Heartworm infection in domestic dogs in Canada, 1977–2016: Prevalence, time trend, and efficacy of prophylaxis

Introduction

Dirofilaria immitis, also known as heartworm, is a mosquito-borne parasite that primarily infects domestic and wild canids (1). Heartworm infection can be acute or chronic, and the disease can be fatal (2). Heartworm can also be transmitted to humans by mosquitoes, but human infections are rare (3).

Mosquitoes are the only known intermediate host of D. immitis. The rate of parasite development from microfilariae to the infective third-stage larvae is influenced by temperature. For the parasite to be transmitted from a mosquito to a dog there is a threshold temperature of 14°C that is required to support larval maturation to the infective stage and a linear relationship exists between the rate of development and temperature from 18°C to 34°C (4). The amount of heat required for microfilariae to reach that stage is 130 heartworm development units (HDUs) and is calculated using the daily sum of degrees Celsius above the threshold (4,5). In Canada, the transmission period for heartworm has historically been determined by summing daily HDUs from May 6th (date of earliest blood feeding for Aedes stimulans — an early season vector) until 130 HDUs are reached; this is regarded as the start of transmission season for that year (4). The end of the transmission period was determined by looking at the last day 130 HDUs were accumulated within a 30-day period based on the late season vector Aedes vexans (4). The start of the transmission season was not limited to 30 d because the early vector has a long life-span and can be found in August (4). An assumption with these calculations was that mosquitoes do not overwinter (4). Based on the most southern region of Ontario, for the years 1957 to 1986, the transmission period for Canada was estimated to be from June 1st to October 9th (4). As a result, monthly heartworm preventive medication is suggested to be given from June 1st to November 1st although the treatment on June 1st is likely not needed as preventives are approved with 1-month’s reach back activity (2).

Heartworm infection was first considered to be endemic to Canada in the 1970s and since then annual heartworm testing has been conducted at many veterinary practices (6). Historically, the risk in Canada was considered highest in southern Ontario, with 2 other foci in southern Manitoba and southern Quebec (6). For many years mail-back surveys were sent to veterinary clinics across Canada requesting information on the heartworm infection status of dogs, cats and “other” animals. These annual surveys were conducted from 1977 to 2010 and collected data from 1977 to 1989, 1991, 1996 to 1998, 2000 to 2002, and 2010. The surveys reported an aggregated prevalence of 3.7% for all dogs infected with heartworm and assessed by a veterinarian, with a prevalence range from 2.4% in 1977 to 0.15% in 2010. However, heartworm prevalence among dogs not on preventive medication was 0.7% in 2010 (6,7).

In Canada, heartworm is considered a rare infection compared to some other countries. For example, randomly selected companion dogs in Grenada displayed a prevalence of 25.3% (8).

As indicated, southern Ontario has historically been one of the foci of heartworm infection in Canada; however, the risk of infection across Ontario has been debated as heartworm maturation within mosquitoes requires sustained warm temperatures that are uncommon in more northern regions of the country (9). The last nationwide mail-back survey in Canada was conducted in 2010. Anecdotally, heartworm risk may have changed in Canada since 2010 due to a changing climate, which is expected to increase prevalence and extend endemic areas (10). For example, climate change has been associated with the emergence of the parasite in Italy and parts of the United States (3,11). If climate change is implied in the emergence of heartworm infections in these countries, the same could hold true for Canada. For example, Ontario has seen an increase in the number and extent of heatwaves (summers 2010 to 2012), i.e., ideal weather conditions for the development of heartworm in mosquitoes (12).

The goal of this study, therefore, was to estimate the prevalence of heartworm infection, and to study its geographic and temporal distribution in Canada from 1977 to 2016. The 3 objectives were to i) determine if there has been a temporal trend in heartworm infection prevalence among domestic dogs in Canada; ii) explore the spatial extension of heartworm infection across Canada using choropleth maps; and iii) assess the efficacy of preventive drugs using the estimated Attributable Fraction Exposed (AFe).

