A case-control survey study of environmental risk factors for primary hypoadrenocorticism in dogs

1 INTRODUCTION

Primary hypoadrenocorticism is an immune-mediated endocrinopathy in which the adrenal cortex is destroyed, resulting in dangerously low concentrations of glucocorticoids and mineralocorticoids.^{1, 2} Although prevalence in the general dog population is low (ie, 0.06%-0.31%)³ the disease is concentrated in certain breeds.^4-10 Standard Poodles (STPDs) and Portuguese water dogs (POWDs) have among the highest prevalence, with some studies estimating 8% to 10% being affected.^{6, 7} Breed predisposition suggests a hereditary component, and genetic contributions have been studied.^{5, 8, 11-17} Major histocompatibility class (MHC) class II alleles,^{11, 13, 18} cytotoxic T-lymphocyte protein 4 (CTLA4),^{12, 18} and other immune system-related genes^{5, 19, 20} have shown modest associations with the disease (henceforth referred to as hypoadrenocorticism) in dogs. Heritability estimates of 0.49 for POWD⁶ and 0.75 for STPD⁷ suggest genetic variation explains a large portion of the disease phenotype. However, a complex etiology with roles for both genetic and nongenetic factors is likely.²¹

Environmental risk factors for hypoadrenocorticism in dogs remain poorly understood. Few studies have investigated roles for nongenetic factors. In addition to breed,^{4, 22} associations with age,²² sex,^{4, 22} body weight,²² insured status,²² and spay/neuter status^22-24 have been identified. However, some of these variables (eg, breed, body weight)²⁵ are not independent of genetic influence and are not exclusively exogenous environmental exposures. To the best of our knowledge, no published study has comprehensively evaluated environmental exposures for association with hypoadrenocorticism in dogs.

Our objective was to conduct a survey of STPD and POWD owners to identify environmental risk factors associated with hypoadrenocorticism in these breeds. We hypothesized that environmental exposures associated with autoimmune disease in humans (eg, pollution, diet, pesticides, tobacco)^26-35 would be present at a higher frequency in dogs with hypoadrenocorticism compared to dogs without the disease. Identifying environmental factors associated with hypoadrenocorticism in dogs may benefit the health of dogs through improvements in clinical management using targeted interventions of associated exposures, and by identifying novel therapeutic targets by research investigating the molecular consequences of these exposures.

2 METHODS

3 RESULTS

A total of 6286 survey responses were attempted. Duplicates (n = 6), ineligible dogs (owners selecting “no” to the consent/eligibility question; n = 7), and incomplete surveys (<20% of key questions answered; n = 1226) were excluded, leaving 5047 responses for initial analysis (Figure 1).

4 DISCUSSION

We evaluated common environmental exposures for association with hypoadrenocorticism in 2 dog breeds with a high prevalence of the disease. Significant associations were found for certain variables across multiple analysis models, including community type (eg, urban, rural, farm), ingestion of canned food, and exposure to lawn fertilizer (Tables 1 and 2, Table S1). Although biologically plausible, the modest effect sizes of these associations suggest low practical relevance.⁴⁵ Modest effect sizes may reflect indirect effects (eg, as a proxy for some other variable), insufficient power, or the influence of sample population characteristics (eg, overrepresentation of breeder-owned dogs in the control group).^{45, 46} Nevertheless, our study provides the first identification of environmental risk factors associated with hypoadrenocorticism in dogs. Follow-up studies that target these variables using probabilistic sampling in a larger number of cases will determine the reproducibility and clinical relevance of these findings.

The strongest association found was for spay/neuter status, a factor previously reported to be associated with hypoadrenocorticism in dogs.^22-24 Despite being a secondary aim, the potential high clinical relevance suggested by the large effect size compelled us to further investigate this variable. The OR here was >2 times higher than previously reported for neutered males in our study (6.0; 95% CI, 2.6-13.9) vs (2.1 ± 0.5)²³ and (2.5; 95% CI, 1.4-3.5) in a study which evaluated males and females together.²² The OR for spayed females resembled those of prior studies: (2.5, 95% CI [1.4-4.5]) in our study vs (1.5 ± 0.3)²³ and (2.5; 95% CI, 1.4-3.5) in previous studies.²² Also consistent with a prior study²³ was the higher OR for neutered males compared to spayed females. Although our CIs were wider than those in prior studies, the differences likely result from the smaller number cases vs controls.^{45, 46} Furthermore, they do not overlap with effect size estimates reflecting significant differences.

The breed-restricted approach of our study may explain the higher ORs for spay/neuter status and hypoadrenocorticism compared to those of previous studies. Breed has been shown to influence the likelihood of developing hypoadrenocorticism in dogs.⁴ Prior studies evaluating spay/neuter status for association with hypoadrenocorticism were open to dogs from any breed.^{22, 23} Additionally, each study evaluated geographically distinct populations: 1 study evaluated United Kingdom (UK) dogs,²² another study included dogs seen at a university clinic in California,²³ whereas our study was open to dogs across the United States and Canada. That 3 independent studies in different validation populations found the same association, with the same pattern of higher ORs for neutered male vs spayed female dogs emphasizes the robustness of this association.

