Limits...
Spillover of pH1N1 to swine in Cameroon: an investigation of risk factors.

Larison B, Njabo KY, Chasar A, Fuller T, Harrigan RJ, Smith TB - BMC Vet. Res. (2014)

Bottom Line: To date we have found pH1N1 only in the North and Extreme North regions of Cameroon (regions in Cameroon are administrative units similar to provinces), though half of our sites are in the Central and Western regions.Random forest analyses revealed that the three best predictors of the presence of pH1N1 in swine were contact rates between free-ranging swine and domestic ducks, contact rates between free-ranging swine and wild Columbiformes, and contact rates between humans and ducks.Sites in which swine were allowed to range freely had closer contact with other species than did sites in which swine were kept penned.

View Article: PubMed Central - HTML - PubMed

Affiliation: Center for Tropical Research, Institute of the Environment and Sustainability, University of California, 619 Charles E, Young Drive East, Los Angeles, California 90095, USA. blarison@ucla.edu.

ABSTRACT

Background: The 2009 pH1N1 influenza pandemic resulted in at least 18,500 deaths worldwide. While pH1N1 is now considered to be in a post-pandemic stage in humans it has nevertheless spilled back into swine in at least 20 countries. Understanding the factors that increase the risk of spillover events between swine and humans is essential to predicting and preventing future outbreaks. We assessed risk factors that may have led to spillover of pH1N1 from humans to swine in Cameroon, Central Africa. We sampled swine, domestic poultry and wild birds for influenza A virus at twelve sites in Cameroon from December 2009 while the pandemic was ongoing, to August 2012. At the same time we conducted point-count surveys to assess the abundance of domestic livestock and wild birds and assess interspecific contact rates. Random forest models were used to assess which variables were the best predictors of influenza in swine.

Results: We found swine with either active pH1N1 infections or positive for influenza A at four of our 12 sites. Only one swine tested positive by competitive ELISA in 2011-2012. To date we have found pH1N1 only in the North and Extreme North regions of Cameroon (regions in Cameroon are administrative units similar to provinces), though half of our sites are in the Central and Western regions. Swine husbandry practices differ between the North and Extreme North regions where it is common practice in to let swine roam freely, and the Central and Western regions where swine are typically confined to pens. Random forest analyses revealed that the three best predictors of the presence of pH1N1 in swine were contact rates between free-ranging swine and domestic ducks, contact rates between free-ranging swine and wild Columbiformes, and contact rates between humans and ducks. Sites in which swine were allowed to range freely had closer contact with other species than did sites in which swine were kept penned.

Conclusions: Results suggest that the practice of allowing swine to roam freely is a significant risk factor for spillover of influenza from humans into swine populations.

Show MeSH

Related in: MedlinePlus

Variable importance from random forest analyses. All variables included in the model are shown listed from top to bottom in order of importance. Variables are the pairwise contact rate (as defined in the text) between the two taxa listed. A variable’s importance is determined by the decrease in the predictive accuracy of the model when that variable is permuted. Waders include all water-associated birds excluding Anseriformes and Charadriformes (shorebirds). Other land birds include all terrestrial birds excluding Passeriformes, Columbiformes and Galliformes and Village Weavers.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4016523&req=5

Figure 2: Variable importance from random forest analyses. All variables included in the model are shown listed from top to bottom in order of importance. Variables are the pairwise contact rate (as defined in the text) between the two taxa listed. A variable’s importance is determined by the decrease in the predictive accuracy of the model when that variable is permuted. Waders include all water-associated birds excluding Anseriformes and Charadriformes (shorebirds). Other land birds include all terrestrial birds excluding Passeriformes, Columbiformes and Galliformes and Village Weavers.

Mentions: A salient difference among sites was whether swine were kept penned or allowed to range freely. Swine were free ranging at 5 of the 6 northern sites, but were nearly always penned in the other sites (Table 1, Figure 1). Domestic poultry were allowed to range freely at all sampled sites. Tree classification using the full set of variables resulted in a single split wherein spillover is predicted to occur at sites with a human-duck contact rate of 1.3027 or greater. Random forest classification analyses indicated that the top three predictors of pH1N1 being present in swine were contact rates between ducks and free-ranging swine, contact rates between swine and wild columbiform birds, and contact rates between humans and ducks (Figure 2). Seven of the top 10 predictors involved free-ranging swine (Figure 2, Table 1). The error rate of the full classification model with all variables was 8.33%. In the full model, sites in which we did not detect influenza were always correctly classified, but one of the four sites in which we did detect influenza was consistently misclassified as a non-influenza site. A model using only the top three predictors, however, successfully predicted both positive and negative sites 100% of the time, suggesting that some variables served only to confound classifications when used as predictors. Table 2 shows the effect of allowing swine to range freely on contact with ducks and Columbiformes. In most sites where swine were penned, swine and these two avian groups did not co-occur within the same scan, or even within the same plot at different times. Where swine were free-ranging they often co-occurred in the same plot either during the same scan or in the same plot at another time. Additionally, behavioral notes taken during the scans by observers included many instances of close contact between swine and other species, all but one of which occurred in sites with free-ranging swine. These included three swine eating a dead duck (Malape), 11 cattle egret following swine in a marsh area (Gounougou), swine foraging in a hut (Vounaloum), and swine sleeping in a hut with humans (Malape). The only such observation including penned swine was of 7 chickens feeding under an elevated pen with swine in it (Obala).


