated the PK of three NSAIDs in five bird species noted the large variability in plasma clearance and half-life of meloxicam between species, with ostriches clearing the drug especially rapidly.
As a similar large ratite species, emus may exhibit similar PK of meloxicam and the empirical dose reported may be too low in this species.
Variability of PK parameters between individuals can result in differing plasma concentration-time courses despite administration of the same dose per unit body weight. This is due to intrinsic variation within a population of factors including genetics, disease status, age and concurrent drug therapy (Buclin et al., 2009).
Within paediatric populations, there are notable differences in physiology compared to adults, which alter PK. These included increased relative size of the liver and kidneys and therefore clearance, as well as varying body fat and extracellular fluid composition resulting in increased volume of distribution (Rane, 1992).
Hepatic maturation also brings changes in enzyme expression and activity which can result in altered metabolite production (Batchelor & Marriott, 2015). These differences should be considered when adjusting doses for paediatric individuals as scaling dose directly by weight may not be appropriate. In juvenile animals, similar changes in PK parameters have been reported, which can be linked to maturational changes in kidney and liver function, as well as in body water/fat composition (Gasthuys et al., 2017;. In this respect, growing conventional piglets have been recently studied as surrogate large animal model in paediatric drug research . This study hypothesizes that there are significant differences in the PK of meloxicam between emus of three different age groups and the empirical dosage regime may be too low in this species. This will be achieved through the following objectives: (a) investigate the effect of age on the main parameters after a single IV bolus by performing PK analysis of meloxicam for the three age groups under general anaesthesia, (b) compute the potency of meloxicam against COX-1 and COX-2 from whole blood assays through PD analysis and (c) calculate the duration of COX-2 and COX-1 inhibition and define an ideal dosage regime and anaesthetic protocol in this species.

| Study population & anaesthetic
The emus in this study were anaesthetised during a biomechanical ratite study for surgical implantation of bone strain gauges and electromyogram wires (Cuff et al., 2019;Main & Biewener, 2007).

The study was approved by the Royal Veterinary College Clinical
Research (CRERB approval number: CR2018-006-2) and Ethics and Welfare Research boards. It was carried out under the Animal (Scientific Procedure) Act (PPL70/7122).

| Blood sampling
Blood samples were collected for anaesthetic monitoring from another lumen of the right jugular catheter in heparin tubes prior to meloxicam administration (t = 0) and at predetermined times following (t = 2, 5, 10, 20, 40, 60, 90, 120, 180, 240 min) or until the end of surgery. The total sampled volume did not exceed 10% of each individual's blood volume. Samples were centrifuged (2,400 g for 5 min) to separate plasma and stored at −80°C until analysis.

| Determining plasma concentration of meloxicam
The method for determining plasma concentrations of meloxicam using validated high-pressure liquid chromatography with ultraviolet detection (HPLC-UV) was based on Baert and De Backer (2003). Sample preparation involved addition of 100 µl plasma to 100 µl methanol, 250 µl water, 25 µl internal standard (piroxicam, 10 µg/ml in methanol), 150 µl 0.1 M hydrochloric acid and 5 ml diethylether. Samples were extracted for 20 min, then centrifugated at 2,400 g for 5 min, the organic layer was transferred and evaporated under nitrogen at 40°C. The remaining residue was re-dissolved in 200 µl of the mobile phase (80 water/20 acetonitrile, v/v) and vortexed, and 50 µl was injected into the HPLC-UV system (see Appendix 1), with a mobile phase consisting of 0.1% acetic acid in water and acetonitrile in gradient elution. The flow rate was 0.5 ml/min. This method was validated prior to analysis. A calibration curve was created by spiking blank plasma with known concentrations of meloxicam, ranging from 0.075 to 5.0 µg/ml. Where concentrations fell below the LOQ of the HPLC-UV method, samples were re-run using liquid chromatography-tandem mass spectrometry (LC-MS/MS). A UHPLC system was coupled to a Xevo TQ-S triple quadrupole mass spectrometer (see Appendix 1). Mobile phase comprised 0.01% formic acid in water and 0.01% formic acid in acetonitrile. A gradient elution was performed on a Hypersil Gold column (50 x 2.1 mm i.d., dp: 1.9 µm) with a precolumn of the same type. The mass spectrometer operated with positive electrospray ionization and multiple reaction monitoring (MRM) was performed. Transitions monitored were m/z 332.30 (precursor ion) →95.00 (product ion) for piroxicam and m/z 352.30 →115.00 for meloxicam. Retention times were 2.29 min for meloxicam and 2.07 min for piroxicam. The limit of quantification was 10 ng/ml.

| Pharmacokinetic analysis
We have used the principle of nonlinear mixed effect modelling for the analysis of this unbalanced dataset (6/ 4/ 3 birds) (Bon et al., 2018). The coding of covariates in the model resulted a single PK model supported by the data from 13 richly sampled emus, from 3 age categories.

