Seroprevalence and Risk Factors for Canine H3N8 Influenza Virus Exposure in Household Dogs in Colorado
Corresponding author: G.A. Landolt, DVM, MS, PhD, DACVIM, College of Veterinary Medicine and Biomedical Science, Colorado State University, 1678 Campus Delivery, Fort Collins, CO 80523-1678; e-mail: firstname.lastname@example.org.
Background: Since 2004, canine influenza virus (CIV) has spread throughout the United States. While studies suggest that CIV is commonly detected in shelter dogs, little is known about its prevalence in household dogs.
Objectives: To evaluate the seroprevalence of CIV in pet dogs presented for care in a veterinary hospital in Colorado and to investigate risk factors that might predispose these dogs to CIV infection.
Animals: One hundred and forty dogs presenting to the Community Practice service, 110 dogs seen at other clinical services at Colorado State University's Veterinary Teaching Hospital in 2009, and samples from 75 dogs seen before 2004.
Methods: In this prospective study, samples were tested with hemagglutination inhibition assays, using 3 CIV isolates. To identify risk factors for CIV infection, 140 owners completed questionnaires at time of sampling.
Results: CIV seroprevalence was 2.9% (4/140) for dogs seen by the Community Practice service and 4.5% (5/110) for dogs seen by other hospital services (P= .48). All sera collected before 2004 tested negative for CIV. No differences were seen in antibody titers to the 3 CIV isolates tested. Data from the questionnaires indicated an association between CIV seropositivity and canine daycare visits (P < .001).
Conclusion and Clinical Importance: CIV seropositivity in household dogs in Colorado is low, although it has increased since 2004. Antibody titers to the 3 CIV isolates were comparable, suggesting that measurable antigenic drift has not yet occurred. Finally, dogs boarded in kennels or attending daycare might be at an increased risk of CIV infection.
canine influenza virus (H3N8)
- CSU VTH
Colorado State University Veterinary Teaching Hospital
receptor destroying enzyme
Influenza A viruses infect a wide range of species, including humans, horses, pigs, sea mammals, and birds.1 Historically, dogs were not considered to be natural hosts for influenza. However, in 2004, an outbreak of respiratory disease in racing Greyhounds at a track in Florida provided evidence of transmission of an influenza virus closely related to the contemporary equine H3N8 “Florida” lineage.2 Over the course of that same year, outbreaks of canine influenza occurred at 14 Greyhound racetracks in 6 states, and by 2005 the virus had spilled over into the non-Greyhound dog population.2,3 To date, canine influenza virus (CIV) infections have been detected in at least 27 states and the District of Columbia.a Because the virus represents a novel pathogen, all dogs irrespective of age, breed, and sex are potentially susceptible to infection with CIV. The virus is shed in respiratory secretions and transmitted in aerosols created by coughing and sneezing. As a consequence, close contact and closed environments appear to favor virus transmission among susceptible animals.3,4
While it is impossible to determine the total number of CIV infections that have occurred since first identification of the virus, a recent 1-year study performed in our laboratory demonstrated that CIV was found in 11 of 16 Colorado humane shelters that experienced outbreaks of canine respiratory disease (unpublished data). These results suggest that CIV is a frequently detected pathogen in Colorado humane shelters that have a history of respiratory disease in dogs. To date, only a few studies have investigated CIV seroprevalence in nonshelter dog populations,b5,6 and no information is available on the prevalence of CIV infection in the household dog population of Colorado. To better understand the extent of CIV infection in household dogs in a state with a presumed high incidence of CIV infection,a we sought to estimate the rates of CIV seropositivity among dogs from nonshelter environments that were seen at Colorado State University's Veterinary Medical Center (CSU VTH). In addition, as identification of factors that increase the risk of CIV infection would provide a factual foundation on which to base disease prevention and control measures, a second objective of this study was to investigate risk factors for CIV infection in Colorado household dogs.