Materials and methods

Data on the prevalence of heartworm infection in Canadian companion dogs were obtained from 2 sources. The first data set was retrieved from the Ontario Veterinary College (OVC) archives. It consisted of survey data collected for the years 1996, 1997, 1998, 2000, 2001, 2002, and 2010 (13). Additional, similar data from the years 1977 to 1989 and from 1991 were extracted from surveys conducted earlier (7). The surveys described the number of cases per province (or grouped provinces by location, e.g., Atlantic Provinces), the number of dogs tested, the number of dogs on preventive medication diagnosed with heartworm, and the number of dogs on preventive medication. Surveys were sent to provincial, federal, industrial, institutional veterinarians (e.g., veterinary colleges) as well as mixed and small animal practitioners (7,13). The surveys collected annual data, except for the 2010 survey which collected responses from 2010 and early 2011 (13). The definition of heartworm infection was determined by veterinarians participating in the surveys; most (~85%) used blood tests to determine heartworm infection status (7,13). Data for Quebec were not available for the years 1996 to 1998 and 2000 to 2002. For privacy reasons, only the location of the veterinary clinics was provided. Information on the patient or the owner, including confirmatory diagnostics and clinical outcome for the patient were not available. The surveys included information on the travel history of only case dogs.

The second dataset analyzed was provided by IDEXX Laboratories Canada. The data were collected from 2007 to 2016 and were a combination of D. immitis antigen test results performed in clinics and results from submissions to the IDEXX Laboratory. The years 2007 and 2016 were partial years; 2007 had submissions from March to December and 2016 had submissions from January to July. Three heartworm blood antigen tests were used: the SNAP 3Dx, SNAP 4Dx, and SNAP 4Dx Plus tests. The SNAP 4Dx test was released in 2007 and used concurrently with the 3Dx test for the years 2007 to 2011. The SNAP 4Dx Plus test was released in 2012 and that year all 3 tests were used. From 2013 onward only the 4Dx Plus Test was used. Dog signalment, test date, and postal code of the veterinary clinic were provided. Information on the patient or the owner, including travel history, confirmatory diagnostics, preventative usage history and clinical outcome for the patient were not available.

The survey data were used to estimate the canine heartworm prevalence for Canada and separately by province, when/where possible. The denominator for Canada was all dogs tested that year, which was available for all years included. Provincial estimates of prevalence depended on whether the published survey results that year included a breakdown of the reported positive dogs and number of dogs tested for that province. The laboratory data were used to estimate the prevalence of heartworm infection in Canada, and by individual province. To investigate the presence of a temporal trend in heartworm prevalence, the Cochran-Armitage trend test for proportions was applied (14). The early surveys for 1977 to 1980 only reported data at a national level and were used to extend the national trend test from 1977 to 2010. In contrast, provincial data were only available from 1981 to 2010; provincial trend tests were examined for those years.

The reported sensitivity and specificity of the 4Dx Plus Test are 99.0% and 99.3%, respectively (15). The true prevalence for the study population was estimated using the Rogan and Gladen estimator based on the apparent laboratory prevalence and the sensitivity and specificity of the antigen test (16,17). The predictive value for dogs testing positive was also estimated (17). The prevalence referred to in this paper is the estimated apparent prevalence of heartworm infection and will hereafter be referred to as prevalence. It should be recognized that the apparent prevalence may differ from the true prevalence due to false negatives or false positives.

To visualize national temporal trends (as indicated by the trend test), sample size weighted trend lines based on smoothing splines were added to respective scatterplots for the survey data and the laboratory data separately. Provincial trend lines for Manitoba, Ontario, and Quebec were added to the scatterplots.

To visually display the provincial prevalence estimates for both survey and laboratory data, a boundary map file was constructed that outlined the country and provincial borders. The boundary file for Canada and the province boundaries were retrieved from Statistics Canada for the 2011 census (18). The boundary file projection was converted from GCS North America 1983 to Universal Transverse Mercator 15N for better visualization of the provinces’ landmass. The denominator was not consistent across provinces; thus, Empirical Bayesian smoothing was used to internally standardize the differences in sample size across provinces (19–21). The annual prevalence was smoothed for each province and the smoothed prevalence estimates were aggregated over time using the mean to create a choropleth map for each dataset.