General model
Pharmacokinetic analysis was performed with Phoenix NLME 8.1 (Certara USA). All concentrations below the LOQ of the LC-MS/MS were excluded from analysis (8 out of 96 samples).
Noncompartmental analysis (NCA) was used to compute the mean standard PK parameters for each age group (Gabrielsson & Weiner, 2012) and estimate the effect of the age categories on clearance (CL) and volume of distribution (V).
Pharmacokinetic modelling using nonlinear mixed effect identified the optimal base model and assessed variability between individuals. One-and two-compartment models were compared using the maximum likelihood estimation objective function value (OFV), expressed as the negative sum of the log of the likelihoods (−2LL), to rank models by goodness of fit. A multiplicative residual error was used for fitting plasma concentration data. As the two-compartment population PK model had a significantly lower −2LL, this model was used to assess the effect of covariates on parameters. Random effects were added progressively to the model for PK parameters until no further reduction of the −2LL could be obtained while keeping eta(η)-shrinkage of the basic model around 20% maximum. A model with random effect on CL, V and V2 fitted the data the best.

Covariate model
Weight was explored as a continuous covariate to explain the differences in CL and V between the three age groups. Weight was used instead of age as it was accurately measured for each individual and would allow greater distribution of data within groups. The model with the lowest −2LL with minimal compromise of η-shrinkage was with CL and V adjusted by the covariate weight, according to following parametrization (average weight 20 kg) in Equations (1) through (4): where tv = typical value of the parameter for the population, η = individual's deviation from the typical value. The covariate terms dCl dweight and dV dweight adjust the population typical value Cl and V to account for weight.
The final two-compartment population PK model with weight as a covariate for CL and V was simulated with 1,000 iterations in a visual predictive check (VPC) plot (diagnostic plots in Appendix 3).

| Whole blood assays
The method for performing the whole blood assay in emus was adapted from . At the conclusion of the biomechanics study, blood was collected from 4 adult emus immediately after euthanasia in heparinized tubes (20 Unit/ml), then divided into 500 µl aliquots to be used for either the COX-1 or COX-2 assay.
For the COX-1 assay, following incubation at 37°C for 1.5 hr, aliquots were spiked with concentrations of meloxicam 0.02-500 µM or saline as a positive control. After 4 hr, 50 µM calcium ionophore (A23187, Sigma Aldrich Co Ltd, Dorset, UK) was added to induce platelet activation and eicosanoid generation, in particular thromboxane (Tx)B 2 . Four hr later, the reaction was terminated by addition of 800 µM meclofenamic acid (Sigma Aldrich) to inhibit any secondary COX stimulation and chilling to 4°C. Plasma was separated off by centrifugation prior to freezing.
In the COX-2 assay, aliquots were pretreated with 10μg/ml aspirin in dimethyl sulfoxide (Sigma Aldrich) to inhibit COX-1  and incubated at 37°C for 1.5 hr. Samples were spiked with concentrations of meloxicam from 0.005-200 µM or saline as a positive control. After 0.5 hr, lipopolysaccharide (LPS Escherichia coli serotype 026:B6, Sigma Aldrich) was added to samples to 100 µg/ ml to induce COX-2 activity in blood monocytes. The reaction was terminated at 8 hr as for the COX-1 assay.
Samples were prepared by transferring 50 µl of plasma to an extraction tube, adding 25 µl internal working solution (25 ng/ml), 25 µl methanol and 150 µl of water. Samples were vortexed with 25 µl 0.1 M hydrochloric acid. 3 ml n-hexane/ethyl acetate (1/1, v/v) was added for extraction over 25 min. Following centrifugation for 10 min at 1,800 g, the supernatant was transferred and evaporated under a nitrogen stream (40 ± 5°C). The dry residue was re-dissolved in 125 µl of methanol in water (1/9, v/v) and filtrated using a Millex PVDF 0.22 µm syringe filter, then transferred to an autosampler vial. 10µl aliquots were injected onto the same LC-MS/MS system as mentioned above. Chromatography was performed on an Acquity UPLC® BEH C18 column (50 × 2.1 mm i.d., dp: 1.7 µm), using a gradient elution with 0.05% acetic acid in water (A) and acetonitrile (B) as mobile phases. The flow rate was 0.3 ml/min. The mass spectrometer was operated in the positive electrospray ionization mode, and two MRM transitions were monitored per component, as shown in Appendix 2. The method was validated, and the following parameters were evaluated: linearity, within-run precision and accuracy, between-run precision and accuracy, limit of quantification and carry-over and specificity. The limit of quantification was between 0.2 and 0.5 ng/ml, depending on the prostanoid.

