The effect of a GnRH vaccine, GonaCon TM on the growth of juvenile tammar wallabies 1 Bob Forrester,...
-
Upload
gage-hogsed -
Category
Documents
-
view
213 -
download
0
Transcript of The effect of a GnRH vaccine, GonaCon TM on the growth of juvenile tammar wallabies 1 Bob Forrester,...
The effect of a GnRH vaccine, GonaConTM on the growth of juvenile tammar wallabies
1Bob Forrester, 2 Melissa Snape and 2Lyn A. Hinds
1Statistical Consulting Unit, ANU, Canberra, ACT2 Invasive Animals CRC, Canberra, ACT
Outline of talk
• Introduction• Data• Questions of interest• Modelling growth responses• Concluding remarks
Introduction
Vaccination against gonadotrophin releasing hormone disrupts hormonal regulation of reproduction
GonaConTM is effective in eutherian mammals (eg deer, horses, bison)
Tested here on marsupials (Tammar wallabies – relatively small and easy to handle)
Introduction – why vaccinate animals?
Control of overabundant populations
eg Possums in New Zealand, kangaroos in Australia
More humane than poison baits
Avoids emotional responses
More politically acceptable
Data
35 juvenile male Tammar wallabies
3 treatments – sham control, Vac1 (week 0), Vac2 (week 0 and week4)
12 variables measured
Key variables associated with testes size
20 measurement times, up to 131 weeks after treatment
Data
Repeated measurements
Unequally spaced intervals
All animals measured at the same time
Two animals have incomplete records (both animals died)
All analyses carried out using GenStat
Data
Week 0 7 11 15 19 24 28Interval 7 4 4 4 5 4
Week 34 40 48 54 62 71Interval 6 6 8 6 8 9
Week 79 91 101 108 115 123 131Interval 8 12 10 7 7 6 8
Questions of interest
Is the vaccine effective?
Is Vac2 more effective than Vac1?
Does the effectiveness wear off over time?
Are body measurements other than testes affected?
Summary statistics approach
Method effective for measurements such as arm length
Answers overall question of treatment effect
Does not explore treatment interaction with time
Cannot handle the complex responses observed in testes measurements
Fitting problems with short response runs
Arm length – exponential model
Means
Parameter Control Vac1 Vac2 SEDA 130.4 110.2 105.5 3.13B -48.7 -27.6 -23.2 2.91R 0.98891 0.986 0.98538 0.001315
Arm length = A + B*R**Weeks , R = exp(-K)
Model variance structure
Measurements unequally spaced, but all animals measured at the same time
Antedependence or power models
Fixed effects of Treatment, Week, Treatment.Week
Suitable for arm length and also testes measurements
Arm length
Antedependence model order 1 (change of deviance)
Additional animal variance component (9.219, SE=2.331)
Fixed term Wald statistic d.f. Wald/d.f. chi prweek 3389.51 19 178.4 <0.001treatment 174.73 2 87.37 <0.001week.treatment 359.66 38 9.46 <0.001
Arm length
week Control Vac1 Vac2 Av. SED0 85.67 84.4 83.92 1.2047 85.02 85.67 83.59
101 114.78 102.42 100.22108 115.43 102.33 99.68115 116.01 103 100.58123 116.42 103.55 100.83131 117.28 104.17 101.41
Greater insight into treatment effects over time
ln(Testes volume)
Log transform needed to stabilize the variance
Antedependence order 1 model
When variance component included for animal, estimation problems with week 115
Antedependence order 2 model better, but effect on the predicted means is slight (smaller SED)
ln(Testes volume)
week Control Vac1 Vac2 Av. SED0 0.258 0.791 0.566 0.22457 0.957 0.536 0.197
11 1.366 0.314 0.08415 1.485 -0.038 -0.29819 1.907 0.032 -0.19224 2.096 0.013 -0.265
115 2.953 0.789 0.503123 3.025 0.995 0.684131 3.078 1.015 0.777
ln(Testes volume)
Significant interaction due to lack of differences until week 11
Large differences between treated and untreated animals thereafter
No significant differences between Vac 1 and Vac 2 at any stage
One Vac 1 animal effect perhaps wearing off