Spillover of pH1N1 to swine in Cameroon: an investigation of risk factors.

Larison B, Njabo KY, Chasar A, Fuller T, Harrigan RJ, Smith TB - BMC Vet. Res. (2014)

Variable importance from random forest analyses. All variables included in the model are shown listed from top to bottom in order of importance. Variables are the pairwise contact rate (as defined in the text) between the two taxa listed. A variable’s importance is determined by the decrease in the predictive accuracy of the model when that variable is permuted. Waders include all water-associated birds excluding Anseriformes and Charadriformes (shorebirds). Other land birds include all terrestrial birds excluding Passeriformes, Columbiformes and Galliformes and Village Weavers.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4016523&req=5

Figure 2: Variable importance from random forest analyses. All variables included in the model are shown listed from top to bottom in order of importance. Variables are the pairwise contact rate (as defined in the text) between the two taxa listed. A variable’s importance is determined by the decrease in the predictive accuracy of the model when that variable is permuted. Waders include all water-associated birds excluding Anseriformes and Charadriformes (shorebirds). Other land birds include all terrestrial birds excluding Passeriformes, Columbiformes and Galliformes and Village Weavers.
Mentions: A salient difference among sites was whether swine were kept penned or allowed to range freely. Swine were free ranging at 5 of the 6 northern sites, but were nearly always penned in the other sites (Table 1, Figure 1). Domestic poultry were allowed to range freely at all sampled sites. Tree classification using the full set of variables resulted in a single split wherein spillover is predicted to occur at sites with a human-duck contact rate of 1.3027 or greater. Random forest classification analyses indicated that the top three predictors of pH1N1 being present in swine were contact rates between ducks and free-ranging swine, contact rates between swine and wild columbiform birds, and contact rates between humans and ducks (Figure 2). Seven of the top 10 predictors involved free-ranging swine (Figure 2, Table 1). The error rate of the full classification model with all variables was 8.33%. In the full model, sites in which we did not detect influenza were always correctly classified, but one of the four sites in which we did detect influenza was consistently misclassified as a non-influenza site. A model using only the top three predictors, however, successfully predicted both positive and negative sites 100% of the time, suggesting that some variables served only to confound classifications when used as predictors. Table 2 shows the effect of allowing swine to range freely on contact with ducks and Columbiformes. In most sites where swine were penned, swine and these two avian groups did not co-occur within the same scan, or even within the same plot at different times. Where swine were free-ranging they often co-occurred in the same plot either during the same scan or in the same plot at another time. Additionally, behavioral notes taken during the scans by observers included many instances of close contact between swine and other species, all but one of which occurred in sites with free-ranging swine. These included three swine eating a dead duck (Malape), 11 cattle egret following swine in a marsh area (Gounougou), swine foraging in a hut (Vounaloum), and swine sleeping in a hut with humans (Malape). The only such observation including penned swine was of 7 chickens feeding under an elevated pen with swine in it (Obala).

Bottom Line: To date we have found pH1N1 only in the North and Extreme North regions of Cameroon (regions in Cameroon are administrative units similar to provinces), though half of our sites are in the Central and Western regions.Random forest analyses revealed that the three best predictors of the presence of pH1N1 in swine were contact rates between free-ranging swine and domestic ducks, contact rates between free-ranging swine and wild Columbiformes, and contact rates between humans and ducks.Sites in which swine were allowed to range freely had closer contact with other species than did sites in which swine were kept penned.

View Article: PubMed Central - HTML - PubMed

Affiliation: Center for Tropical Research, Institute of the Environment and Sustainability, University of California, 619 Charles E, Young Drive East, Los Angeles, California 90095, USA. blarison@ucla.edu.

ABSTRACT

Background: The 2009 pH1N1 influenza pandemic resulted in at least 18,500 deaths worldwide. While pH1N1 is now considered to be in a post-pandemic stage in humans it has nevertheless spilled back into swine in at least 20 countries. Understanding the factors that increase the risk of spillover events between swine and humans is essential to predicting and preventing future outbreaks. We assessed risk factors that may have led to spillover of pH1N1 from humans to swine in Cameroon, Central Africa. We sampled swine, domestic poultry and wild birds for influenza A virus at twelve sites in Cameroon from December 2009 while the pandemic was ongoing, to August 2012. At the same time we conducted point-count surveys to assess the abundance of domestic livestock and wild birds and assess interspecific contact rates. Random forest models were used to assess which variables were the best predictors of influenza in swine.

Results: We found swine with either active pH1N1 infections or positive for influenza A at four of our 12 sites. Only one swine tested positive by competitive ELISA in 2011-2012. To date we have found pH1N1 only in the North and Extreme North regions of Cameroon (regions in Cameroon are administrative units similar to provinces), though half of our sites are in the Central and Western regions. Swine husbandry practices differ between the North and Extreme North regions where it is common practice in to let swine roam freely, and the Central and Western regions where swine are typically confined to pens. Random forest analyses revealed that the three best predictors of the presence of pH1N1 in swine were contact rates between free-ranging swine and domestic ducks, contact rates between free-ranging swine and wild Columbiformes, and contact rates between humans and ducks. Sites in which swine were allowed to range freely had closer contact with other species than did sites in which swine were kept penned.

Conclusions: Results suggest that the practice of allowing swine to roam freely is a significant risk factor for spillover of influenza from humans into swine populations.

Show MeSH
Related in: MedlinePlus