Geographic locations of veterinary clinics with laboratory submissions were extracted as centroids using Canadian postal code area information (22). The clinic locations were overlaid as points on top of the choropleth map of the smoothed prevalence. Point sizes were weighted by respective sample sizes (i.e., the number of reported tests).

The estimated Attributable Fraction Exposed (AFe) was estimated from the survey data to examine the efficacy of prophylaxis in preventing heartworm infection using the equation:

Where: D stands for infection and E represents exposure and the positive and negative symbols explain the status (23). The AFe estimates the proportion of infection among a population that is due to an exposure (the assumption being the relationship is causal) (23). If exposures, i.e., prophylaxis, are negatively associated with infection, the AFe estimates the “lack of exposure” to the protective factor as increasing the risk (23). Therefore, the exposed population were dogs not on heartworm preventive medication and the unexposed population were dogs on preventive medication. The assumption for this study is that all parasites are drug susceptible, and that preventive medications were properly administered. The national annual efficacy was estimated by averaging the AFe for each province and year.

All data analyses were conducted in R and RStudio using a significance level of α = 0.05 (24,25). The choropleth maps were created in ArcGIS (26).

Results

The heartworm prevalence for dogs in Canada in 1977 was estimated to be 2.4% (95% CI: 2.2, 2.6) and for 2010 (the last survey year) estimated to be 0.15% (95% CI: 0.14, 0.17). There was no information on the number of dogs on preventive medication in 1977 but 16% of the veterinarians who had participated in the 1977 survey said they recommended a preventive program for their clients (27). In contrast, the 2010 survey reported that 83% of dogs were on preventive medication (13). The prevalence based on laboratory data in 2015 was estimated to be 0.17% (95% CI: 0.16, 0.20), and for the partial year of 2016 (months January to July) estimated at 0.12% (95% CI: 0.10, 0.14). Using the laboratory data, the aggregated prevalence for all dogs tested from 2007 to 2016 was highest in Manitoba at 0.35% (95% CI: 0.31, 0.40). The aggregated prevalence for all dogs tested from 2007 to 2016 for Ontario was estimated to be 0.12% (95% CI: 0.11, 0.13), and the aggregated prevalence for all dogs tested from 2007 to 2016 for Quebec was estimated to be 0.27% (95% CI: 0.24, 0.31).

The surveys consistently had a larger sample size (Tables 1 and 2) than the laboratory register; however, the laboratory annual sample size steadily increased from 8082 samples in 2008 to 181 205 samples in 2015. Notably, most (75% to 90%) data were collected from dogs residing in the province of Ontario (Tables 1 and 2).

There were differences in information provided between survey and laboratory data. Veterinarians from the territories did not mail back surveys and only 1 veterinarian from the Yukon submitted blood samples to the diagnostic laboratory. The surveys reported national heartworm information starting in 1977, but it was not until 1981 that the surveys displayed information at a provincial level.

Figure 1 is a proportional symbol or “bubble” plot showing the annual prevalence of heartworm across Canada, where the bubble size represents the sample size. Additional smoothing lines visualize time trends for selected provinces: Ontario, Manitoba, and Quebec. These 3 provinces were selected because they had the largest sample sizes and had previously been suggested as provinces with an elevated risk of heartworm infection in companion dogs (6). The Cochran-Armitage trend test suggested a temporal trend at both the national level and for some provinces based on the survey data (P < 0.05) (Table 1). Based on the laboratory data, the Cochran-Armitage trend test suggested a temporal trend from 2007 to 2016 for Manitoba and Quebec (P < 0.05); with respect to the trend line for these provinces, prevalence increased (Figure 1 — right). The trend test for Ontario suggested a temporal trend from 2007 to 2016 (P < 0.05); the trend line showed a decrease in prevalence (Figure 1 — right).

