Materials and Methods

Study design—This study was designed as a cross-sectional study in which serum samples collected by veterinarians from pet and shelter dogs with ILI (determined on the basis of cough, sneezing, nasal discharge, or pneumonia) from 2005 to June 2009 were analyzed. Canine serum samples were voluntarily submitted to the University of Florida College of Veterinary Medicine or the Cornell University Animal Health Diagnostic Center for serologic detection of antibodies against H3N8 CIV via an HI assay. Some serum samples were submitted indirectly to the Cornell University Animal Health Diagnostic Center by personnel at other national diagnostic reference laboratories. All serum samples of dogs were tested prior to availability of an inactivated whole virus H3N8 CIV vaccine in June 2009. The frequencies of intrinsic and exposure factors for seropositivity to H3N8 CIV were compared between seropositive and seronegative dogs. The University of Florida Institutional Animal Care and Use Committee approved diagnostic testing conducted at that university for this study; committee approval was not required for the Cornell University Animal Health Diagnostic Center because the center is a national reference diagnostic laboratory and testing was performed for diagnostic purposes.

Laboratory procedures—The HI assay was performed as described.¹ Briefly, serum samples were pretreated with receptor-destroying enzyme^a or periodate, then incubated at 56°C for 30 minutes to inactivate nonspecific inhibitors of hemagglutination. Serum samples were serially diluted 2-fold in PBS solution and incubated with 4 hemagglutinating units of H3N8 CIV (A/canine/Florida/43/2004) antigen (diluted in 25 μL of PBS solution) in 96-well V-bottom microtiter plates for 30 minutes at room temperature (approx 23°C). An equal volume of turkey RBCs in PBS solution (0.5% [vol/vol]) was added to each well and incubated for 30 minutes at room temperature. The endpoint antibody titer was defined as the reciprocal of the highest dilution of serum that completely inhibited hemagglutination. The HI assay cutoff titer for H3N8 CIV seropositivity was 32, as previously determined.¹ Serum samples from specific pathogen–free dogs housed in a barrier research facility were used as negative control samples, and serum samples from dogs with confirmed H3N8 CIV infection were used as positive control samples.^1,2 Control serum samples had previously been validated via a serum microneutralization assay.¹ The HI assay is highly sensitive (99.6%) and specific (94.6%) for serologic diagnosis of H3N8 CIV infection when serum samples are collected from dogs at the proper times.¹⁵

Study inclusion and exclusion criteria—Dogs with ILI and dogs involved in outbreaks of ILI were considered for inclusion in the study. Inclusion of dogs was determined on the basis of a serologic diagnosis of H3N8 CIV infection via the HI assay. Seropositive and seronegative dogs with paired acute and convalescent serum samples were included in the study because infection status was confirmed via detection of presence or absence of seroconversion (a ≥ 4-fold increase in anti-H3N8 CIV antibody titer between acute and convalescent serum samples). All seropositive dogs with 1 unpaired serum sample were also included in the study. However, seronegative dogs with 1 unpaired serum sample were further evaluated to confirm that seronegative status. Antibodies against the H3 protein of CIV detected via the HI assay are produced approximately 7 days after infection of dogs.^1,5–7 Therefore, the date of onset of illness of a dog and the date of serum sample collection were evaluated to determine whether an unpaired serum sample for a seronegative dog had been collected during the acute (< 7 days after infection) or convalescent (≥ 7 days after infection) phase of disease. To ensure seronegative status of dogs from which such samples had been obtained (ie, to avoid misclassification of dogs that may have seroconverted after serum sample collection), seronegative dogs with 1 unpaired serum sample collected < 10 days after the date of onset of illness were excluded from the study. On the basis of these criteria, 1,268 pet and shelter dogs in 42 states evaluated because of ILI from 2005 through June 2009 were included in the study.

