Study Population and Design

The COHERE (15) and STROBE (16) statements were used to guide reporting of the findings and the preparation of this article. We defined a household as >1 persons >18 years of age living with >1 pet that does not live solely outdoors. Pets were defined as dogs, cats, ferrets, and hamsters, based on previous research documenting experimental COVID-19 infection in these species (17,18). We conducted this study in King, Snohomish, Yakima, Whitman, Pierce, Spokane, and Benton Counties in Washington and Latah County in Idaho during April 2020‒September 2021. The COVID-19 and Pets Study is a cross-sectional study with individual-level and household-level data collection. Study participation involved 2 components: an online survey, followed by animal sampling.

Recruitment and Eligibility

Households were recruited through partnerships with other COVID-19 clinical trials and community studies, as well as by social media, word of mouth, community partners, and contact tracers from Public Health Seattle and King County during case investigation/contact tracing calls. We screened persons for eligibility by using the UW Research Electronic Data Capture (REDCap) system (19), a Health Insurance Portability and Accountability Act‒compliant web tool for clinical research, which had criteria including county of residence, pet ownership, and >1 household member with confirmed SARS-CoV-2 infection by PCR or antigen testing by a provider or laboratory. Animals with known fearful or aggressive behavior were excluded. However, other animals in the corresponding household were eligible.

Ethics Statement

This study received ethics approval from the University of Washington Institutional Review Board (STUDY00010585) and the Office of Animal Welfare (PROTO201600308: 4355–01). We obtained informed consent by using REDCap or over the telephone with the study coordinator, after the nature and possible consequences of study involvement had been explained.

Survey

A household member completed a survey before the sampling visit was scheduled. Human survey items included COVID-19 symptoms, onset, and severity; concurrent conditions; vaccination status, dates, and type; and reported COVID-19‒like illness of any other household members, including those without confirmatory testing. Animal survey items included veterinary clinical variables, history of illness compatible with SARS-CoV-2 infection, and contact with specific members of the household. Environmental survey items included type and size of home, type of flooring (e.g., carpet, wood), and availability of outdoor space for pets to roam.

At the sampling visit, the field team inquired about updates for human and animal household members, including new hospitalizations, symptoms, or COVID-19 diagnoses. The study team also reviewed SARS-CoV-2 test results to confirm date and positive result; self-test results were not accepted.

Animal Sampling

A team of 2 study personnel, including at least 1 veterinarian, performed sampling in the home of a participant or at a veterinary hospital. No chemical restraint was used because of biosafety concerns, and no muzzles were used.

The team used species-appropriate restraint standard techniques for venipuncture and collection of 3 mL of blood into a labeled serum separator tube. Swab specimen samples collected from rostral nares and the caudal oropharynx were placed into 1 Primestore Molecular Transport Medium Tube (Longhorn Vaccines and Diagnostics LLC, https://lhnvd.com). A fecal swab specimen collected from the rectum was placed into a separate tube. All participants received educational information from the field team about measures to mitigate household COVID-19 transmission. Swab and serum samples were transported on ice within 24 hours to the Washington Animal Disease Diagnostic Laboratory (WADDL) for PCR and antibody testing.

Testing

We performed RNA extraction and SARS-CoV-2 real-time reverse transcription PCR (RT-PCR) for the SARS-CoV-2 RNA-dependent RNA polymerase gene (RdRp) as described for respiratory and fecal swab specimens (20). We also performed a second RT-PCR targeting the N1 region on the nucleocapsid gene as described for RdRp-detected samples (CDC 2019-Novel Coronavirus real-time RT-PCR [2019-nCoVEUA-01] (21). There was 100% concordance (agreement) between the RdRp PCR and N1 PCR. After initial viral detection by PCR, 3 dog samples and 1 cat sample were submitted to the University of Minnesota Genomics Center (Oakdale, MN, USA) for whole-genome sequencing (WGS) (22). A second cat sample was submitted to the USDA National Veterinary Services Laboratory (NVSL; Ames, IA, USA) for WGS. Mutational analysis was performed by using the GISAID EpiFlu Database CoVsurver: Mutation Analysis of hCoV-19 (23,24). We deposited all 5 sequences into GISAID (accession nos. EPI_ISL_7845315–8, and EPI_ISL_8897004) and assigned SARS-CoV-2 lineages by using the PANGO lineage tool (25,26).