| Pharmacodynamic modelling
Inhibition of TxB 2 production by meloxicam was used as a marker of both COX-1 and COX-2 activities and expressed as percentage of the control value (meloxicam concentration = 0 µM). Four curves for COX-1 inhibition of TxB 2 and one of COX-2 inhibition could be used for PD modelling of the four birds, using Phoenix NLME.
COX-1 inhibition was initially modelled as a naïve average of the 4 birds with additive residual error with the Hill equation (Equation

5):
Where I = percentage inhibition of COX-1 activity compared with control, I 0 = baseline inhibition, I max = maximal possible inhibition, IC 50 = concentration of meloxicam producing 50% inhibition of COX-1 activity, C = total concentration of meloxicam and γ (gamma) = slope of the concentration-effect curve. A population-based PD model with additive residual error using a Laplacian algorithm was used to investigate inter-individual variability. Models with random effect on single and multiple parameters were compared with identify the one with the lowest −2LL and shrinkage. Simulation with 500 iterations of the final model was performed to produce a VPC plot (diagnostic plots in Appendix 4).
As only one curve of COX-2 inhibition was produced, this curve was modelled without random effect, using an I max model with additive residual error. Models were compared with varying γ and with γ equal to 1. The simple I max model was selected as it had the lowest Akaike Information Criterion (AIC). TA B L E 1 PK parameters of meloxicam in emus of three different age groups (Chicks n = 3, Juveniles n = 4, Adults n = 6) after administration of a single IV bolus of 0.5 mg/kg

| | Statistics
The statistical significance of differences between PK parameters computed by NCA in each age group was tested with a Kruskal-Wallis test (p < .05) using GraphPad Prism (version 8.1.1, GraphPad Software, La Jolla California, US).

| Noncompartmental analysis
Mean PK parameters determined by NCA are summarized in Table 1.
Meloxicam CL and V (per unit body weight) were significantly higher in chicks than adult emus (p < .05). Figure 1 describes the relationship between meloxicam concentration (observed and fitted) and time in different age groups (VPC plot obtained with 1,000 iterations). It was not possible to predict beyond 2 hr for the chicks and beyond 3 hr for the adults due to lack of data above the LOQ. Hence, the predictions beyond 3 hr are based solely on the data from juveniles and may not be representative of the population.

| Population-based two-compartmental PK model
All birds cleared the drug relatively rapidly, with chicks having the lowest area under the curve (AUC) due to an especially high CL.   (6) and (7)

| Pharmacodynamic analysis
Pharmacodynamic parameters describing inhibition of TxB 2 production by COX-1 and COX-2 by meloxicam are summarized in Table 3, and concentration-inhibition curves are represented in Figure 3.  TA B L E 2 PK parameters of meloxicam in emus as determined by a twocompartment population pharmacokinetic model with first order elimination with Body Weight as a covariate F I G U R E 2 Relationships between meloxicam PK parameters and weight as predicted by population-based two-compartment PK modelling. Figure  occurred. This effect was similarly noted by  for meloxicam and carprofen and by Kay-Mugford et al. (2000) for meloxicam, tolfenamic acid, carprofen and ketoprofen. The percentage inhibition was therefore rescaled (0%-100%), and the total meloxicam concentrations converted from µM to µg/ml to calculate selectivity indices of meloxicam for COX-1 versus COX-2 (Table 4), as in Giraudel et al. . Meloxicam was COX-2 selective with values of IC 50 COX-1 of 158 µg/ml and IC 50 COX-2 of 17.5 µg/ml, yielding a ratio IC 50 COX-1/COX-2 of 9:1. As the two inhibition curves were not parallel due to the higher sensitivity of meloxicam for COX-1 versus COX-2 (γ = 1.232 COX-1 and γ = 1 for COX-2), the ratios of COX-1:COX-2 inhibition were largest at IC 20 and smallest at IC 80 . Other IC ratios have been included to assess clinically relevant levels of inhibition in terms of balancing therapeutic efficacy (high inhibition of COX-2, ideally more than 80%) with safety (low inhibition of COX-1, ideally less than 20%).

