Aspergillosis has gained clinical significance in both the veterinary medicine and public health fields. It affects a wide variety of hosts such as dogs, cats, birds, and humans causing morbidity and mortality worldwide (1–11). Canine sinonasal aspergillosis (SNA) is a common presentation of aspergillosis in dogs (1, 3, 12, 13) and also one of the most frequent causes of chronic sinonasal disease in dogs (6, 14–16). SNA is mainly caused by Aspergillus fumigatus, and less frequently by other species, such as Aspergillus niger, Aspergillus flavus (16, 17), and Penicillium spp. (16). A. fumigatus is a ubiquitous filamentous saprophytic fungus (18), and its conidia (spores) are found in soil, air, water, decaying vegetation, and dust. It is an airborne pathogen (19–24), thus transmission occurs through inhalation of conidia from the environment. SNA generally involves the nasal cavity and/or frontal sinuses (17, 25) of otherwise apparently systemically healthy dogs (17), causing destruction of nasal turbinates (9). Extensive damage to the nasal bones, cribriform plate, and orbit can occur in severe cases (26, 27). Clinical signs include mucoid to mucopurulent nasal discharge, sneezing, depigmentation and/or ulceration of the nares, nasal pain, and epistaxis (1, 26). Ocular discharge occurs as a result of nasolacrimal duct destruction and orbit invasion in advanced cases (1). Neurologic signs may occur if the cribriform plate is affected (26, 27). An important differential diagnosis for SNA in dogs is nasal neoplasia (14). Diagnosis is costly and requires a mixture of invasive and non-invasive diagnostic tests such as advanced imaging, serology, rhinoscopy for identification of fungal plaques, and cytology and/or histopathology (12, 27–31). Treatment is often challenging and involves a combination of tissue debridement, topical, and systemic antifungal therapy (9, 26).

We focused our study on SNA because it is a frequent form of aspergillosis seen in dogs (1, 3, 12, 13), and because the etiology and epidemiology of SNA are distinct from the other types of aspergillosis in dogs and may have implications for human disease. Nasal aspergillosis accounts for 7–34% of dogs with nasal disorders and is the second most common cause of chronic nasal discharge (6, 14, 32, 33). A study in humans shows that worldwide approximately 2.5% (4.8 million people) of adults who have asthma also have allergic bronchopulmonary aspergillosis (34). Invasive aspergillosis in humans generally affects immunocompromised people and is one of the most common fungal infections on organ transplant recipients (35). Studies have shown that A. fumigatus are present and propagate under a variety of environmental and climatic conditions such as water, soil, decaying vegetation, and compost. A. fumigatus present in the air and water for example have been associated with aspergillosis in humans (11, 19–23, 36, 37). A. fumigatus spores can also disperse easily in the air when compared to other types of fungi. Moisture and inappropriate temperature have been associated with the disease in birds (10).

We hypothesized that environmental and climatic factors significantly favor pathogens’ growth, spread, and disease transmission. Studies of geographical distribution with identification of potential high-risk areas and quantification of the influence of environmental and climatic factors on canine SNA are lacking.

Our specific study aims were as follows: (1) to analyze the spatial patterns of canine SNA in Northern California from electronic medical records of the University of California Davis William R. Pritchard Veterinary Medical Teaching Hospital (VMTH) to aid identification of high-risk areas for disease occurrence; (2) to assess the association between environmental/climatic risk factors and disease occurrence in Northern California using a multivariate logistic model. Although SNA does not pose zoonotic potential, dogs may be important sentinels for human exposure as dogs and humans cohabit the same environment. Therefore, this study has potential implications for both animal and public health with the ultimate goal of early detection, prevention, and mitigation of disease.

A total of 250 cases (out of 286 cases retrieved from ZCTA in the entire state of California) and 190,894 dogs (reference dog population) were part of the study catchment area ZCTAs and were considered for the analysis. The mean number of cases per ZCTA in the study area was 0.33 (SD = 0.71; min = 0; max = 5). The overall cumulative incidence (incidence proportion) for the study area was 1.31 SNA cases per 1,000 dogs for the entire studied period. The mean SIR was 1.34 (SD = 4.96; min = 0; max = 54.54). The study area contained 768 ZCTAs (43% of the total number of ZCTAs in California), from which 140 ZCTAs had an SIR > mean.

