Canine serology as adjunct to human Lyme disease surveillance.

To better define areas of human Lyme disease risk, we compared US surveillance data with published data on the seroprevalence of Borrelia burgdorferi antibodies among domestic dogs. Canine seroprevalence >5% was a sensitive but nonspecific marker of human risk, whereas seroprevalence <1% was associated with minimal risk for human infection.


Paul Mead, Rohan Goel, and Kiersten Kugeler
To better defi ne areas of human Lyme disease risk, we compared US surveillance data with published data on the seroprevalence of Borrelia burgdorferi antibodies among domestic dogs. Canine seroprevalence >5% was a sensitive but nonspecifi c marker of human risk, whereas seroprevalence <1% was associated with minimal risk for human infection.
L yme disease is caused by Borrelia burgdorferi and transmitted in North America by Ixodes spp. ticks. Routine surveillance for human illness indicates that risk for infection within the United States is highly localized. Residents of 10 states accounted for >93% of the ≈248,000 cases reported to the Centers for Disease Control and Prevention (CDC) during 1992-2006 (1). Annual countylevel incidence ranged from 0 to >1,000 cases per 100,000 population (1).
Accurate information about risk is necessary for targeting and motivating Lyme disease prevention efforts (2). In addition, health care providers require knowledge of local disease risk to properly interpret clinical and laboratory fi ndings (3,4). Although risk often can be inferred from surveillance data, reporting practices are subject to bias. Independent measures of disease risk are therefore valuable for validating surveillance fi ndings.
Like humans, domestic dogs are susceptible to opportunistic infection with B. burgdorferi. These infections are often subclinical and pose no risk for direct transmission to humans. Nevertheless, they elicit a robust antibody response. Given the greater proclivity of dogs for tick exposure, canine seroprevalence has been proposed as a sensitive and independent measure of human Lyme disease risk (5)(6)(7). We compared US national surveillance data on Lyme disease with recently published data on B. burgdorferi antibody seroprevalence in dogs (8) to determine the degree of concordance between these 2 measures of Lyme disease risk and to assess the potential for canine seroprevalence to predict areas of Lyme disease emergence among humans.

The Study
State and territorial health departments report Lyme disease cases to CDC as part of the National Notifi able Diseases Surveillance System (1). Data on canine seroprevalence of B. burgdorferi antibodies were obtained from a 2009 publication by Bowman et al. that reported results for 982,336 dogs tested throughout the United States by using a commercial C6-based assay during 2001-2006 (8). We obtained state-specifi c seroprevalence from Table  1 of this publication and county-specifi c seroprevalence as categorical values (0%, 0.1%-0.5%, 0.51%-1%, 1.1%-5%, >5.1%) from Figure 2 of this publication after digital enlargement. We excluded counties too small for the value to be determined reliably. We Detailed canine seroprevalence data were available for 46 US states. In linear regression analysis, state canine seroprevalence and human Lyme disease incidence were positively correlated ( Figure 1; r 2 0.75, p<0.001). On the basis of this relationship, human Lyme disease incidence was effectively zero when the canine seroprevalence was <1.3%. States generally fell into 2 distinct categories according to canine seroprevalence ( Figure 1). Median Lyme disease incidence was uniformly low (median 0.3 cases/100,000 population) and not correlated with canine seroprevalence (r 2 0.0, p>0.4) among 32 states with canine seroprevalence <5%. Among 14 states with canine seroprevalence >5%, median annual human Lyme disease incidence was ≈100- fold higher (24.1 cases/100,000 population) and positively correlated with canine seroprevalence (r 2 0.33, p = 0.03). Categorical canine serologic data were available for 866 (28%) of 3,141 counties in the 46 states (8). Median population in 2004 was 85,699 for counties for which data were available, compared with 25,505 for all counties in the 46 states. As in the state-level analysis, human incidence and canine seroprevalence were positively associated at the county level. Median annual reported Lyme disease incidence for humans was 0.2 per 100,000 population in counties with canine seroprevalence <1%, 1.4 in counties with canine seroprevalence 1.1%-5%, and 25.9 in counties with canine seroprevalence >5% (p<0.001; Figure 2). Five (1%) of 520 counties with canine seroprevalence <1% had rates of human illness above the overall county mean of 4.7 cases per 100,000 population annually, compared with 171 (85%) of 201 counties with canine seroprevalence >5%.

Conclusions
Our results confi rm an overall correlation between canine seroprevalence and reported human incidence of Lyme disease as measured through national surveillance. Canine seroprevalence <1% is associated with extremely low rates of human illness in both state-and county-level analyses. Because human cases are reported according to county of residence rather than county of exposure, infections acquired during travel will occasionally be reported from areas without local transmission. Similarly, low levels of canine seropositivity are expected on the basis of the specifi city of assay (up to 2% false positivity [9]), data from fi eld surveys (7,10), and relocation of dogs from areas of high endemicity (8). Low levels of canine seroprevalence or human incidence should not be misinterpreted as confi rmation of local transmission of B. burgdorferi. Conversely, the overall agreement between human and canine data support the conclusion that risk for B. burgdorferi infection is generally low to nonexistent outside the highly Lyme disease-endemic areas of the Northeast, mid-Atlantic, and upper Midwest.
At the other end of the spectrum, canine seroprevalence >5% was invariably associated with above average Lyme disease incidence in state-level analyses. In county-level analyses, the situation was more nuanced. Although 85% of counties with canine seroprevalence >5% also had above average Lyme disease incidence, 15% did not. In more than half of these counties, incidence increased to above average rates in the following 3 years, suggesting some predictive potential for high canine seroprevalence, especially in counties geographically clustered with other high seroprevalence counties. In other counties, however, high seroprevalence appears to be an anomaly resulting from small sample sizes and local demographics. For example, Routt County, Colorado, is a small rural county in a state where locally acquired Lyme disease has never been documented. Although canine seroprevalence for the county was >5%, a survey of all county veterinarians indicated that 11 of 12 seropositive dogs had lived in or traveled to  known Lyme disease-endemic areas (CDC, unpub. data). Selective testing of dogs with exposure histories may yield misleading results with respect to local endemicity.
Our fi ndings suggest that canine seroprevalence >5% can be a sensitive but nonspecifi c marker of increased risk for human Lyme disease. Because dogs do not transmit infection directly to humans (or humans to dogs), this association refl ects similar susceptibilities to tick-borne infection. In some circumstances, high canine seroprevalence appears to anticipate increasing rates of human infection at the county level. Conversely, canine seroprevalence <1% is associated with little to no local risk for human infection. Canine seroprevalence is a useful adjunct to human surveillance for Lyme disease.