Population modelling in veterinary medicine : - PowerPoint Presentation

1. Population modelling in veterinary medicine : an introductionPierre-Louis ToutainRoyal veterinary College London & project officer at the ENV of Toulouse Wuhan University October 2017

2. To explain the main concepts related to population PK/PD investigations To illustrate with some examples how this approach may be useful in veterinary medicine: Ojectives of the presentation2

3. Objectives of the presentationIt is not to learn population PK/PD analysis…

4. POP PK analysis : reviews, textbooks and regulatory guidelines

5. Pharmacokinetic-Pharmacodynamic Modeling and Simulation. PL Bonate;. 2011 second edition . New York: SpringerThis book is written for the kineticist who performs PK/PD modelingIt is expected the reader has basic knowledge of PK and simple PD models.The reader is also expected to have had a 1-year introductory course in statistics that covers basics of probability, regression, and analysis of variance. A 1-semester course in matrix algebra is desired but not needed.

6. Reviews for an introduction to Pop PK/PD (open access)

7. Publications in population PK: reviews

8. Guidelines: EMA 2007

9. Guidance FDA 1999

10.

11. POP PK: definition

12. Population PK/PD: generic definitionStudy of the variability in drug concentration or pharmacological effect between individuals when standard dosage regimens are administeredA. Arons Br. Clin. Pharmacol. 1991, 32: 669

13. Goal of Population kineticTo identify the sources (factors or covariates) of PK and PD variability in the target animal population as well as the magnitude of that variability, in field or clinical condition (not in lab condition) and to explain variability by covariates to make recommendations

14. Ultimate goal of population PK/PDAdjust the dosing regimen for a given patient or a group of patients (e.g. for a given breed) taking into account relevant covariatesAnd also to make recommendations: management issues e.g. size of a pen to reduce BSV , withdrawal time and health status etc,

15. The difference between optimal population and optimal individual doseThe range of dose-responses for efficacy ( green) and adverse effects (blue) in a population is shown, with the optimal dose selected to provide the best efficacy without too many adverse effects. However, patient A (orange curves) experiences many adverse effects at the population dose yet responds well to a lower dose, whilst patient B ( purple) experiences little efficacy at the population dose but can tolerate a higher dose.

16. NONMEMNon-Linear Mixed Effect Model A statistical model non-linear in the regression parameters and non-linear in random effectsThe tool for pop PKPD

17. 1-Population PK to investigate variability

18. Variability is a biological fact of interest

19. Variability is a biological fact …..

20. Variability is not a “noise” but a biological fact that needs to be measured and explained by covariatesPopulation PK/PD modeling

21. Veterinary medicine can face huge sources of variability between animals due to the existence of breeds, collective treatments….

22. 22Sources of variability well documentedspeciesfoodagesexdiseasesPKPDDosePlasma concentrationEffectBODYReceptorOften ignoredSubpopulations/responders/non-responders; health status….

23. Sources of variability in drug response

24. Sources of Intersubject Variability Demographic Age, Body Weight or Surface Area, gender, raceGenetic: CYP2D6, CYP2C19Environmental: Diet, temperature (fish)…Physiological/Pathophysiological: Renal (Creatinine Clearance) or Hepatic impairment, Disease State Behavior associated to collective treatmentsSocial, food behavior…Concomitant Drugs

25. Two types of covariatesContinuousAge, BW, creatinine….CategoricalSex, breed, polymorphism, formulation, pain score

26. Age as covariate: bioavailability of Cefadroxil in foalAge(months)0.51235F%99.667.635.119.514.4Duffee JVPT 1997 20 427POP modeling: continuous covariate

27. Polymorphism: COX-2 inhibitorsPOP modeling: categorical covariate

28. Variability vs. noise

29. Should be: recognized measured & Explained by covariablesShould be suppressed (bias) or minimized (precision)

30. Experimental vs. population approach:the status of variabilityExperimentalviewed as a nuisance that has to be overcomeObservational Populationrecognized as an important feature that should be identified, measured and explained (covariates)

31. Experimental vs POP PK/PD trialObjective: document mean and variability of PK/PD data for THE whole population Hence the term “Population”This implies to study a representative sample of this population Individuals are randomly sampled in the populationObjective: test one hypothesis (e.g. fed/unfed, pionneer/generic)Hence consider interindividual variability as a nuisance that should be reducedThis implies to select homogenous subjects Individuals may not be considered as randomly sampled in the population Analysis subject per subject; two stages approach Simultaneous analysis of subjects altogetherPopulation (POP) PK & PDExperimental PK & PD