Data collection—For each dog, the following data were requested from veterinarians submitting serum samples for serologic diagnosis of H3N8 CIV infection: clinical signs, date of onset of illness, date of serum sample collection, age (classified as < 1 or ≥ 1 year), breed (classified as pure vs mixed), sex (male vs female), potential virus exposure setting (shelter facility, boarding kennel, or other exposure setting [ie, setting in which dog was housed at the time of onset of illness or potential exposure to CIV]), and geographic region. The US regions were defined as follows: northeast (Connecticut, Delaware, Massachusetts, Maryland, Maine, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont, and West Virginia), southeast (Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia), Midwest (Iowa, Illinois, Indiana, Kansas, Michigan, Minnesota, Missouri, Nebraska, Ohio, North Dakota, South Dakota, and Wisconsin), southwest (Arizona, New Mexico, Oklahoma, and Texas), and west (Alaska, California, Colorado, Hawaii, Idaho, Montana, Nevada, Oregon, Utah, Washington, and Wyoming).

To determine whether certain breeds were associated with H3N8 CIV seropositivity, dogs were classified into 1 of 10 canine genomic categories (ancient-spitz, herding, mastiff-like, retriever, scent hound, sight hound, small terrier, spaniel, toy, or working dog).¹⁶ The genomic categories of purebred dogs and mixed-breed dogs reported to be of 1 predominant breed were coded for analysis. When a breed was not included in any of the genomic reference categories, a dog was classified in accordance with the breed's predominant ancestry (as reported by the American Kennel Club).¹⁷

Potential virus exposure settings of dogs were reported by veterinarians at the time of serum sample submission. These exposure settings were classified into the following 3 categories: shelter facility, boarding kennel, or other exposure setting. Dogs in shelter facilities included dogs housed in animal shelter or rescue group facilities. Dogs in boarding kennels included pet dogs recently housed in commercial boarding kennels, veterinary hospital boarding kennels, training or rehabilitation kennels, or dog day care facilities. Dogs in other exposure settings included pet dogs exposed to CIV in settings other than those characterized as high-density or high-turnover housing. For example, dogs that had visited dog parks or grooming facilities or were purchased from pet stores, exposed to another dog that may have had ILI, or evaluated at a veterinary hospital because of ILI (unknown potential exposure setting) were included in the other exposure settings group.

Statistical analysis—Seroprevalence for H3N8 CIV was calculated as the number of dogs that were seropositive for H3N8 CIV divided by the total number of dogs that were tested. Overall and annual prevalence estimates were calculated for all dogs in the study and for dogs from Colorado, Florida, and New York. The 95% CI was calculated for each seroprevalence estimate. Annual seroprevalence estimates were evaluated to detect trends from 2005 to 2007 and from 2007 to 2009 via the χ² test for trend. Additionally, the proportions of dogs classified as seropositive for H3N8 CIV in Colorado, Florida, and New York were compared for each year from 2005 to 2008 via a χ² test; these comparisons were not conducted for 2009 because the number of dogs tested during that year was small.

Unconditional logistic regression (with missing categorical data) was used to model the prevalence odds for classification of dogs as seropositive for H3N8 CIV as a function of the evaluated intrinsic and exposure factors.^18,b In the univariable analysis, variables with values of P ≤ 0.20 were considered for inclusion in a multivariable unconditional logistic regression model (model 1). Because of the variation in proportions of H3N8 CIV–seropositive dogs during 2005 to 2009,^5,7 an additional model (model 2) including the same variables that were included in model 1 was evaluated via multivariable conditional logistic regression that controlled for the year in which dogs were tested.¹⁹ To remove confounding effects of the year in which dogs were tested for H3N8 CIV, unconditional logistic regression models with the same variables included in model 1 were examined for 2005, 2006, 2007, and 2008; data for 2009 were not evaluated because the number of dogs tested during that year was small. A forward stepwise approach was used to identify variables associated with H3N8 CIV seropositivity with 2-sided P values-to-enter and P values-to-remove of 0.05 and 0.10, respectively. For each year, following the fitting of the main effects model, the interaction term between geographic region and exposure setting was tested for significance via the Wald test and the likelihood ratio test. Separate analyses were conducted for dogs in different exposure settings (ie, other low-risk setting vs shelter facility and other low-risk setting vs boarding kennel). Values for the final models were considered significant at P < 0.05. The adjusted OR and 95% CI were reported for each model.