Materials and Methods
We utilized a prospective convenience sampling method in which eligible households were defined as a domestic unit located in the state of Colorado consisting of the members (or a member) of a family who live together along with a dog (or dogs) of undefined age, breed, vaccination status, or sex. While dogs that qualified under this definition were eligible for inclusion regardless of the reason for presentation, they were excluded from the study if their owner declined to provide consent. In addition, dogs were excluded if the clinician determined that blood collection would have caused undue stress to the animal or if the dog had received CIV vaccination. Based on the results of a serological surveillance study for swine influenza virus infections among unvaccinated pigs in the north-central United States,7 the sample size was calculated using an estimated seroprevalence of 20%. Before initiation, this study was reviewed and approved for conduct by the Colorado State University Institutional Animal Care and Use Committee.
Serum samples were collected from 3 different household dog populations presented to the CSU VTH. Over a collection period of 9 months (March to December 2009), 140 serum samples were obtained from household dogs admitted to the Community Practice service at the CSU VTH. These dogs were generally in good health and were presented for reasons such as wellness examinations, preventive health care, dentistry, and heartworm testing. In addition, serum samples were obtained from 110 dogs that had been submitted by CSU VTH clinicians to the Clinical Pathology laboratory from dogs of various ages, breeds, and clinical histories. The blood samples were submitted as part of the diagnostic investigation of a range of clinical conditions. Lastly, 75 serum samples collected before 2004 from Colorado household dogs of all ages obtained from a serum bank maintained at the Animal Cancer Center at CSU VTHc were used for the estimation of the pre-2004 CIV seroprevalence.
A questionnaire was administered to the 140 dog owners that presented their animals to the Community Practice service at CSU VTH at the time of sample collection. The questionnaire included the dog's signalment (age, breed, sex), vaccination status, and past and current medical histories. In addition, the questionnaire addressed the length of ownership of the dog, the number of other dogs living in the same household, the dog's travel history, and information about social interactions with nonhousehold dogs, such as visits and/or stays at dog parks (including the number of visits), daycare facilities (including frequency of stays), and boarding facilities (including number and length of stays).
The levels of hemagglutination inhibiting (HI) antibodies in serum samples were determined by HI assays that were performed as described previously7,8 and in accordance with procedures recommended by the World Organization for Animal Health (OIE). Briefly, sera were incubated overnight at 37°C with 4 volumes of receptor destroying enzyme (RDE)d prepared from Vibrio cholera. After inactivation of the RDE by incubation of the samples at 56°C for 60 minutes, 2-fold serial dilutions of sera were mixed with 4 hemagglutinin units of CIV in 96-well microtiter plates. The assays were developed by adding 0.5% (v/v) chicken red blood cells, and the HI antibody titers were defined as the reciprocal of the highest dilution causing complete inhibition of agglutination. To account for potential antigenic drift, the following CIV strains were used as antigens: A/Canine/Florida/43/04 (H3N8), A/Canine/CO-1/224986/06 (H3N8), and A/Ca/CO-5/234550/09 (H3N8). Each sample was run in duplicate and each assay included both positive (positive control canine serum) and negative (physiological saline only) controls. Seropositivity was defined at an HI titer >1 : 8.
Data from the questionnaires were collated with dog signalment information and results of serological testing. These data were validated and entered in a computer spreadsheet and data were summarized using contingency tables. “Plus four” confidence intervals were calculated for the proportions of seropositive dogs by adding 2 successes to the seropositivity outcome and 4 to our sample size for computation of a more appropriate 95% confidence interval.9 Associations between seropositivity and potential risk factors such as age (≥2 years [mature dog], <2 years [immature dog]), sex (male, female), history of cough in the past 6 months (yes, no), other dogs living in the same household (yes, no), dog acquired from shelter or rescue (yes, no), dog attended daycare in the past 6 months before sample collection (yes, no), dog was boarded in the past 6 months before sample collection (yes, no), dog greeted other nonhousehold dog in the past 6 months (yes, no), and dog visited dog park in the past 6 months before sample collection (yes, no) were investigated using the Fisher's Exact Test (http://www.langsrud.com/fisher.htm) with a critical α of 0.05.