32. Variability may be noise in some circumstances There is sometime advantage (for statistical reasons) in limiting or even suppressing sources of variability when performing some study (tests of hypothesis)Bioequivalence trial (only the formulation factor is investigated)Fed vs unfed bioavailability

33. Experimental vs. observational population studies and the inference space??

34. Variability is often viewed as a nuisanceVariability can be viewed as a nuisance rather than a valuable information and may be deliberately suppressed :e.g.: healthy animals rather than patientse.g.: beagle dogs rather than mongrel dog

35. Why to suppress variability?There is advantage (for statistical reasons) in limiting or even suppressing sources of variability when performing a laboratory studyQuestion: what is the universality (relevance) of the conclusion generated under these conditions ? (question of the inference space)

36. Variability is a biological and relevant fact, not a noiseThere is sometime advantage (for statistical reasons) in limiting or even suppressing sources of variability when performing a studyBioequivalence trial (only the formulation factor is investigated)There is often disadvantage to suppress variability e.g.: healthy animals rather than patientse.g.: beagle dogs rather than mongrel dogQuestion: what is the universality of the conclusion generated under these conditions

37. Standardization: the proPressure to promote standardized approachSimple for reviewersWelcome by regulatory bureaucrats under financial, staffing and time pressureReduce opportunities for arbitrary regulatory discretionFavor global trade (WTO)e.g. bioequivalence criteria

38. Standardization: the consForcing science to offer universalized solutionStandard, narrow, technical version of sound scienceBlock innovation Not ethical: average patient do not existPublic health issues: e.g. what is the value of an universal withdrawal time as obtained in 16-20 cattle?

39. An example of biological variability to explain :What are covariates able to explain between subject variability of doxycycline exposure in pigs when doxycycline is given as a mass medication in feed?

40. PK variability of doxycycline administered to pigs215 pigs (30-110 kg)12- 15 pigs/pen (15 m2)Doxycycline given in meals twice a day at 5mg/kg/mealPK : large interindividual variabilityAUC0-24h distributionPlasma concentrationsDel Castillo et al. (2006)

41. PK Variabilityn = 215Doxycycline

42. What is the origin of this large AUC distribution ?

43. Doxycycline concentration variability: population vs. experimental trial1/27/2018Pop Kin sep/07 - 43Number of data pointsTrialPopulation n=215Experimental n=15 to 190461224Time (h)0.00.51.01.5DOXYCYCLINE (µg/mL)

44. Doxycycline concentration variability: population vs. experimental trial for time 6h post-administration1/27/2018Pop Kin sep/07 - 4401230.00.51.01.5DOXYCYCLINE (µg/mL)Number of data points1: Population n=2152: Experimental n=163: Experimental n=64

45. Pop breakpoint rvc 2008Doxycycline : sex effect Time (h)DoxycyclineSexe 0Sexe 1

46. Pop breakpoint rvc 2008Variability analysis: AUC vs. body weight

47. Doxycycline : disease effectTime (h)Concentrations (µg/mL)healthydiseased

48. Doxycycline : body temperature effectRectal temperatureDoxycycline

49. PK doxycycline variability analysis: conclusionsNo observed covariable was able to explain variabilitySocial behaviour (food behaviour, hierarchy etc. is likely the most important factor of variability

50. Pharmacokinetic variability of fosfomycin administered to pigs in food and water: impact of social rank

51. The aim of this trial was to document the effect of social ranking on the internal exposure of pigs to antibiotic administered either in food or in drinking water in a commercial setting. Fosfomycin was selected as the test antibiotic

52. Fosfomycin given in food or drinking waterFeed and water consumption was recorded. Animals were identified using a system of video cameras (equipped with night vision and wide-angle lens)

53. Comportamiento alimenticio del los animales: Consumo Pharmacokinetic variability of fosfomycin administered to pigs in food and water: impact of social rankde agua y alimento

54. Feeding ParametersValuesAmount of food consumed/day/animal (g)1318 ± 190Ingestion rate (g/min)35.0 - 40.4Number of meals/day7.57Number of visits to the feeders and drinkersFeeder 136.05Drinker 87.30Intake Duration/day (min)39.1 ± 2.2Intake Duration /meal (min)5.04 ± 1.40Amount of food consumed/meal (g)156 - 202Table 2: Mean (±SD) values for dietary behavior of pigs (n = 36).