Results

Seroprevalence of antibodies against H3N8 CIV and geographic distribution of dogs—The seroprevalence of H3N8 CIV and geographic distribution of the 1,268 dogs in the study were summarized (Table 1). Data regarding geographic location at the level of state were available for all dogs; however, the name of the city or facility in which dogs were housed was not available for any of the dogs in 12 states (California, Colorado, Connecticut, Massachusetts, North Carolina, New Hampshire, New Jersey, New York, Ohio, Pennsylvania, Utah, and Wyoming). Serum samples were analyzed for dogs in 42 states (≥ 296 cities or facilities). For 24 states, serum samples collected from < 10 dogs were tested. The overall seroprevalence of H3N8 CIV in each geographic region was as follows: northeast (265/427 [62%] dogs), west (144/252 [57%] dogs), southeast (192/490 [39%] dogs), southwest (11/43 [26%] dogs), and Midwest (6/56 [11%] dogs). Colorado (n = 101 dogs), Florida (444), and New York (205) had the highest number of dogs in the west, southeast, and northeast geographic regions, respectively.

The overall seroprevalence of H3N8 CIV was 49% (618/1,268 dogs; 95% CI, 46% to 51%). The number of dogs for which serum samples were submitted for analysis decreased from 2005 (n = 462 dogs) to 2009 (40), and serum samples of most (1,111/1,268 [88%]) dogs were tested from 2005 through 2007. As determined by use of a χ² test for trend analysis, the annual seroprevalence of H3N8 CIV in dogs was significantly (P < 0.01) lower in 2005 (201/462 [44%] dogs) than it was in 2006 (223/419 [53%] dogs) or 2007 (143/230 [62%] dogs). Seroprevalence for H3N8 CIV was significantly (P < 0.01) lower in 2008 (45/117 [38%] dogs) and 2009 (6/40 [15%] dogs) than it was in other years. The overall geometric mean HI anti-H3N8 CIV antibody titer for seropositive dogs was 302 (range, 32 to 4,096).

Seroprevalence of antibodies against H3N8 CIV in dogs with ILI in Colorado, Florida, and New York during 2005 to 2009—Most (750/1,268 [59%]) dogs in the study were from Colorado, Florida, or New York. The overall seroprevalence of H3N8 CIV was significantly (P < 0.05) higher in Colorado (89/101 [88%] dogs) than it was in New York (142/205 [69%] dogs) or Florida (187/444 [42%] dogs; Table 2). The overall seroprevalence of H3N8 CIV was significantly (P < 0.05) higher in New York than it was in Florida. When the annual seroprevalence values for Colorado, Florida, and New York were combined, an increase in seroprevalence was detected from 2005 to 2007. In 2005, the seroprevalence of H3N8 CIV was significantly (P < 0.05) higher in Florida (135/221 [61%] dogs) than it was in New York (7/28 [25%] dogs). From 2006 through 2008, the seroprevalence of H3N8 CIV was significantly (P < 0.05) higher in Colorado and New York than it was in Florida.

Factors associated with seropositivity for H3N8 CIV—In the univariable analysis for factors associated with seropositivity for H3N8 CIV, the variables for age, genomic breed category, exposure setting, geographic region, and year that serum samples of dogs were tested had values of P ≤ 0.20 (Table 3). In the multivariable analysis, the final unconditional logistic regression model (model 1) did not include the year that serum samples of dogs were tested because the variable for annual prevalence of seropositivity of dogs did not fit well in the model (because the trend for annual seroprevalence was not linear [the trend increased then decreased]). Comparison of results determined via the unconditional and conditional logistic regression models (Table 4) indicated the odds of H3N8 CIV seropositivity for the variables exposure setting and geographic region were substantially (> 10%) different between these models, which indicated that the prevalence odds for these 2 variables were confounded by the year that serum samples of dogs were tested.

Results of unconditional logistic regression indicated the odds of H3N8 CIV seropositivity for the variables age, exposure setting, and geographic region varied substantially among years. During 2006 and 2007, the prevalence odds of seropositivity were significantly (P < 0.01 for all comparisons) greater for dogs ≥ 1 year old than they were for dogs < 1 year old, after adjusting for exposure setting and geographic region (Table 5). The odds of seropositivity were significantly (P < 0.05 for all comparisons) greater for dogs housed in shelter facilities and boarding kennels in 2005 and 2006 than they were for dogs housed in other settings during those years, after adjusting for age and geographic region. During 2005, the odds of seropositivity were significantly (P < 0.01 for all comparisons) lower for dogs from the southwest, west, and northeast than they were for dogs from the southeast, after adjusting for age and exposure setting. However, in years after 2005, the odds of seropositivity were significantly (P < 0.01 for all comparisons) greater for dogs from the west (except during 2008) and northeast than they were for dogs from the southeast.