The final model contained six variables (five main effects and one interaction term), three of which were environmental [i.e., traffic density (OR = 1.7; p = 0.0311), active composting sites (OR = 1.2; p = 0.0299), agriculture (OR = 0.67; p = 0.00345)] and three were climatic [i.e., wind difference (OR = 1.3; p = 0.0621), temperature difference (OR = 0.69; p = 0.0104), interaction between wind and temperature differences (OR = 1.6; p = 0.0134)] (Table 2). The predictive ability of the model based on the AUC value of the ROC curve was 73.1%. The autologistic regression model showed similar results than the final logistic regression model, with similar regression coefficients, deviance residuals and AUC and, for that reason, the simplest logistic regression model was selected.

The spatial distribution of SNA and significant predictors included in the final model are shown in Figure 1. The effect of temperature difference interacting with wind difference at different levels in the occurrence of SNA is shown in Figure 2. This figure shows that when there was high wind difference, the probability of SNA was high, and the protective effect of temperature difference on SNA occurrence was minimized.

The resultant risk map for canine SNA shows that high-risk areas were concentrated close to the San Francisco Bay area, other coastal areas, and also in northern and central CA (Figure 3). The top 10 counties containing ZCTAs that exhibited the highest risk for canine SNA occurrence were Stanislaus, Santa Clara, Sonoma, Santa Cruz, Napa, Contra Costa, Placer, Monterey, Marin, and Sacramento. Residuals from the final logistic model were relatively randomly distributed within the study area (Figure 4).

To the best of authors’ knowledge, this is one of the first studies to provide a canine SNA risk map and to analyze the association between environmental and climatic factors associated with canine SNA in California. Results provide valuable insights on climatic, environmental, and anthropogenic risk factors associated with the disease.

We compared results from both models (logistic and autologistic) by evaluating improvements in the intercept, regression coefficients, deviance residuals, and goodness of the fit (AUC). Since the autologistic model did not significantly change the results over the logistic model, we assumed that spatial autocorrelation was not playing an important role in this particular case. Moreover, residuals map for the final logistic model did not show values too far from 0, and the highest values were fairly dispersed. The majority of the regions with the highest deviance residual values (residual map values >2) are located mainly in the sierra and do not overlap with the highest risk areas (risk map values >0.4), located mainly in the coastal areas. Overall, the deviance residual values are low, indicating a good fit of the model, and are generally dispersed, which, in addition to the results of the autologistic regression model suggest no concerning spatial autocorrelation. Only values deviance residuals >2 may be indicative of lack of fit for those few counties in the sierra suggesting that our model seems to predict better coastal cases than cases in the sierra. Thus, the non-spatial logistic regression model was chosen for simplicity. Traffic density was the main factor associated with SNA (OR = 1.7), suggesting that pollution may predispose dogs to this disease. It is possible that the pathogenicity of A. fumigatus in dogs varies in the presence of different pollutants, or that pollution damages mucosal defense mechanisms. One recent study of the effect of urban air pollution on allergenicity of A. fumigatus spores in the laboratory found increased allergenicity in polluted urban environments during the first 12 h of exposure (46).

Wind and temperature were also significantly associated with SNA occurrence. Both wind and temperature differences between the beginning and the end of the study period (i.e., 1990 and 2014, respectively) were used as proxies of areas where climatic conditions are more drastically changing. Both wind and temperature measurements were higher in 2014 compared to 1990. The protective effect that temperature difference had on SNA occurrence was attenuated by wind (Figure 2). Wind may facilitate dispersion of spores and thus increase the risk of SNA in dogs. Areas with increased wind in the last years are mostly in urban and coastal settings, which tend to experience landscape modifications such as construction, leading to soil disturbance. Urban and coastal areas also appear to have higher humidity and traffic density compared to other areas. Similarly, this study suggests that dogs living in areas experiencing lower temperature differences are those at highest risk of SNA, although again this effect was influenced by wind. In general, these areas have lower elevations and higher temperatures (Figure 1). Active composting sites (for food, green waste, wood, biosolid, and agricultural waste) were significantly associated with an increased risk for SNA. A. fumigatus thrives in composting facilities due to the presence of organic matter (22, 37, 39). The thermotolerant nature of A. fumigatus allows this pathogen to grow in compost and crops at elevated temperatures (optimal growth at 37°C and maximum growth at 52°C) (47, 48).