55. Fosfomycin given in food or drinking waterFeed and water consumption was recorded. Animals were identified using a system of video cameras (equipped with night vision and wide-angle lens) Hierarchy levelFosfomycin AUCFosfomycin given in feedFosfomycin given in waterHierarchy levelFosfomycin AUCEffect of social ranking on fosfomycin exposure R=0.53R=0.63

56. Variabilidad concentraciones plasmáticasATB administrado en el alimento

57. Variabilidad concentraciones plasmáticasATB administrado en el agua

58. Use of population PK/PD What about covariate-based dose adjustment?If animal social behavior is the main factor, variability (inter-individual but also intra-individual variability) cannot be reduced by usual covariate dose adjustment but by better farm practice (husbandry, food preparation and administration)Antibiotic in foodReferenceTestHigh competitionLow competitionAUC

59. Veterinary examples of Population PK to investigate variability

60. Examples of questions to document using POP PK In pigs: Does the same dose of an antibiotic should be given for prophylaxis, metaphylaxis or for curative treatment?In cattle for susceptibility testing: what is the PK/PD breakpoint for a new antibiotic; what is the withdrawal time for an extralabel use of a drugIn cats: do we have to adapt the dosage regimen in case of a mild renal failure?In dogs oncology: the body surface area matter ?In horses: what is the appropriate withdrawal time for doping controlfish: what is the influence of water temperature on an antibiotic exposure

61. POP PK in dogs & cats

62. Population kinetics in dogs: possible objectivesTo investigate (simultaneously in clinical conditions) Breed effectDisease (kidney, liver..)Food effectsSexAgeBWDrug-drug interaction

63. Breed effects & Dose individualisation in petsSame dose for all ?Does the beagle dog relevant for domestic dogs? And for cats?

64. Breed effects & Dose individualisation in petsSame dose for all ?Does the beagle dog relevant for domestic dogs? And for cats?

65. Population PK to describe and explain variability: antibiotics

66. Use of Population PK to revise the dosage regimen of mavacoxib in dogsThe PK of mavacoxib differs considerably between young adult laboratory Beagle dogs (or Beagle-sized Mongrel dogs) and the typical geriatric large-breed osteoarthritic patients. The typical Beagle t1 ⁄ 2 is estimated to range from about 15 days (Supporting Information) to 17 days (Cox et al., 2010)but the typical osteoarthritic patient t1 ⁄ 2 (i.e. the t1 ⁄ 2 in a 35 kg, 10-year-old patient that is not a German shepherd or Labrador retriever) is 44 daysApproximately 5% of the osteoarthritic patients were found to have an unusually prolonged t1 ⁄ 2, i.e. t1 ⁄ 2 > 80 days and at least twice the typical t1 ⁄ 2. No covariate factor was associated with prolonged t1 ⁄ 2, and the long t1 ⁄ 2 patients spanned a diverse range of BWs, breeds and ages.

67. Half-lives of mavacoxib in dogs15-17 Days44 daysHealthy beagleOld osteoarthritic dogs

68. Polymorphism within breed:the case of celecoxibPKExtensive Metaboliser (45%)Poor metabolizer (55%)Half-life (h)1.72  0.795.18  1.29Clearance (ml/kg/min)18.2  6.47.15  1.41Paulson et al (1999) Drug Metabolism and Disposition 27, 1133-1142

69. POP PK in pigs

70. Population kinetics in pigs: possible objectivesTo investigateHusbandry/managementSocial behavior (hierarchy, food intake…)BreedAge (piglets vs. adult)DiseaseOther drugsDietBreakpoint for AST…..

71. Population PK to describe variability :e.g. to establish a breakpoint for Antimicrobial Susceptibility Testing (AST)

72. POP PK in cattle

73. Population kinetics in cattle: possible objectivesTo investigateAge (preruminant vs. ruminant)Type of production/husbandryBreed effectPregnancy /LactationDiseaseManagement/nutritionResidues/withdrawal time

74. POP and Withdrawal time for residuesThis study questions the use of small groups of healthy animals to determine WDTs for drugs intended for administration to large diverse populations. This may warrant a reevaluation of the current procedure for establishing WDT to prevent violative residues of flunixin.