For 2006, addition of the interaction term between exposure setting and geographic region to the model indicated that the interactions between boarding kennel and northeast region, shelter facility and northeast region, and shelter facility and west region were significant. The prevalence odds of H3N8 CIV seropositivity were significantly (P = 0.01) greater for dogs housed in boarding kennels in the northeast than they were for dogs housed in other settings considered low risk in the southeast (OR, 33.41; 95% CI, 2.30 to 483.90). The odds of H3N8 CIV seropositivity were significantly (P = 0.01) greater for dogs housed in shelter facilities in the northeast than they were for dogs housed in other low-risk settings in the southeast (OR, 26.13; 95% CI, 2.17 to 314.10). The odds of H3N8 CIV seropositivity were significantly (P = 0.01) greater for dogs housed in shelter facilities in the west than they were for dogs housed in other low-risk settings in the southeast (OR, 16.39; 95% CI, 1.91 to 140.70). A significant likelihood ratio statistic for 2005, 2007, or 2008 was not detected when the interaction term between these variables (exposure setting and geographic region) was added to the model.

Discussion

In the present study, serum samples obtained from pet and shelter dogs with ILI in 42 states were tested from 2005 through June 2009 to detect CIV H3 antibodies via an HI assay. The overall and annual estimated seroprevalences of H3N8 CIV were high for all dogs in the study and for dogs in Colorado, Florida, and New York (states in which approx 60% of the dogs included in the study lived). The variables year, geographic region, and exposure setting were associated with seropositivity of dogs for H3N8 CIV. In addition, a significant interaction between exposure setting and geographic region was detected for H3N8 CIV seropositivity in tested dogs during 2006.

This study had several limitations that should be considered when interpreting the results. Dogs with ILI were included in the study because, in clinical veterinary practice, such dogs are typically the only dogs tested to determine the cause of a respiratory disease. In addition, serum samples were voluntarily obtained from dogs with ILI and submitted for CIV H3 antibody testing by veterinarians; therefore, not all US states were equally represented, the number of serum samples submitted varied among years, pet dogs and shelter dogs were not equally represented, and the proportion of dogs with ILI that were not tested was unknown. Furthermore, dogs were not selected via a randomization procedure; thus, results of this study cannot be extrapolated for all dogs in the United States or for dogs in each state regarding susceptibility to H3N8 CIV infection. Moreover, diagnostic testing of serum samples was not conducted to detect antibodies to canine respiratory pathogens other than H3N8 CIV; therefore, estimates of seroprevalence for such other pathogens were not determined, and factors associated with seropositivity for such pathogens could not be compared with those for H3N8 CIV. Another limitation of this study was that follow-up convalescent serum samples were not submitted for many of the dogs that had negative serologic results for antibodies against H3N8 CIV; seronegative status could not be confirmed for these dogs. Exclusion of such dogs could have decreased the number of seropositive or seronegative dogs included in the study, which would have resulted in under- or overestimation of seroprevalence. In addition, seropositive dogs for which only 1 unpaired serum sample was submitted could have been exposed to H3N8 CIV prior to the current ILI. Thus, because of the cross-sectional nature of this study, it was difficult to determine the temporal relationship between exposure setting and seropositivity for H3N8 CIV.

In this study, the geographic regions with the highest overall seroprevalence of H3N8 CIV in dogs were the northeast, west, and southeast. However, 1 state in each of those regions was overrepresented. This may have been attributable to the fact that the 3 states with highest representation (New York, Colorado, and Florida [in the northeast, west, and southeast regions, respectively]) had diagnostic laboratories that actively solicited veterinarians for submission of canine serum samples for CIV H3 antibody testing. In addition, although serum samples of dogs from 42 states were tested in this study, more than half of these states were represented by < 10 dogs. Exclusion of some CIV-seronegative dogs in this study affected the number of serum samples that were included for some states. The low numbers of serum samples submitted from some states may have been attributable to a lack of interest by veterinarians or dog owners regarding CIV testing; therefore, the numbers of dogs included in the study from some states may not have represented the actual number of dogs with ILI in those states. In the present study, underrepresentation of many states and absence of serum samples obtained from dogs in other states likely resulted in under- or overestimation of overall seroprevalence of antibodies against H3N8 CIV in dogs for some geographic locations.