Agricultural land use by ZCTA reduced the risk of canine SNA. This was somewhat unexpected because the fungus is frequently found in environments rich in nutrients and humidity and where the soil is usually disturbed (as it is usually the case of agricultural areas). However, one possible explanation for the negative association is fungicide application to crops. Unfortunately, detailed information about fungicide use in California was not available. Further studies should be conducted to clarify this effect.

Risk maps provide useful information for clinicians when evaluating animals with chronic nasal discharge that reside in high-risk areas and may help to accelerate accurate diagnosis, and in turn enable more prompt treatment and better prognosis. Furthermore, because A. fumigatus causes disease in both dogs and humans, and because dogs and humans share the same environments, our findings may also have relevance for identification of high-risk areas for human aspergillosis. Human invasive aspergillosis mostly affects immunocompromised patients and can be life-threatening (4, 38, 49, 50). Because A. fumigatus spores can be found in air, dust, and water in hospital environments (11, 20, 36, 38, 41), hospitals located in high-risk areas might focus on improved ventilation systems to reduce risk of infection.

One of the limitations of this study was the availability of data from a single hospital, which can introduce selection bias. Because canine SNA requires specialized diagnostic equipment and treatments not commonly available in general veterinary practice, a diagnosis of canine SNA is often made at the VMTH or cases are referred in for treatment. In order to further reduce selection bias, the 145-mile catchment area around the VMTH was empirically delineated to represent areas of hospital caseload influence. This radius was assumed to be a reasonable distance for a pet owner to drive to seek treatment at the VMTH. Thirdly, the dog reference hospital population from the 25-year study period was taken into account for the SIR calculation; therefore, we adjusted the expected number of SNA cases observed by ZCTA with the population of dogs visiting VMTH from that ZCTA. Therefore, the use of SIR provides a better estimate of the SNA situation per region than the use of the number of cases per se. Lack of access to patient travel history was also a limitation of this study since canine SNA is a chronic disease, and some dogs may have become infected at a different residence prior to diagnosis. Similarly, a long study period was used because of the relatively low number of cases/year, and the relatively stable SIR over that 25-year period (mean: 1.34; SD: ±4.96). The variability of 4.96 is not very high, but it is important to consider. We tried to apply Bayesian zero-inflated Binomial and zero-inflated Poisson models that allow for overdispersion in INLA to model these data, but models did not converge. Authors acknowledge the limitation of merging cases over time and unfortunately it was not possible to include variables per year to minimize ecological fallacy due to the small sample size. However, changes over time were verified and variables such as difference in temperature, in wind, and in precipitation were included and used to account for changes over time and hopefully mitigate the impact of time on the study results.

Future studies that include data from more veterinary hospitals over a wider geographic range may provide more detail in regards to the spatiotemporal patterns of this disease and the varying impact that environmental and climatic conditions may have had in different regions over time.

Despite the limitations, this study provides preliminary insights into the spatial distribution and risk factors contributing to SNA occurrence in dogs in Northern California. These results may be useful to increase awareness and guide diagnosis and risk mitigation strategies in high-risk areas, ultimately opening new doors for further investigation of A. fumigatus infections not only in dogs, but also in humans.

All authors contributed to the design of the study. PV, JS, and MM contributed to the data gathering, cleaning, interpretation, and validation. MM and BM-L wrote the R codes, conducted the data analyses, and drafted the manuscript. All the authors contributed to the critical discussion of results and reviewed and edited the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.