75. POP PK in horses

76. Population kinetics in horse: possible objectivesTo investigateBreed (ponies vs thoroughbred vs draft horse…)Age (foal vs. adult)Sex (male, gelding, female)Exercise (doping control)DiseaseOther drugsDiet…..

77. Individual prediction (Bayesian)Type of Q: what is the delay for a given horse to be lower than a screening limit for its urine given the population model plus 2 plasma concentrations obtained from this horse?

78. Other applications of the NLME

79. Other applications of NLMEThe NLME modelling can be (should be) used to perform:Meta-analysis by aggregating optimally data, more or less unbalanced, collected from different sources (our example for florfenicol)To analyze +/- small data set of sparse data : e.g. toxicokinetics, exotic species, synovial fluids

80. 4-Population modeling for meta-analysis

81. Veterinary publications in Pop PK : meta-analysisStatistically, meta-analysis is a tool designed to summarize the results of multiple studies. It has been utilized in human drug development to assess the clinical effectiveness of healthcare interventions by combining data from different trials.Combined data can be analyzed using nonlinear mixed effect (NLME) modeling approaches not specifically designed for this purpose. This review outlines the procedures

82. Meta-analysis: a reviewStatistically, meta-analysis is a tool designed to summarizethe results of multiple studies

83. Population PK to describe variability :e.g. to establish a breakpoint for Antimicrobial Susceptibility Testing (AST)

84. Body weight & Allometric relationship

85. Meta-analysis to document residues and Withdrawal time for penicillinLi et al Interspecies Mixed-Effect Pharmacokinetic Modeling of Penicillin G in Cattle and Swine AAC 2014 pp4495-4503

86. Meta-analysis to document residues and Withdrawal time for penicillin

87. Making use of sparse samples taken in clinical studies with NLME:

88. The objective of this study was to model the PKs of robenacoxib using NLME to leverage the variety of information obtained from sparse and rich datasets for the appropriate assessment of the drug kinetics and its between-subject variability in cats.

89. Question: should NSAIDs be given before or after surgery?Spay clinic, 60 clinical cats recruited36 cats received robenacoxib (2mg/kg) on admission, just before or just after surgery.Need to estimate individual exposure to relate to effects on renal function1 to 2 blood samples per cat

90. Data from both IV and SQ routes were modeled Influence of parameter correlations and available covariates (age, gender, bodyweight, and anesthesia) on population parameter estimates were evaluated; No dosing adjustment based on available covariates information is advocated .

91. Simultaneous modeling of sparse and rich PK dataCombining IV and SC routes to investigate flip-flopMaking use of sparse samples taken in clinical studies with NLME: Use preclinical data from highly informative individuals to leverage information about PK in clinical patients (sparse). Possible with planning and modeling effort

92. Pooling data from different clinical trials A PPK model can be developed using data pooled from healthy volunteers who participate in phase 1 studies and from patient enrolled in specific clinical trials

93. Population modeling for sparse and unbalanced data

94. Two points not to be confuseThe objectives of pop PK/PDDocument variabilityThe tool that is usedThe NLME modelingFor pop PKPDBut also for non pop PKPD when sparse or unbalanced data

95. Sparse & unbalanced data Sparse data - only a small number of samples obtained from each subject. Not possible to fit each subject’s data separately. Toxicokinetics in ratsFish PKNot enough sampling times for some subjectsUnbalanced design: a study design in which all the subjects participating to the study do not supply the same number of observations

96. Population PK for unbalanced data Unbalanced design: a study design in which all the subjects participating to the study do not supply the same number of observationsTulathromycin

97. Population PK modeling for sparse dataAJVR 2003

98. Pop PK for exotic species

99. Sparse data analysis vs. population analysisThe 2 concepts are not synonymsSparse data analysis in toxicokinetics is not a « population question »The goal is not to document variability in a population of rats but to analyze properly a set of sparse data to avoid major bias due to possible unbalance …Conversely, it is possible to perform population kinetics with rich data

100. Sparse data for non-population investigation

101. 3-Population PK for predictions

102. Bayesian forecasting The full PK of a given animal can be predict using only a few blood samples if information of the corresponding whole population exists

103. Prediction of a withdrawal timeExtralabel use of a drugThe EU definition of WT for milk is typically a POP definition

104. POP PK/PD

105. Non POP PK/PD investigation