CIV H3N8 Seroprevalence
A total of 325 serum samples were tested. All dogs included in the study lived in the state of Colorado at the time of sample collection. One-hundred and forty serum samples were collected from dogs seen at the Community Practice service at CSU VTH between March and December 2009 (mean age 4.8 years [range 6 months–15 years]; 73 males, 67 females). One-hundred and ten samples were obtained through the Clinical Pathology laboratory from dogs that were seen by other clinical services at CSU VTH for various reasons between March and December 2009 (mean age 8 years [range 3 months–16 years]; 51 males, 59 females). Lastly, 75 serum samples that had been collected before 2004 were obtained. Of all serum samples collected from dogs in 2009, 3.6% (9/250; 95% CI = 1.1%, 6.1%) were antibody positive, including 2.9% (4/140; 95% CI = 0.01%, 6.1%) of dogs seen by the Community Practice service and 4.5% (5/110; 95% CI = 0.1, 9.0%) of dogs seen by other services at CSU VTH. This difference in seroprevalence between groups was not statistically significant (P= .48). All serum samples collected from dogs before 2004 were seronegative. The antibody titers to A/Ca/Florida/43/04 were similar to those for the 2006 and 2009 isolates. The mean geometric log base 2 HI antibody titer for the positive serum samples was 7.5 (1 : 176), ranging from 5.8 to 8.4 (1 : 58 to 1 : 341). These patterns of inhibition demonstrated the antigenic similarly between the CIV isolates tested and indicate that measurable antigenic drift has not yet occurred.
Risk Factor Analysis
One hundred and forty questionnaires were completed by owners presenting their dogs to the Community Practice service at CSU VTH. The only factor statistically associated with CIV H3N8 seropositivity was attendance at canine daycare within the last 6 months (OR = 53.8; 95% CI = 37.2, 77.8; P < .001) (Table 1). Additionally, although not statistically significant, dogs boarded in kennels within 6 months before serum sample collection appeared to be predisposed to CIV infection (OR = 8.5; 95% CI = 6.5, 10.9; P= .06). No other factors related to the dog's signalment (age, breed, sex), vaccination status, medical and travel histories, or other social interactions with nonhousehold dogs, such as at dog park visits, were associated with CIV seropositivity.
Table 1. Results of analyses of various potential risk factors for association with CIV seropositivity among dogs presented to the Community Practice service at Colorado State University Veterinary Medical Center.
|Age of dog|
| ||≥ 2 years||4||85||5.4||1.1–26.6||0.16|
| ||< 2 years||0||51||Reference|| || |
| ||Male||2||71||Reference|| || |
|History of cough in past 6 months before sample collection|
| ||No||4||118||Reference|| || |
|Other dogs living in the same household|
| ||No||1||54||Reference|| || |
|Dog acquired from shelter or rescue|
| ||No||2||96||Reference|| || |
|Dog attended daycare in the past 6 months before sample collection|
| ||Yes||2||2||53.8||37.2–77.8||< 0.001|
| ||No||2||134||Reference|| || |
|Dog boarded in the past 6 months before sample collection|
| ||No||2||122||Reference|| || |
|Dog has greeted other nonhousehold dog in the past 6 months|
| ||No||0||25||Reference|| || |
|Dog visited dog park in the past 6 months before sample collection|
| ||No||4||110||Reference|| || |
Findings of the present study suggest that CIV H3N8 seroprevalence in household dogs in Colorado is lower compared with CIV seroprevalence in humane shelters,2,10 although CIV seroprevalence has increased since 2004. Despite the fact that the total number of canine influenza infections that have occurred since first identification of CIV is unknown, a recent 1-year study conducted by our laboratory found that 11 of 16 Colorado humane shelters that experienced outbreaks of canine respiratory disease (kennel cough) housed at least 1 CIV real-time PCR positive dog (unpublished data). Similarly, an earlier study found a 97% CIV seroprevalence in shelter dogs in Florida and in dogs housed at veterinary clinics in Florida and New York that had a recent history of a respiratory disease outbreak.2 Taken together, these data suggest that CIV plays an important role in dogs with clinical respiratory disease that are housed in closed environments in Colorado, Florida, and New York.