The true prevalence for the laboratory data year 2015 was estimated to be 0.165% based on the calculation using apparent prevalence and sensitivity and specificity of the antigen test, which was very close to the apparent prevalence of 0.17% (17). The comparison of true prevalence to apparent prevalence was used with the aggregate prevalence for the years 2007 to 2016 for all dogs tested. The apparent prevalence for Manitoba was underestimated at 0.35% compared to the true prevalence of 0.37%. Ontario had similar apparent and true prevalences of 0.12% and 0.11%, respectively. True and apparent prevalences were the same for Quebec at 0.27%. The positive predictive value was estimated at 96.1% for all dogs tested for the laboratory diagnostic year of 2015, indicating that only 3.9% of the positive tests that year were false positives.

The choropleth map for survey data indicated that Alberta and Quebec had the highest aggregate smoothed prevalence from 1981 to 2010 for all dogs (Figure 2). The choropleth map based on the laboratory data illustrated that Manitoba and Quebec had the highest aggregate smoothed prevalence from 2007 to 2016 for all dogs (Figure 3). Figure 3 also shows the spatial distribution of submitted samples from across Canada; the southern parts of provinces, especially southern Ontario, were the origin of most samples. A spatial comparison of the 2 maps shows an increased prevalence in Manitoba and a decreased prevalence in Ontario (Figures 2 and 3).

Descriptive statistics for the apparent efficacy of preventive drugs are presented in Table 3. The AFe is interpreted here as the total amount of heartworm infections prevented by chemoprophylaxis aggregated for all years and was estimated to be 93.08% (95% CI: 92.85, 93.31). The AFe fluctuated between years, but the estimate for the year 1996 (AFe = 62%) is far below that of the other years, which range between AFe = 93% and AFe = 98%. The maximum AFe was attained in 2010 (Table 3).

Discussion

The prevalence of heartworm in Canada varied over time. However, results from surveys between 1977 and 2010 displayed a consistent decrease among all provinces. In contrast, the prevalence estimated from laboratory diagnostics conducted between 2007 and 2016 increased for Manitoba and Quebec (Figure 1 — right). It should be noted that the sample size has increased over time for both data sources, which may limit the potential for selection bias of clinics whose veterinarians consistently report cases. The increase in sample size over the years could also mean that more veterinary clinics are suggesting routine heartworm testing to their clients, or that there are more owners concerned about heartworm, leading to an increase in testing. The antigen test is also used for detection of antibody to Borrelia burgdorferi and other tick-borne infections. Thus, the increase in testing may also be due to increased awareness/testing for tick-borne infections. Testing occurred year-round; however, there was no indication of the reason behind the test. Most tests occurred from April to June; suggestive for routine heartworm testing, but the proportion of tests outside this period increased over 2007 to 2016.

There were major differences between the survey and laboratory datasets. The annual surveys were based on veterinarian-reported cases of heartworm in which the veterinarian acted as a “gold standard.” However, there can be differences of opinion or method among veterinarians and without a diagnostic test it is not possible to estimate the true prevalence. The prevalence reported from the survey data from 1977 to 2010 could have been seriously over- or underestimated. An antigen test was used for laboratory data, and although the test was updated from 2007 to 2016, the technique remained consistent. Despite the difference in methodology, the comparison of results was similar nationally and for Ontario. However, the possible variance in reporting measures used in the survey data makes a fair comparison to the laboratory data across space and time difficult.

The focus of the analysis was on Ontario due to the large sample size for that province, with additional specific study of Quebec and Manitoba, the provinces with second and third highest number of tests. A significant decrease in temporal trends for survey prevalence from 1981 to the early 2000s was noted in both Manitoba and Quebec. However, the prevalence among laboratory tested dogs from 2007 to 2016 depicted a significant increasing temporal trend (Figure 1 — right). Due to the large sample size from Ontario, the prevalence pattern seen for Canada closely follows the observed Ontario prevalence; despite the significant temporal decrease for Ontario laboratory prevalence reports there was not a trend at the national level (Figure 1 — right).