| Dose computation
The method for evaluating the empirical dose of meloxicam (0.5 mg/ kg) was adapted from  using results of the PK/ PD analysis. As PD values obtained with a whole blood assay account for drug protein binding, in vitro inhibitory concentrations are comparable with in vivo therapeutically relevant concentrations.
When predicting doses using plasma clearance from PK analysis, a Note: I = percentage inhibition of COX-1 activity compared with control, I 0 = baseline inhibition, I max = maximal possible inhibition, IC 50 = concentration of meloxicam producing 50% inhibition of COX-1 activity, C = total concentration of meloxicam and gamma = slope of the concentrationeffect curve. Stdev0 = residual error. IIV inter-individual variation and η-shrinkage eta-shrinkage (random effect). −2LL: negative sum of the log of the likelihoods AIC: Akaike information criterion TA B L E 3 Summary of pharmacodynamic parameters of COX-1 and COX-2 inhibition by meloxicam by whole blood assay. COX-1 results were obtained from a population-based PD model with additive residual error with random effect on IC 50 and I 0 (n = 4). COX-2 result results were obtained from a simple I max model with additive residual error with γ (slope gamma) equal to 1 (n = 1)

F I G U R E 3
Sigmoid curves of percentage inhibition COX-1 (blue diamonds for observed average % inhibition ± SEM from 4 birds and continuous blue line for fitted regression) and COX-2 (green circle and regression line) by meloxicam

Meloxicam cyclooxygenase inhibition curves
COX1 fitted COX2 fitted COX-2 data COX-1 data mean +/-SEM multiplying factor of 1.5 was used to convert to whole blood clearance (emu blood haematocrit averaged 33.5%). Predicted doses were calculated using Equation (8) and displayed in Table 5: Where the target concentration is IC x% COX-1 or COX-2 and F (bioavailability) = 1. The current empirical dose (0.5 mg/kg/24 h) in emus is too low as COX-2 inhibition is only above 50% for an average of 2 hr (72 min in chicks versus 138 min in adults). Predicted doses required for 80% inhibition of COX-2 are between 1.9 (adults) and 6.7 (chicks) times higher than the empirical dose. Using the relationship between weight and clearance (Figure 2), Figure 4 illustrates predicted doses for 80% inhibition of COX-2. For birds of all ages, COX-1 inhibition up to 20% for between 48 min (chicks) and 105 min (adults). No doses have been evaluated for safety of in vivo dosing in emus.