SARS-CoV-2 Spike Protein Receptor Binding Domain ELISA

WADDL developed canine and feline SARS-CoV-2 ELISAs by using recombinant SARS-CoV-2 spike receptor‒binding domain (S-RBD) protein as antigen. The recombinant S-RBD protein was obtained from the University of Washington Center for Emerging and Reemerging Infectious Disease Laboratory of Wesley Van Voorhis through an institutional material transfer agreement. WADDL used an in-house standard operating procedure for indirect ELISA of SARS-CoV-2 in 96-well format based on a previous report for humans (27).

The major components of the assay were recombinant S-RBD coating of plates as target antigen (2 μg/mL in carbonate-bicarbonate buffer; Sigma-Aldrich, https://www.sigmaaldrich.com); a 1:100 dilution of test serum diluted in ChonBlock ELISA Buffer (Chondrex Inc., https://www.chondrex.com); anti-dog IgG‒horseradish peroxidase conjugated as linker (goat anti-canine IgG; Southern BioTech, https://www.southernbiotech.com); and the 3,3′,5,5′-tetramethylbenzidine liquid substrate system (Sigma-Aldrich) to develop the optical density (OD). Plates were blocked with ChonBlock ELISA buffer per the manufacturer’s instructions, washing solution consisted of phosphate-buffered saline plus 0.1% Tween 20 (Sigma-Aldrich), and plates were read on a plate reader at 450 nM. Serum samples were tested in triplicate and used at the test OD.

For the dog RBD ELISA, negative controls consisted of serum samples collected from 6 dogs before COVID-19, archived at WADDL and tested for 5 canine viruses: adenovirus, distemper virus, coronavirus, parainfluenza, and parvovirus. All 6 samples had antibody on >1 of the tests performed. However, no serum sample reacted in the SARS-CoV-2 canine RBD ELISA.

For the cat RBD ELISA, negative controls consisted of serum samples collected from 3 cats before COVID-19 from WADDL archives and tested for feline coronavirus IgG and feline panleukopenia virus IgG. Two of the 3 samples had antibody on >1 of the tests performed (including 2 for feline coronavirus). However, neither sample reacted in the SARS-CoV-2 feline RBD ELISA.

We tested negative controls in triplicate and used the mean as the negative control OD. We used a ratio of test OD:negative control OD to determine the results. The positive cutoff of 2.0 test OD:negative control OD ratio equated to the mean of negative controls +3 SDs of the mean. Use of +2 or +3 SDs from the mean OD of defined negative control serum samples is a commonly used method when no standard negative or positive control serum samples are available. Use of + 3 SDs from the mean of defined negative control serum samples was chosen as the most conservative SARS-CoV-2 RBD ELISA cutoff to reduce the risk for false-positive results.

We performed SARS-CoV-2 RBD ELISA in triplicate on 3 different occasions for all samples and tabulated the final results as a mean value obtained from the repeated testing. Initially, because no dog or cat in Washington or Idaho had previously been confirmed to be SARS-CoV-2 seropositive, the first antibody-positive case for each species and state was sent to the USDA NVSL for confirmation by virus neutralization (VN) assay in keeping with regulatory recommendations. Subsequently a subset of 30 SARS-CoV-2 RBD ELISA‒positive serum samples that had a range of ELISA output (20 dogs and 10 cats) and 4 SARS-CoV-2 RBD ELISA‒negative serum samples (2 dogs and 2 cats) were compared by inter-laboratory comparison to live SARS-CoV-2 VN testing performed at the USDA NVSL. Although a VN test is not a validation of an ELISA because they detect different biologic functions of antibody that could involve different epitopes, avidity or affinity, the SARS-CoV-2 RBD ELISA to VN comparison showed 91% overall agreement (31/34), and a Cohen κ of 0.68 (substantial agreement), a metric that takes into account agreement by chance.