We initially were surprised by the high odds of developing hypoadrenocorticism in spayed/neutered dogs. We therefore took several steps, including a small follow-up survey, to increase our confidence in these findings. For example, we observed that, in our initial study population, the median age of control dogs was significantly lower than the median age of cases (3 years [IQR 1-6] for controls and 7 years [IQR 4-10] for cases; P < .001). We hypothesized that dogs in the younger control population may not have been eligible for a spay or neuter procedure. We therefore restricted our analysis to include only dogs with owners who had answered the spay/neuter status question, and dogs that were old enough to have been eligible for a spay or neuter procedure (conservatively estimated at ≥25 months at the time of the survey). Age distributions of cases and controls were not significantly different after this adjustment (P = .53; Table 1). Although the values of the initially observed OR contracted, they remained high (6.6; 95% CI, 3.2-13.7) for neutered males and (3.6; 95% CI, 2.1-6.4) for spayed females (Table 1).

In our final MLR analysis, these ORs decreased slightly for males (6.0; 95% CI, 2.6-13.9) and compared to females (2.5; 95% CI, 1.4-4.5). Therefore, we concluded that these associations were significant and not driven by factors we could otherwise control for in our study population. We acknowledge the many approaches to MLR analysis. Given our case-control study design, we approached our analysis with a focus on identifying independent predictors of hypoadrenocorticism. In our exploratory analysis and ULR models, we found spay/neuter status had the largest effect size of all potential exposures. This observation led to our decision to explore spay/neuter status as the primary exposure and consider other potential confounding factors in our analysis. Because we had a large sample size and many potential confounders, we applied both a stringent cut-off (P value of .05) and a priori knowledge to move variables into the MLR model. Because we intended to identify the independent association between spay/neuter status and hypoadrenocorticism, all other variables in the MLR model were considered only as confounders. This analytic approach was most appropriate for our objectives, because other approaches may have allowed for only consideration of the statistical significance, and spurious results might have been identified.

Given the strong association observed between hypoadrenocorticism and spay/neuter status, we followed this finding up in a subgroup of study participants to explore the timing between spay/neuter and hypoadrenocorticism diagnosis because our initial survey did not ask questions designed to interrogate this relationship. The majority of dogs (79%) were found to have been spayed or neutered before the hypoadrenocorticism diagnosis. More males (88%) were neutered before the hypoadrenocorticism diagnosis than females (72%). Therefore, we concluded that the higher proportion of spayed/neutered dogs among cases was not entirely explained by owners seeking this procedure for their dogs in response to a hypoadrenocorticism diagnosis to prevent further breeding. Furthermore, the median months for a hypoadrenocorticism diagnosis after the spay/neuter procedure was >1 year later (17.9 months for males/females combined' 14.9 months for females and 21.1 months for males). This finding suggests that, in most cases, the spay/neuter procedure did not trigger an adrenal crisis nor did it uncover pre-existing disease. Although these data are consistent with the hypothesis that being spayed or neutered contributes to the development of hypoadrenocorticism, they are only associations.⁴⁷

Additionally, our findings are supported by other evidence. For example, in a large epidemiological study in a Swedish population of dogs,⁴ POWD and STPD were found to be breeds commonly diagnosed with hypoadrenocorticism.^{6, 7} However, the prevalence in these breeds (1.2% for both POWD and STPD)⁴ was much lower than those found in another study of North American populations (8.6% for STPD⁷ and 7.0% for POWD,⁴⁸ respectively). Lower rates of hypoadrenocorticism might be expected in populations for which spay and neuter procedures are less common if spay/neuter status contributes to the pathogenesis of the disease. In Sweden, only 1.1% of dogs are estimated to be spayed or neutered,⁴⁹ compared to an estimated 64% of dogs in the United States.^{23, 50} Our subgroup analysis was limited by containing only approximately a third of the case population, 29% (104/358) of all eligible cases or 32% (104/330) of age-restricted cases. It is possible that the association between spay/neuter status is a proxy for some other yet to be identified causative variable or a latent confounding variable.^{47, 51, 52}

Other novel insights include evidence of effect modification by spay/neuter status. A univariate stratification analysis showed different environmental variables were associated with spayed vs neutered dogs (Table S3). This finding suggests sex hormone profiles may influence both overall risk as well as susceptibility to other environmental risk factors for developing hypoadrenocorticism in dogs. Future etiological studies should account for the influence of spay/neuter status in study design. Other research is needed to determine if spay/neuter status plays a contributing role to the development of hypoadrenocorticism that can be explained by biological mechanisms or if spay/neuter status is a proxy for other variables such as better access to veterinary care.