| D ISCUSS I ON
This study determined the PK parameters of meloxicam in emus (CL = 0.54 L/kg/h, V = 0.095 L/kg, for a typical 20kg bird) and demonstrated significant species differences to comparable studies ( Table 6). The clearance was similar to ostriches, another larger ratite species. Clearance was high compared with other bird species except for red-tailed hawks. While volume of distribution in emus is not as high as in ostriches, it is similarly higher than in mammalian species (not showed in Table 6). The large variability in PK between avian species is one of the main issues with analgesic therapy in birds as cross-species allometric scaling is not reliable and therefore PK/PD analysis is required in every species (Baert & De Backer, 2003).
The emus in this study were under general anaesthesia, and concurrent drugs may have an effect on PK parameters. The alpha 2-adrenergic agonist xylazine used in this study's anaesthetic protocol may have altered hepatic blood flow. The presence (8) Dose Dosing intervel = plasma clearance * 1.5 * target concentration F TA B L E 4 Indices describing selectivity of meloxicam in emus for COX-1 versus COX-2 as determined by interpolation of inhibition curves using the Hill equation I = I 0 +(I max *C γ )/((IC 50 γ + C γ )) using values obtained from PD analysis Note: IC x : concentration of meloxicam producing x% inhibition TA B L E 5 Dose extrapolation of meloxicam from PK and PD results. IC 80 COX-2 as determined by PD analysis. Plasma clearance used from population PK two-compartment model. Whole blood C max from plasma C max from NCA divided by 1.5 (plasma/blood conversion ratio). Percentage inhibitions were extrapolated from rescaled sigmoid curves. Time above IC X estimated from concentration-time curves in Figure 5 and expressed as time for the 50% quantile of the population (5% quantile -95% quantile). Range to >180 indicates the time upper quantile was greater than 3 hr and therefore not determinable from the adult emus' concentration-time graph of an alpha 2 -adrenergic agonist can reduce clearance by 50% (Bennett et al., 2017;. This may be a confounding factor causing the increased clearance in chicks, as two out of the three received half the xylazine dose (1.5 mg/kg versus 3 mg/kg) of the rest of the study population.
While the chick receiving the 3 mg/kg xylazine dose had the lowest clearance of all chicks (1.22 L/kg/h), its clearance however remained much higher than one of the juvenile and adult emus, suggesting that xylazine dose was not a major influencing factor in this case. for robenacoxib in cats).
From well-stirred venous equilibration model (Benet & Hoener, 2002): Because the renal contribution to the systemic clearance of robenacoxib is negligible in cats (2% excreted as unchanged), and assuming this is the same in birds, we can also write the following: Considering the cardiac output of 37.5 kg Emu at rest 67.9 mL/ kg/min (Grubb et al., 1983), the extraction ratio (CL/ cardiac output) of meloxicam in emus was E meloxicam = 6.45/ 67.9 = 0.095 (present adult Emus clearance 0.387 L/kg/h). This is between a low clearance (E < 0.05) and medium clearance (E = 0.15), as per Toutain and Bousquet-Melou (2004).
For drugs exhibiting low extraction (Q hepatic » f u . Cl int ), the hepatic clearance can be approximated by the product of the unbound fraction and the intrinsic clearance as follows: Hepatic NSAID elimination is therefore perfusion-independent and only dependent on the (a) free fraction of NSAID within the plasma and (b) the intrinsic enzyme activities of the liver (Anderson & Holford, 2007). For the first (a), enzyme activities will change with emu maturation, but not because of anaesthesia provided that normothermia is maintained. For the second (b), anaesthesia may change the free fraction. However in these circumstances, and as described by Benet and Hoener (Benet & Hoener, 2002), the free steady state concentration which is pharmacodynamically active has no dependency upon the free fraction. Consequently, potential changes in protein binding due to anaesthesia would have little impact on the effect of meloxicam for postoperative pain relief medication in emu.
Chicks had significantly higher clearance and volume of distribution per unit weight than juvenile and adult emus (