The overall seroprevalence of H3N8 CIV was 49% (618/1,268 dogs in 42 states had positive results for antibodies against H3N8 CIV). Although the number of dogs included in the study decreased from 2005 through 2009, the seroprevalence of H3N8 CIV increased from 2005 through 2007 and then decreased from 2007 through 2009. The increase in seroprevalence from 2005 through 2007 was attributed to the increases in seroprevalence for dogs from Colorado, Florida, and New York during those years. Because approximately 60% of the dogs in the study were from these states, overall and annual seroprevalence estimates for those states were calculated and compared. In 2005, seroprevalence of H3N8 CIV was higher for dogs in Florida than it was for dogs in New York in this study. This was likely attributable to results of a study¹ published in 2005 indicating detection of antibodies against H3N8 CIV in dogs in Florida in 2004 and 2005, which may have stimulated the interest of Florida veterinarians regarding submission of canine serum samples for CIV testing. In addition, H3N8 CIV infections were not identified in dogs in New York until late 2005. However, during 2006 through 2008, seroprevalence of H3N8 CIV was higher for dogs in Colorado and New York than it was for dogs in Florida. These results of the present study were consistent with reported^5,7 variations in H3N8 CIV activity from 2005 through 2009 and were likely attributable to the frequently occurring outbreaks of H3N8 CIV in boarding kennels, veterinary clinics, and shelters in these regions (particularly in New York City, Philadelphia, and the Denver area).^{4,5,7,10–12}

Differences in H3N8 CIV seroprevalence estimates between results of this study and those of other studies can be attributed to differences in virus transmission (among years and geographic locations), sampling methods, and exposure settings of tested dogs. For example, the overall seroprevalence for H3N8 CIV in dogs in Colorado was higher in the present study (88%) than it was in another study¹² (3.6%) conducted in that state. These differences in findings were attributed to different methods for soliciting submission of serum samples, periods during which samples were collected, and exposure settings of dogs. In the present study, pet and shelter dogs with ILI were tested for CIV from 2005 through June 2009; during that period, most serum samples obtained from dogs in Colorado were submitted from shelters during respiratory disease outbreaks. In the other study,¹² only pet dogs with or without ILI examined at a university teaching hospital from March through December 2009 were tested for CIV. In another study¹⁴ that included pet dogs without ILI participating in a 2-day flyball tournament in Pennsylvania in November 2009, only 3% of tested dogs had antibodies against CIV H3. In comparison, results of another study¹¹ indicated the overall seroprevalence of H3N8 CIV in dogs in a Philadelphia shelter was 42%, and 21 of 74 (28%) dogs tested from October through December 2008 in that study had ILI. In the present study, the overall seroprevalence of H3N8 CIV in dogs (including pet dogs and shelter dogs with ILI) in Pennsylvania was 39%. Such differences in results among studies indicate the high risk of H3N8 CIV exposure for dogs in shelters (particularly shelters located in regions in which the virus is considered to be endemic).

Results of epidemiological analyses of data in the present study indicated that the variables year, geographic region, and exposure setting were associated with seropositivity of dogs for H3N8 CIV. Year was an important predictor of seroprevalence and had an extreme confounding effect on geographic region; the influence of year changed the direction of the association between geographic region and CIV seropositivity of dogs during 2005 versus later years. The prevalence odds for CIV seropositivity associated with exposure setting also varied among years, although the direction of the association did not change.