Given the presumed high prevalence of CIV exposure in Colorado shelter dogs, it was surprising that the seroprevalence found in Colorado pet dogs was so low. As the canine sera demonstrated similar titers to CIVs isolated in 2004, 2006, and 2009, indicating a lack of measurable antigenic drift, the small number of seropositive dogs found in our study cannot be explained by a change in antigenicity of the virus. While other studies conducted in nonshelter dogs have demonstrated similarly low seroprevalences (0.4% and 1.2%),b6 these studies were conducted in geographic areas where CIV prevalence is thought to be low (Ontario, Canada; Iowa).a Considering the frequent detection of CIV in Colorado humane shelters, a potential explanation for the assumed discrepancy in the frequencies of CIV exposure is that influenza virus ecology is different in shelter dogs than in household dogs and spillover of virus from the shelter environment may not pose a common risk for most household pets presented to the CSU VTH.
Another possible reason for the low number of CIV H3N8 positive dogs found in this study is that the current study was a population prevalence study and not a study focusing on animals with clinical signs of respiratory disease. Despite this, the low seroprevalence in this study suggests that CIV is not a common respiratory pathogen encountered by the majority of pet dogs in Colorado. This notion seems to be further supported by the finding that a history of coughing within 6 months before sample collection was not statistically associated with CIV seropositivity. In contrast, our study identified attendance at canine daycare as a risk factor for CIV exposure, supporting the notion that commingling of animals in closed environments favors virus transmission. This result may be used to target preventive strategies (eg, vaccination) at dogs that have the greatest risk of CIV exposure, such as dogs attending daycare or dogs that are being boarded. However, controlled field trials are needed to evaluate the value of such preventative measures in these groups of animals.
We recognize several limitations of this study. Firstly, as the number of seropositive animals was small, the power to define risk factors for CIV exposure is low and, therefore, the data should be weighed carefully. Moreover, the small number of positive serum samples does increase the potential for Type II statistical error (not finding a statistical difference when one truly exists). As a result, it is possible that, although no significance was found between CIV seropositivity and other risk factors analyzed in our questionnaire, there may actually have been associations present. Secondly, a potential bias relating to the sample population also has to be considered. This study involved dogs presented to a primary (Community Practice Service at CSU VTH) and secondary (other veterinary services at CSU VTH) veterinary care facilities. This population is likely biased as there is an expected subset of animals that are rarely or never presented for veterinary care. As the majority of dogs included in this study originated from a limited geographical area, one cannot draw conclusions as to the true CIV H3N8 seroprevalence in pet dogs in the state of Colorado. Lastly, while published data report detection of antibodies in banked dog sera more than 5 years after collectione, the duration of persistence of CIV antibodies in individual animals has not been well defined. It is possible that some of the dogs tested in this study had previously been infected but no longer had detectable antibody titers, which may have resulted in an underestimation of CIV seroprevalence in this study. Nevertheless, while it remains to be determined how long CIV antibodies persist in canine serum, the discrepancy between the frequent detection of CIV in Colorado shelters and the low seroprevalence in household dogs merits further investigation.
aAnimal Health Diagnostic Center at Cornell University, http://diagcenter.vet.cornell.edu/pdf/CIV-Stats.pdf (accessed May 11, 2010)
bLiu S, Kim WI, Johnson J. Characterization and surveillance of canine influenza virus in Iowa. Proceedings of the ASV meeting, Vancouver, July 11–15, 2009, 326 (abstract)
cSera were provided by Dr S. Lana, Animal Cancer Center, Colorado State University
dDenka Seiken Co, Tokyo, Japan
eAnderson TC, Grimes L, Pompey J, et al. Serological evidence for canine influenza virus circulation in racing greyhounds from 1999–2003. J Vet Intern Med; 21:576–577, 2007, 16 (abstract)
This study was supported by a grant from the American College of Veterinary Internal Medicine Foundation. We thank the Professional Veterinary Medicine Students for serum sample collection and for the administration of the questionnaire. Additionally, we thank Dr P.C.Crawford for providing the A/Canine/Florida/43/04 (H3N8) isolate used in this study.