estimation of the probability of freedom from infectious...

Estimation of the probability of freedom from infectious salmon anemia virus in farmed cohosalmon in Chile using scenario tree modeling

A. Alba, Monti, G., R. Ibarra, A. Tello, M. Lara, K. Montecinos, A. Gallardo, E. Sergeant, A. M. Perez, F. O. Mardones

1 Department of Veterinary Population Medicine, College of Veterinary Medicine, University of Minnesota, MN, USA2 Instituto de Medicina Preventiva Veterinaria, Facultad de Ciencias Veterinarias, Universidad Austral de Chile, Valdivia, Chile.

3 Instituto Tecnológico del Salmón (INTESAL de SalmonChile), Av. Juan Soler Manfredini 41, OF 1802 Puerto Montt, Chile.4 Sernapesca, Chile’s National Fisheries and Aquaculture Service, Victoria 2832, Valparaiso, Chile

5 AusVet Animal Health Services Pty Ltd, Toowoomba, Australia.6 Escuela de Medicina Veterinaria, Facultad de Ecología y Recursos Naturales, Universidad Andres Bello (UNAB), Santiago, Chile.

Int'l Symposium on Food Safety - UNAB OCDE

Importance of aquaculture“About half of seafood production globally comes from aquaculture”


Salmon farming industry


Natural life cycle Artificial salmon farming

Chilean salmon industry

• Started mid-80s

• Exotic species

• 2nd largest producer

• USD 4 bill. yearly

• Half of the country

salmonex.comInt'l Symposium on Food Safety - UNAB OCDE

Infectious diseases impacting farmed salmon


Infectious salmon anemia virus (ISAV)

• RNA virus; family Orthomyxoviridae (“fish flu”)

• Affects Atlantic salmon (Salmo salar), rainbow trout (O. mykiss) are considered as carriers.

• World Animal Health Organization (OIE) notifiabledisease

• Reported in Norway 1984, Canada 1995, Scotland 1998, Faroe Island 2000, USA 2001, Ireland 2002, Chile 2007.


Photos: M. GodoyInt'l Symposium on Food Safety - UNAB OCDE

Rationale

Goal. To quantify the sensitivity of the surveillance and probability of freedom from ISAV in the population of farmed coho salmon in Chile.

Why? To evaluate the need for sustaining surveillance activities for the disease in this species, and contribute to increase the efficiency in the use of resources for disease surveillance in farmed salmon in Chile and other regions.


Stochastic scenario tree model (Martin et al, 2007)

• To estimate Sensitivity of the surveillance system in the last 7y

• To estimate the probability of freedom from ISAV in the cohosub-population

• Assessed every 4 months

• The model builds up information from quarter to quarter, taking the confidence gained from consecutive testing and the probability of introduction of the disease into the cohosubpopulation over time.


Scenario tree based on surveillance program


• Coordinated by Sernapesca and paid by both farmers and Sernapesca, is mandatory

• All farms in fresh water and marine sites

• At least two visits per year

Sampling scheme

• 30 to 60 fish depending if marine or freshwater, respectively.

• Samples prioritizes recently died or moribund fish

• If breeders at the farm, additionally 15

• Tissue pools from 3 fish including kidney, heart and gills.

• RT-qPCR and confirmatory sequencing


Scenario tree based on surveillance program


• Surveillance system component (fresh and salt water)

• Test Se• Design Prevalence• Probability of introduction


Nomenclature Definition Values Source of information and/or reference

Inputs

tp Type of production, i.e., fresh water (FW) or saltwater (SW) coho salmon farms

FW or SW Historical data at Sernapesca

q a quarter comprised between September 2007 and December 2014

1,…, 22 Historical data at Sernapesca

r Number of pools sampled in each coho salmon farm

Integer Historical data at Sernapesca

k Pool size, i.e., number of fish in a pool Integer Historical data at Sernapesca

N Total number of coho salmon farms per quarter

Integer within a range of 22 and 167 Historical data at Sernapesca

P*U Design prevalence (expected prevalence to detect within farms)

15% Gustafson et al., 2007; Corsin et al., 2009; Mardones et al., 2013.

P*Reg Design prevalence (expected prevalence to detect among farms)

2% OIE Aquatic Animal Health Code (2015)

Sepool Pool-level sensitivity estimated by RT-qPCR from an official ring test carried out by Sernapesca

Pert(0.90, 0.95, 0.99) Laboratory records at Sernapesca

Sppool Pool-level specificity 1 Assigned by the modelers

nh Number of farms sampled per quarter Integer within a range of 18 and 89 Historical data at Sernapesca

PriorPInfi,tp Probability of being infected before conducting the first survey

0.5 Assigned by the modelers

IAS Quarterly incidence in Atlantic salmon Percentage within a range of 3.4 and 33 Historical data Sernapesca, 2015

PInfbcq,tp Risk of incursion between consecutive samplings

Pert (IAS,q/1000, (IAS,q/1000 + IAS,q/100 )/2, IAS,q/100)

Equation 6; adapted from Rolland and Winton, 2003.