Results of the present study indicated that during 2005, the odds of CIV seropositivity were significantly lower for dogs in the southwest, west, and northeast than they were for dogs in the southeast (after adjusting for age and exposure setting). However, in years after 2005, the odds of CIV seropositivity were significantly higher for dogs in the west (except in 2008) and northeast than they were for dogs in the southeast. The association between CIV seropositivity and the variable southeast region in 2005 was attributed to the first reported¹ outbreaks of H3N8 CIV in dogs in Florida during 2004 and 2005 and to the fact that dogs in other geographic regions were not tested for CIV as often during that year. Despite findings of another study² that CIV outbreaks occurred in Florida through 2007, results of the present study indicated the prevalence odds of H3N8 CIV seropositivity were lower for dogs in the southeast region during years after 2005 than they were during 2005. This finding was unexpected and may have been attributable to a lack of H3N8 CIV testing for dogs in southeastern states other than Florida (because of illness attributable to other canine respiratory pathogens in that region for which serologic testing was not conducted or because of differences in serologic testing of dogs among veterinarians in such states). The finding of the present study that there was an association between CIV seropositivity of tested dogs and the variables west and northeast regions during 2006 and 2007 and northeast region during 2008 was attributed to outbreaks of H3N8 CIV in boarding kennels, veterinary clinics, and shelters in these regions (particularly in New York City, Philadelphia, and the Denver area).^{4,5,7,10–12} Additionally, the low frequency of CIV seropositivity of dogs in the southwest region during 2005 in the present study may have been attributable to low virus activity in that region^5,7 or the small number of serum samples submitted from the southwest.

Results of the present study indicated that the odds of CIV seropositivity were significantly higher for dogs housed in shelter facilities and boarding kennels in 2005 and 2006 than they were for dogs housed in other settings during those years (after adjusting for age and geographic region). The opportunity for virus transmission between infected and susceptible dogs is very high in communal housing settings that have high population densities or turnover rates.^{1,2,4,5,7,11,12} For example, results of a recent study¹¹ conducted in a Philadelphia animal shelter indicated that the risk for H3N8 CIV exposure doubled every 3 days that a dog stayed in the shelter. Although outbreaks may occur in low-risk environments, dogs in communal housing settings with a high population density are highly susceptible to virus exposure and transmission among animals. The finding of the present study that there was no association between exposure setting and CIV seropositivity in 2007 and 2008 can potentially be attributed to low H3N8 CIV activity in shelter facilities or boarding kennels during those years, an inability to determine a temporal relationship between exposure setting and seropositivity (because of inclusion of CIV-seropositive dogs for which only 1 serum sample had been submitted), or a small sample size of dogs during these years.

During 2006 and 2007, the prevalence odds of CIV seropositivity were significantly higher for dogs ≥ 1 year old than they were for dogs < 1 year old in the present study (after adjusting for exposure setting and geographic region). Because dogs of any age are susceptible to H3N8 CIV infection,^1,6,20 this result was attributed to an inability to obtain exposure histories for some dogs and an inability to determine the temporal relationship between exposure setting and CIV seropositivity in this cross-sectional study.

Results of the present study indicated a significant combined effect of exposure setting and geographic region on CIV seropositivity of dogs in 2006 (specifically, between the variables boarding kennel and northeast region, shelter facility and northeast region, and shelter facility and west region). This finding could be attributed to CIV outbreaks in shelter facilities and boarding kennels in the northeast and in shelter facilities in Colorado during 2006 and to the high seroprevalence of CIV in dogs in New York and Colorado in the present study. Failure to detect significant interaction effects between these exposure settings and the northeast and west regions in 2007 and 2008 (despite outbreaks of H3N8 CIV in boarding kennels, veterinary clinics, and shelters in these regions^{4,5,7,10–12}) was attributed to a low number of dogs included in the present study during those 2 years. In 2005, the number of dogs in the west, southwest, and southeast regions that were tested for CIV was low; thus, it was not possible to adequately determine the combined effect of exposure setting and region on CIV seropositivity of dogs.

The overall seroprevalence of antibodies against H3N8 CIV in dogs in the present study was high. An increase in annual CIV seroprevalence was detected from 2005 through 2007, and a decrease in seroprevalence was detected from 2007 through 2009. The variables year, geographic region, and exposure setting were associated with CIV seropositivity of dogs, and the odds of seropositivity associated with geographic region and exposure setting varied substantially among years. Results of this study suggested there is a need for continued surveillance of H3N8 CIV in dogs in the United States and that personnel in communal dog-housing facilities should formulate, implement, and evaluate biosecurity protocols to reduce the risk of CIV transmission among dogs.