The western provinces do not appear to have experienced changes in heartworm prevalence, except for Manitoba. The western United States that borders with the prairies in Canada have low heartworm prevalence compared to eastern states and this border relationship may contribute to the spatial distribution of heartworm among Canadian provinces (28). There was no information on travel history for laboratory data; however, the 2010 survey for western provinces stated most of the positive dogs never left the province (13). It is possible that recent changes in climate may have affected the mosquito populations of Manitoba, thus impacting the heartworm prevalence among dogs.

The positive trend in prevalence seen in Manitoba and Quebec could be due to a changing climate in these provinces providing more favorable conditions for heartworm development in mosquitoes. The province of Ontario, for example, has experienced longer heatwaves during the summer (12). If other provinces, such as Manitoba or Quebec are experiencing similar heatwaves, that could produce a more supportive environment for the mosquito lifecycle. In addition, these heat waves could produce the heat required for quicker maturation to the infective third-stage larvae. Both factors could affect the transmission of heartworm infection among dogs.

The laboratory study limitations include the small sample sizes for the Atlantic Provinces, Saskatchewan, and the Yukon. A small sample size can affect the prevalence estimates as it is based on only a few dogs. The data were aggregated over time for choropleth mapping, as heartworm is a rare infection and some years had limited sample sizes. The aggregation of data to ensure adequate sample sizes was a limitation because it examines heartworm prevalence on a larger temporal scale than what occurs in a year; the choropleth maps represent the study period prevalence and not an individual year.

The limitations of the macroscopic view of heartworm infection in Canada from this study should be addressed. A detailed map of Ontario that depicts where most cases were diagnosed would provide better insight as to where hot spots of heartworm infection exist. Determining the location of hot spots might provide veterinarians with a communication tool for advising their clients on the necessity of preventive treatments. Additionally, regional climate data should be examined for spatial and temporal increases in Dirofilaria-development-units from 2005 to 2016 (the number of days over 14°C) (5). There could be changes in the spatial distribution or seasonality of development units that affect the transmission of heartworm.

Dog ownership can be expensive. A socio-economic status that permits paying for tests and regular veterinary visits may contribute to selection bias. Diagnosis is possible only among dogs which frequent a veterinarian and are tested, thus the prevalence may be under-reported as not all domestic dogs frequent the veterinarian or undergo testing. The subset of dogs not tested could also not have prophylaxis and be at greater risk of heartworm infection, but their infections are also not being captured. In a 2011 United States survey of more than 50 000 households, 18.7% of dog owners reported not taking their dog to the veterinarian (29).

Compliance is a problem with prophylaxis, as some owners may not follow the instructions for dosing. However, information on prophylaxis compliance among dog owners is incomplete: it is only available for a few dogs and based on the assumption that owners correctly recall their compliance (2). The issue of compliance can lead to a misclassification bias, in which the protected group may include dogs that are not effectively protected. Furthermore, for oral products, gastrointestinal complications can lead to lower drug uptake by the patient and leave a dog vulnerable to infection (2). In Canada, the efficacy of prophylaxis among survey dogs from 1996 to 2010 has remained above 62% (Table 3). The low efficacy seen in 1996 appears to have been an anomaly, as efficacy remained high after that year (Table 3). There were no new prophylaxis drugs released that year and the survey did not mention an influx of imported rescue dogs. Several owners of dogs which tested positive while on prophylaxis had told their veterinarian that they had missed a dose, so it is likely the low efficacy seen was due to poor compliance.

This study shows that although previous reports of heartworm infection in Canada have indicated a decrease in cases from 1977 to 2010, in certain provinces such as Manitoba and Quebec, heartworm prevalence was on the rise from 2007 to 2016 (Figure 1). Ontario, which has previously been a heartworm focus in Canada, displayed a decreasing temporal trend in prevalence from 2007 to 2016. Climate change may play a role in the epidemiological differences seen in heartworm infection across the country, although further investigation is required to separate this effect from the effect of preventive medication among dogs. The importance of chemoprophylaxis in preventing heartworm infection was illustrated, with an average of 93% of infections being prevented among dogs. This highlights the importance of regular veterinary visits and obtaining a preventive option for a dog that maximizes compliance.