Outputs

SeReg,q,tp Surveillance sensitivity by q for each tp

Continuous value between 0 and 1 Equations 2 and 3

PFreeReg,q, total Probability of freedom from ISAV by q combining both components

Continuous value between 0 and 1 Equation 8

SeReg,q,totalNn Overall Surveillance sensitivity by q Continuous value between 0 and 1 Equation 9

PFreeReg,q, tpN Probability of freedom from ISAV by q for each tp


AUCtp Probability of freedom for each tp during all the period


AUCtotal Overall Probability of freedom during all the period


𝑆𝑒farm 𝑖 = 1 − 1 − 𝑃𝑈,𝑝𝑡∗ × 𝑆𝑒𝑝𝑜𝑜𝑙

𝑟

SS𝑆𝑒𝑞,𝑝𝑡 = 1 − 1− 𝑃 ∗𝑞,𝑝𝑡 × 𝑆𝑒farm 𝑖 1−𝑃 ∗𝑞,𝑝𝑡 × 𝑆𝑒farm 𝑖

SS𝑆𝑒𝑞,𝑝𝑡 = 1− 1 −𝑛ℎ

𝑁 −𝑁 × 𝑃pt

∗ ∗× 𝑆𝑒farm 𝑖 1 2

2

𝑁×𝑃pt∗ ×𝑆𝑒farm 𝑖

1−𝑛ℎ

𝑁 −𝑁 × 𝑃pt

∗ ∗× 𝑆𝑒farm 𝑖 1 2

2

𝑁×𝑃pt∗ ×𝑆𝑒farm 𝑖

𝑃𝑜𝑠𝑡𝑃𝐹𝑟𝑒𝑒q,pt =1 − 𝑃𝑟𝑖𝑜𝑟𝑃𝐼𝑛𝑓𝑖,𝑝𝑡

1 − 𝑃𝑟𝑖𝑜𝑟𝑃𝐼𝑛𝑓𝑖,𝑝𝑡 × SS𝑆𝑒𝑞,𝑝𝑡

𝑃𝑜𝑠𝑡𝑃𝐼𝑛𝑓q,pt = 1 − 𝑃𝑜𝑠𝑡𝑃𝐹𝑟𝑒𝑒q,pt

𝑃𝑟𝑖𝑜𝑟𝐼𝑛𝑓q,pt = 𝑃𝑜𝑠𝑡𝑃𝐼𝑛𝑓q−1,pt + 𝑃𝐼𝑛𝑓𝑏𝑐q,pt − (𝑃𝑜𝑠𝑡𝑃𝐼𝑛𝑓q−1,pt × 𝑃𝐼𝑛𝑓𝑏𝑐q,pt

𝐴𝑈𝐶𝑡𝑜𝑡𝑎𝑙 = 𝑞=1

𝑞=22

𝑃𝑜𝑠𝑡𝑃𝐹𝑟𝑒𝑒q,Total d(𝑃𝑜𝑠𝑡𝑃𝐹𝑟𝑒𝑒q,Total


Statistical Software

(free)Associated packages

(also free!)

“FFD”

“learnBayes”

“reshape”

“zoo”

Results


Component Saltwater (SW) Freshwater (FW)

Population range undersurveillance (median)

86 – 167 (109) 74-102 (87)

Surveyed range farms 43 – 82 (55) 18-58 (32)

Average coho population per farm 0.73 million (95% CI = 0.4, 1.04) 3.8 million (95% CI = 1.8, 5.8)

• Overall surveillance period sampled a total of 164 FW and 299 SW farms

• 86,382 tested pools

• All negative results RT-qPCR

Assessment of surveillance


Median Se SW = 0.94 (0.61 – 1)Median Se FW = 0.83 (0.61 – 0.99)

Overall surveillance


Median Se ALL = 0.98 (0.81 – 1)

Probability of freedom

• Combining both components PrFreedom by quarter was estimated as a median of 0.99 with a range between 0.96 and 1.

• The PrFreedom over the entire period expressed as AUC was 0.95.


Discussion

• High probability (>0.95) of being free of ISAV

• Kibenge et al 2001 confirmed ISAV from diseased coho salmon farmed with Atlantic salmon, however, Atlantic salmon did not developed any clinical signs.

• Similar in British Columbia


Surveillance Components

• SW surveillance was main contributor to the overall surveillancesensitivity

• Also, SW had a more stable Se compared to FW

• Sampling at SW was more efficient than FW because:• A higher proportion of SW farms were surveyed quarterly (50 vs 38%)

• Number of fish sampled within farm was high enough in SW


Relevance in Disease Management

• Coho salmon has shown a lower suscetibility to most diseases thatcurrently affect farmed salmonids.

• Including sea lice, Piscirickettsia salmonis, infectious pancreaticnecrotic virus and ISAV.

• Epidemiological ‘firebreak’ between more susceptibles species, e.g. Atlantic salmon.


Associated costs to survey for nothing?

• Individual pool sample 22 USD

• Overall study period 86,382 samples tested

• Equivalent to US 1,8 million or roughly 260,000 USD per year

• Resources that can be allocated to other diseases, e.g., Piscirickettsiasalmonis


Acknowledgments

• Maria Jesus Serrano (UNAB)

• Funded by Intesal-SalmonChile and Fondecyt 3140235

• University of Minnesota MnDrive Program


estimation of the probability of freedom from infectious...

Documents