Although we did not evaluate the biological mechanisms for the role of spay/neuter status in the development of hypoadrenocorticism, the literature supports several potential explanations. For example, sex hormones have immune-modulating effects in mammals. Gonadal androgens have broad immunosuppressive effects in both the central and peripheral immune system,⁵³ including suppressing antibody production, T-cell proliferation, natural killer cell activity, and cytokine production.^{54, 55} Loss of these hormones through neutering and their corresponding immunosuppressive functions may contribute to an immune-mediated pathogenesis in hypoadrenocorticism. Additionally, both androgen and estrogen receptors can be found on immune^{56, 57} and adrenocortical cells, and sex hormones influence the physiology of these cells.^{58, 59} Sex hormones also play roles in cellular repair, turnover, and proliferation of the adrenal glands, as well the recruitment of adrenal stem cells.^{58, 60} Loss of these hormones may contribute to dysfunction of these cells resulting in adrenocortical atrophy that is observed in these dogs.⁶¹

We acknowledge several limitations inherent to the methodology of our study. Nonprobabilistic convenience sampling is associated with several issues: (1) sampling error cannot be calculated, making results not generalizable beyond the study population, and (2) potential for participation/nonresponse bias.⁶² That is, participants self-select into the study and those with a higher motivation to participate may be overrepresented (eg, the higher proportion of POWDs relative to STPDs in the cases may reflect higher motivation to participate among owners of POWDs with hypoadrenocorticism). Also, advertising the survey through breed clubs could have led to sampling bias such that breeder-owned dogs are overrepresented, possibly explaining the overrepresentation of certain breeder-associated behaviors in the control group (eg, use of professional grooming and not spaying or neutering). Furthermore, a nonprobabilistically sampled population may not represent a target population because of poor sampling coverage or selection effects.⁶² Comparing our nonprobabilistically derived US sample to a well-defined estimate of the US sample frame using spatial mapping suggested that our sampled US population was geographically representative of the target US population (Figure S4). Despite these known limitations, we used non-probabilistic sampling for certain advantages to performing exploratory research (eg, low cost, expedience, ability to collect a large sample).⁶³

Self-reported survey data are vulnerable to several types of bias (eg, misclassification, recall, response, item, and unit nonresponse bias).^64-66 Among these concerns are potential effects from missing data. We calculated missing data for each explanatory variable in the multivariable analysis and compared the differences between cases and controls using a Chi-squared test; statistically significant differences in missing data were not detected for these variables. Self-reported data are also susceptible to misclassification bias. For example, some respondents may have incorrectly categorized their dogs as not having hypoadrenocorticism because of misdiagnosis or misunderstanding of the term. However, misclassification of hypoadrenocorticism status is unlikely in our target population for the following reasons: (1) hypoadrenocorticism is common in these breeds and most owners are aware of its importance; (2) hypoadrenocorticism is fatal if not treated and treatment is highly specific; (3) there were several opportunities to self-correct classification (including several hypoadrenocorticism-specific questions); and (4) pilot testing of the survey tested comprehension of terms.

Spatial data have other limitations. For example, the scale of GIS data influences whether or not an association might be detected.^{67, 68} Differences in how data from diverse municipalities are collected, managed, and stored may introduce standardization errors. Other limitations stem from the data analysis (eg, type II error). To illustrate, although tobacco exposure was evaluated, its use was low for both case and control cohorts (approximately only 3% of households sampled; Table S1). Our study was likely underpowered to detect a real effect for this variable. Therefore, the lack of association for a particular variable in our study should not be regarded as evidence of a lack of risk. Future studies with sufficient power are required to detect an effect for variables with low minimum effect sizes.

Likewise, significant results are not inherently meaningful. For example, several air quality variables had statistically significant associations with hypoadrenocorticism (eg, PM₁₀, NO₂, O₃, yearly total and median number of days above the unhealthy air quality index [MD > UAQI]; Table S2A,B). However, medians compared between cases and controls for many of these variables were often near zero (eg, PM₁₀, NO₂, and MD > UAQI), small enough to question if these differences are meaningful even if a statistically significant differene exists. Different study design methods may be required in future studies to more effectively make use of GIS databases in understanding environmental exposures associated with hypoadrenocorticism in dogs.

Although the main objective of our study was to identify environmental exposures associated with hypoadrenocorticism in dogs, the most important finding in our study population was the association of spay/neuter status. This finding confirms prior work^{22, 23} and lays the groundwork for future studies to test the reproducibility of our work and to investigate underlying biological mechanisms that might contribute to the disease. Future epidemiologic studies might use probabilistic sampling (eg, prospective studies) with attention to the temporality between spay/neuter status and hypoadrenocorticism diagnosis. Other studies in cell-based model systems such as canine adrenal organoids or adrenocortical cell culture could generate insights into how different hormone profiles affect these tissues. These studies would represent a promising new approach to understanding underlying triggers of this common disease that carries a high degree of morbidity in certain dog breeds.

CONFLICT OF INTEREST DECLARATION

Steven G. Friedenberg serves as Associate Editor for the Journal of Veterinary Internal Medicine. He was not involved in review of this manuscript. No other authors declare a conflict of interest.

OFF-LABEL ANTIMICROBIAL DECLARATION

Authors declare no off-label use of antimicrobials.

INSTITUTIONAL ANIMAL CARE AND USE COMMITTEE (IACUC) OR OTHER APPROVAL DECLARATION

Authors declare no IACUC or other approval was needed.

HUMAN ETHICS APPROVAL DECLARATION

Authors declare human ethics approval was not needed for this study.