Statistical Analyses

The primary aim was to estimate the burden of household SARS-CoV-2 transmission from humans to their pets. Secondary aims included describing the nature of human‒animal contact within households and identifying risk factors for household transmission, including human‒animal contact.

Outcome

We defined animal infection or illness with SARS-CoV-2 as an animal meeting >1 of the following criteria: SARS CoV-2 RBD ELISA‒seropositive status, PCR-positive status, or illness consistent with SARS-CoV-2 infection, hereafter referred to as illness, defined as participant answer of yes to the survey question “Since the time of COVID diagnosis/symptom onset in the household, has this animal had any new issues with difficulty breathing, coughing or decreased interest in playing, walking, or eating?” We parameterized serostatus as ELISA ratio, log-transformed for interpretability, and PCR-positive status and illness as binary variables.

Regression Models

We defined outcome as an animal case of SARS-CoV-2. Separate regression models were fit for each outcome definition.

Household-level exposures included residence in house versus apartment or condominium, home size in square feet, and the number of human confirmed SARS-CoV-2 cases. Animal-level exposures included sharing beds or bowls (separately) with human household members and SARS-CoV-2 positive household members taking precautions to prevent transmission to their pets. We also examined the association between canine seropositivity and illness compatible with SARS-CoV-2 infection in the animal and between seropositivity and time since the animal was first exposed, defined as 2 days before the first date any household member had symptoms of COVID-19 or a positive result.

Figure 1

Figure 1. Directed acyclic graph for human‒animal transmission of SARS-CoV-2, Washington and Idaho, USA. Squares indicate exposures of interest and circles indicate outcomes (approximated by serostatus, PCR result, and illness in separate...

We identified possible confounders a priori by using a directed acyclic graph (Figure 1). We defined the minimum sufficient adjustment set by using this graph and appropriate software (DAGitty, http://www.dagitty.net) separately for each exposure (28). Animal species was explored as an effect modifier by using a multiplicative interaction term, and stratified results presented for all cases in which this interaction term reached statistical significance (p<0.05).

For each exposure of interest, we implemented a generalized estimating equation approach with an exchangeable working correlation structure, household as the clustering variable, and binomial models with a logit (binary outcomes) or Gaussian (continuous outcomes) link by using the geepack package in R (29). For regression of ELISA ratio on illness and time since first exposure, we performed linear regression by using the glm() function in R.

Recruitment

Figure 2

Figure 2. Flowchart indicating serologic and PCR sampling for study of household transmission of SARS-CoV-2 from humans to pets, Washington and Idaho, USA. Of 119 dogs and 57 cats corresponding to 105...

A total of 107 eligible households enrolled and completed the survey; 83 households, corresponding to 100 dogs and 47 cats, had a sampling visit conducted (Figure 2). Of those animals, 6 dogs and 8 cats were not sampled because of temperament, leaving 94 dogs and 39 cats that had PCR results. An additional 13 dogs and 9 cats were safe to restrain for swab (PCR) samples but not for serum collection, leaving 81 dogs and 32 cats that had serologic results.

Descriptive Statistics

Figure 3

Figure 3. SARS-CoV-2 RBD ELISA serologic data for cats in study of household transmission of SARS-CoV-2 from humans to pets, Washington and Idaho, USA. PCR testing was complete for 39 cats, and...

Figure 4

Figure 4. SARS-CoV-2 RBD ELISA serologic data for dogs in study of household transmission of SARS-CoV-2 from humans to pets, Washington and Idaho, USA. PCR testing was complete for 94 dogs, and...