Dose-weight curve
Meloxicam is metabolized to inactive metabolites by cytochrome P450 isozymes and oxidation, which are then excreted in faeces and urine (Grude et al., 2010). Hepatic metabolism is affected by the increased relative size of the liver in neonates resulting in increases hepatic blood flow relative to body weight and therefore clearance (Strolin Benedetti et al., 2005). Maturation of enzyme expression and activity also affects clearance to a variable degree (Koukouritaki et al., 2004). Renal clearance is increased with increased relative size of the kidneys in neonates; however, as less than < 0.25% of meloxicam is excreted as its active compound, this will not have significantly affected clearance in chicks (Batchelor & Marriott, 2015).
Selectivity of NSAIDs for COX-1 and COX-2 varies between species and assays used (Schmid et al., 2010). The ratio IC 50 COX-1:COX-2 in emus was 9.1:1, similar to that in dogs (value of 7.3:1 from ), indicating meloxicam is COX-2 preferential in this species. Whole blood assays most closely represent in vivo physiological conditions and are most useful at predicting clinical relevance. IC 20 COX1:IC 80 COX-2 indicates selectivity relating to clinical use, essentially a ratio of therapeutic efficacy to side effects.
This was 0.735:1 in the emus in this study, higher than that in dogs (0.46:1) and cats (0.25:1) Schmid et al., 2010). Two limitations of the COX-2 assay were that that only one curve was  Lees et al. (1991) produced, and the best signal obtained was from a TxB 2 assay. In mammalian assays, PGE 2 conventionally indicates COX-2 activity; however, no PGE 2 was detected in the COX-2 assay in this study. This may be due to PGE 2 playing a less significant role in avian inflammation than in mammals, as indicated by Golemboski et al. (1990) and Fraifeld et al. (1995), and therefore, a different assay may be required (using TxB 2 for example). While both COX-1 and COX-2 genes are encoded within the emus genome, their roles in the inflammatory response remain uncertain (Altschul et al., 1997) and there is a lack of information on inflammation and biomarkers in avian species.
In a urate synovitis model (Anhut et al., 1979), the PGD 2 response was higher in chickens synovial fluid compared with PGE 2 , but the PGE 2 response was still measurable (and more substantial than the TxB 2 one). In our COX-1 assay, PGE 2 was also produced and measured (though not in quantities as high as TxB 2 ), but not in the COX-2 assay. In the validation of our ex vivo assay, PGE 2 could however be measured in the absence of aspirin (either with or without LPS, corresponding to a COX-1 contribution), but could not be mea- This response could be significantly suppressed by ketoprofen, but this does not rule out an in vivo effect of LPS on COX-1 predominantly compared with COX-2.
NSAIDs have variable safety in avian species with diclofenac associated with significantly high mortality in vultures due to carcass consumption (Swan et al., 2006). High dosing with NSAIDs is associated with intestinal lesions, degenerative bone changes and renal impairment (Palocz et al., 2016). The results obtained indicate the current empirical dose of 500 µg/kg/24 hr will inhibit 50% COX-2 for 2 hr or less and therefore is too low. The predicted therapeutic dose required for 80% COX-2 inhibition for 24 hr was 3.375 µg/kg in chicks, down to 949.3 µg/kg in adult emus. While meloxicam is the safest NSAID in birds, raising the dose to the level predicted in this study may result in adverse side effects (Cuthbert et al., 2007).
A dose of 1 mg/kg has been recommended for African Grey and Hispaniolan Amazon parrots and may be more appropriate for adult emus (Hawkins et al., 2018). Therapeutic analgesia in emus could be achieved through a single higher IV dose, repeated IV administration, a longer-acting subcutaneous formulation or using a multi-modal approach including opioids and nonopioids. The safety and practicality of this should be considered in particularly with younger emus, due to their higher dose requirements.

| CON CLUS ION
Meloxicam is a widely available analgesic with safety demonstrated in many avian species; however, doses used are often empirical due to lack of data in individual species. This may lead to inadequate analgesia due to the high degree of variability of PK and PD in different veterinary species. This study demonstrates that emus clear meloxicam more rapidly, have a higher volume of distribution than many mammalian species and display similar PK parameters to ostriches.
This difference in drug exposure (AUC) was lower in younger emus than in adults, which may be due to weight and developmental stage or due the anaesthetic protocol used. Our PD analysis using the first whole blood assay generated in birds, further indicates the empirical dose used was too low in this species for once daily dosing and requires revision.

ACK N OWLED G M ENTS
We would like to thank the animal technicians from BSU, the staff from the structure and motion laboratory and Dan Hampshire who helped during the biomechanical and ratite anaesthesiology studies.

CO N FLI C T O F I NTE R E S T
The author declares no conflict of interest.

AUTH O R CO NTR I B UTI O N
The authors contributed to this study as follows: LP contributed to conceptualization, project administration, overall supervision, data modelling and resources (  319.20 c 19 30 C ion used for quantification. For the COX-1 assay, PGF2α and to some extent PGE2 were also measured and suppressed by meloxicam. In one Emu, PGFM, BC-PGE2, 6-keto-PGF 1α , 11-dehydro-TXB 2 , PGB 2 and Δ-12-PGJ2 were measureableFor COX-2, TxB 2 was measured in 3 birds (1 curve exploitable), PGF 2α was measurable in in 2 birds.For both assays, PGEM was positively correlated with meloxicam concentration (R 2 = 0.9360) and initially shared the same m/z ratio until the setup was refined. None of the other prostaglandins or metabolites ever reached quantifiable levels. a m/z = mass to charge ratio. b CE = collision energy.

A PPE N D I X 3
Diagnostic plots of PK analysis of meloxicam in Emu of different age groups after administration of 0.5 mg/kg intravenously