One hundred sixty-four Maine clinics were contacted in February 2003 and invited to join the study; 69 of these agreed to participate. Clinics were instructed to record results of all IDEXX 3Dx Lyme disease tests that were conducted as part of a routine health screen, to record town of residence, and to record if a Lyme disease vaccine had ever been administered. Lyme disease vaccines can be highly effective (2); however, since vaccination rates are unevenly distributed, inclusion of vaccinated dogs would bias estimates of disease risk. This protocol was approved by the Maine Bureau of Health Institutional Review Board.

Canine seroprevalence rates were calculated for minor civil divisions, including towns and unorganized townships. Rates were calculated only for divisions that had results of 10 or more tests. The relationships between the canine prevalence rates and human Lyme disease reports to the Bureau of Health (217 division-matched reports) and tick submissions to the Vector-borne Disease Laboratory (12,482 division-matched submissions) for the 2 years before this study, 2001–2002, were tested with Spearman rank correlation. Canine C₆ antibodies persisted in experimentally infected, untreated dogs for ≥65 weeks, with no endpoint described (9); exposure status of the dogs in the present study could not be determined. Using data from 2 years allowed us to include sufficient numbers of human reports for meaningful statistic testing without sacrificing the ability to look at a "snapshot in time" of the Lyme disease spread.

Figure 1

Figure 1. Canine Lyme disease seroprevalence rates based on the IDEXX 3Dx test for minor civil divisions with ≥10 tests, Maine, 2003.

Figure 2

Figure 2. Regional canine Lyme disease seroprevalence rates calculated from minor civil division pools created within 15-minute quadrangles, Maine, 2003.

Two maps were created. The first map (Figure 1) showed prevalence rates of minor civil divisions with ≥10 tests. The second map (Figure 2) showed pooled data from all divisions, including those with small sample sizes. For this map, an overlay of the state with 15-minute quadrangles was used. Each division from which data were collected was assigned to the quadrangle that contained the largest portion of its area. Seroprevalence rates for quadrangles were calculated by combining test results from all divisions within a quadrangle to find the average rate. Divisions were then assigned the average seroprevalence rate of their quadrangle for mapping. Quadrangles having a pooled total of <10 tests were not included in this map.

Test results from 9,511 dogs that had not been vaccinated for Lyme disease were submitted from 343 minor civil divisions. Tests were performed from March to July 2003. The overall seroprevalence rate was 8%. One hundred and eighty-three divisions met the criterion of a minimum sample size of 10 for calculating prevalence rates. At the division level, seroprevalence rates significantly correlated with the number of ticks submitted to the Maine Medical Center Research Institute's Vector-borne Disease Laboratory from 2001 to 2002 (r = 0.41, p<0.001), and human Lyme disease reports to the Bureau of Health (r = 0.15, p<0.05) from 2001 to 2002.

Regional seroprevalence rates were calculated for 65 quadrangles representing 297 minor civil divisions. Seroprevalence rates ranged from 0% to 47%. Rates were highest along southern coastal Maine (≤47%), with regional rates of 11% as far east as Columbia and along the mid-New Hampshire border as far north as Upton. Forty-four divisions with ≥10 tests had prevalence rates of 0%; 12 of these had ≥30 tests and 3 had ≥60.

This study demonstrates how canine serosurveys using the IDEXX 3Dx test can serve as an active surveillance system for potential human Lyme disease risk. This method overcomes the limitations of human Lyme disease reporting systems by relying on routine screening of populations of healthy dogs to calculate true seroprevalence rates. In this study, a large volume of data from across the state was generated for the most extensive and detailed measure of regional Lyme disease risk in Maine to date. In contrast, passive human Lyme disease surveillance during the previous 2 years yields cases from <90 towns, approximately two thirds of which had only 1 or 2 cases.

Canine seroprevalence rates were congruent with I. scapularis submissions and human Lyme disease reports during a 2-year period when dogs could have been infected, reinforcing the effectiveness of this method for predicting geographic human risk. One previous study has calculated canine seroprevalence rates in Maine (6), but a different assay technique was used (4), which limited our ability to compare those rates to those of the current study. In spite of substantial agreement between canine seroprevalence and rates of tick submissions, mapping of canine seroprevalence data shows high-risk foci in inland areas that were not previously identified by 14 years of tick submissions to the Vector-borne Disease Laboratory or from human Lyme disease reporting to the Bureau of Health; this suggests that canine serosurveys may identify new areas of disease transmission. These are areas of low human population density, and repeat surveys may demonstrate the value of canine serosurveillance in detecting disease spread where human populations are low.

Mapping of pooled data on a regional scale allows geographic patterns of disease to be viewed. Most notably, our data show a concentration of infected dogs in southern and coastal areas. Patterns of infection are suggested in inland areas as well. The significance of these patterns with respect to environmental variables favoring disease transmission is unknown but could be clarified by comparing prevalence rates with patterns of land use, deer herd density, habitat, and other ecologic attributes.

The widespread acceptance of the IDEXX 3Dx test facilitates the use of canine serosurveys for public health. In many Maine veterinary offices, virtually every dog tested for heartworm in the spring is tested for B. burgdorferi antibody; however, well below 100% of canine patients are vaccinated against Lyme disease. Test results can be collected opportunistically from collaborating veterinarians with minimal effort. Previous serosurveys have involved much more intensive effort because of the need for veterinarians to collect extra blood samples. The ease of data collection based on this manner of testing enhances real-time as well as long-term monitoring of disease. Furthermore, the large volumes of test results generated from routine B. burgdorferi screening, and the ability to collect information on dog residence, make large-scale studies of disease geography possible. That we did not exclude in our analyses dogs that have traveled suggests that caution should be used when considering the importance of low prevalence rates or prevalence rates calculated from low sample sizes. However, our finding of dozens of towns with 0% prevalence suggests that the effect of dogs that have traveled on calculated seroprevalence rates is small.