On average, at least 6 weeks (dogs) and 2 weeks (cats) elapsed between the last human COVID-19 diagnosis in the household and animal sampling (Table 1). Of the 119 dogs and 57 cats who had completed surveys, 20 dogs (20.4%, 95% CI 12.9%‒29.7%) and 19 cats (38.8%, 95% CI 25.2%‒53.8%) had reported illness. Of the 94 dogs and 39 cats who were PCR tested, 4 dogs (5.3%, 95% CI 1.8%‒12%) and 3 cats (7.7%, 95% CI 1.6%‒20.9%) were positive for any swab specimen; of the 81 dogs and 32 cats who had serum collected, 33 dogs (40.2%, 95% CI 29.6%‒51.7%) and 13 cats (40.6%, 95% CI 23.7%‒59.4%) were seropositive. SARS-CoV-2 RBD ELISA OD:negative control OD ratios in seropositive animals ranged from 2.03 to 21.22 for dogs (Figure 3) and from 3.01 to 30.35 for cats (Figure 4).

Of the 94 dogs and 39 cats who were PCR tested, 5 dogs (cycle threshold [Ct] 26.0–37.67 for RdRp PCR and Ct 26.07–37.67 for N1 PCR) and 3 cats (Ct 27.03–39.97 for RdRp PCR and 27.03–39.97 for N1 PCR) were PCR positive by nasal/oropharyngeal swab specimens; 1 of these dogs was also PCR positive by a fecal swab specimen (Ct 39.20). Five PCR positive samples (2 cats and 3 dogs) had Ct values sufficient for WGS (Ct<30): The earliest cat sample (April 2021) that underwent WGS was in the Pango clade B.1.2. Another dog sample was identified as the Delta sublineage B.1.617.2.103 (AY.103), and the other 3 samples (1 cat and 2 dogs) were identified as Delta sublineage B.1.617.2.25 (AY.25). Of the 5 PCR-positive dogs, 3 were PCR positive before being seropositive and 2 were simultaneously PCR positive and seropositive.

There were 11 households that had >2 positive animals, and among multi-pet households that had >1 positive pet, mean prevalence (PCR or serology) was 91%. Of 8 PCR-positive cases, all were detected after April 2021, when the first case of the Delta variant was documented in Washington.

Nearly one third of dogs engaged in activities outside the household during periods of human isolation or quarantine. More than 50% of cats and dogs resided in households whose residents reported awareness of CDC guidelines to prevent human‒animal transmission of SARS-CoV-2, and 48 (41%) dogs and 17 (30%) cats resided in households that reported taking precautions to prevent transmission to household pet(s). No cats and only 2 dogs resided in a household in which an infected person was hospitalized for COVID-19. Nearly all dogs (83%) and most cats (72%) had access to yards or gardens and were allowed on furniture (86% of dogs and 100% of cats), and most dogs and cats were kissed by (75% of dogs and 68% of cats) and shared beds with (69% of dogs and 73% of cats) human household members. Almost all bowls for dogs (91%) and cats (95%) were washed in the kitchen.

Regression Models

We calculated results of regression models as prevalence odds ratios for the binary outcome of illness, reflecting the cross-sectional design of this study, and as exp^β for the outcome of ELISA ratio, which can be interpreted as the relative change (ratio scale) in ELISA ratio for a 1 unit change in the exposure (Table 2). Because so few animals were PCR positive, we did not run regression models for that outcome. With the exception of house size, which was adjusted for house type because the minimum sufficient adjustment set was small for that exposure, confounders were not adjusted for because of concerns regarding overfitting arising from the small sample size. Effect modification by species was found only for house type.

Dogs residing in houses on average had a 79% (95% CI 2%‒211%) higher ELISA ratio than dogs residing in apartments or condos, and the inverse association was detected for cats (49% lower mean ELISA ratio, 95% CI 75% lower to 3% higher) and for the outcome of illness in both cats and dogs (48% lower prevalence odds, 95% CI 80% lower to 34% higher). This association reached statistical significance for dogs only. No other effect estimates reached statistical significance. However, there were positive trends across both outcome definitions for bed sharing with humans, sharing bowls, and being indoor only and a negative effect for precautions taken to prevent SARS-CoV-2 transmission after diagnosis. We also found that the ELISA ratio was positively associated with illness. However, we did not find evidence of an effect of time since first exposure on ELISA ratio or of house square footage on either outcome.