2.1 Study population
This was a retrospective study of laboratory data from regular sample submissions to three different laboratories in North America, including the states of Georgia (laboratory A), New York (laboratory B), in the United States, and in the province of Ontario, Canada (laboratory C). Inclusion criteria comprised samples from respiratory specimens from domestic cats submitted to the participating diagnostic laboratories for the detection of feline respiratory pathogens between 2011 and 2020. Most cases had a case date, age, sex, accession number, assay code, assay name, and result (positive or negative). Clinical signs were provided for most cases except for those from Laboratory B. Incomplete data were identified as unknown for the purpose of this study.
Some combination of conjunctival, nasal, oropharyngeal, eye swabs, transtracheal washes, or lung tissues had been submitted (Supplementary Table S1) for each patient for detection of the following common feline respiratory pathogens: Bordetella bronchiseptica, Chlamydia felis, Mycoplasma, FeHV-1, FCV, and influenza A virus. Results were sent to us as de-identified data, and no additional clinical information was gathered.
To investigate the presence of pathogens in cats without clinical signs of respiratory disease (control group), nasal swabs were collected 4h to 24h postmortem from clinically normal cats (n = 51) that had been submitted to Laboratory A for post-mortem evaluation. Inclusion criteria of these control animals were based on the absence of a history of respiratory clinical disease according to the submitting veterinarian, and the absence of any post-mortem signs of respiratory disease. A board-certified pathologist performed complete postmortem and histological examinations to confirm that the animals were not affected by respiratory disease at the time of death. Cats with any macroscopic or histological lesion associated with respiratory diseases were excluded from the control group.
The information collected in this retrospective study was part of routine diagnostic procedures. Hence, it did not require Institutional Animal Care and Use Committee approval. Animal ID and client information were kept confidential.
2.1 Pathogen detection methods
Clinical samples submitted to laboratory A were analyzed by a quantitative PCR panel targeting Bordetella bronchiseptica 19, Chlamydia felis 20, Mycoplasma spp. 21, FeHV-1 20, FCV 22 and influenza A virus 23. At laboratory B, nucleic acid extraction was performed as previously reported 24, and quantitative PCR was performed for Bordetella bronchiseptica 25, Chlamydia felis 26, Mycoplasma felis 27, and influenza A virus 28. Detection of FeHV-1 and FCV was performed by virus isolation on cell culture. At laboratory C, quantitative PCR was performed for FeHV-1 and FCV.
2.2 Predictor and outcome variables
Information about age, sex, dates (seasonality), and clinical signs was obtained from the original sample submission form provided by the submitting veterinarian. The occurrence of pathogen by age was evaluated in four categories defined as kitten (1- to < 7-month-old, coded as 1), junior (7 month to < 3-year-old, coded as 2), adult (3 - ≤11-year-old, coded as 3) and senior (>11-year-old, coded as 4). To investigate whether seasonality has an impact on the occurrence of pathogens, data were divided into cold and warm seasons: the cold season was defined as October 15th to April 15th; the warm season was defined as April 16th to October 14th.
To assess the effect of the severity of clinical signs on the occurrence of pathogens, we categorized information regarding clinical signs of infectious respiratory diseases from the data obtained from the submission forms from laboratories A and C. Diseased cats were categorized according to severity of clinical signs: clinical score 1 (upper respiratory mild disease: coughing, sneezing, conjunctivitis, or nasal/eye discharge), clinical score 2 (upper respiratory severe disease: the same signs as clinical score 1 plus ulcers in the mouth, lethargy, depression, inappetence or fever), clinical score 3 (lower respiratory disease: pneumonia followed by one or more signs such as dyspnea, lethargy, depression, inappetence or fever (Table 1).
2.3 Statistical analysis
Differences in the proportion of pathogens across season, sex, clinical sign categories, and age were assessed by Fisher’s Exact test using Holm post-hoc test to adjust for multiple comparisons using R 4.3.0 software. P-values < 0.05 were considered significant.
A co-infection was categorized as 0 when there was no infection or infection with one pathogen, whereas it was categorized as 1 when cats were infected with more than one pathogen.
To determine whether specific pathogens were more likely to be present in co-infections, a network analysis was performed using Spearman's correlation coefficient via the 'igraph’ and ‘bootnet’ packages in R software 29. The network analysis was performed using only submissions that had results for all six pathogens: Bordetella bronchiseptica, Chlamydia felis, FCV, FeHV-1, Mycoplasma spp., and Influenza A virus.
A univariable ordered logistic regression model was employed to predict the four-tiered outcome of interest. The outcome categories were defined as follows: clinical score 0 represented no infection, score 1 represented upper mild respiratory tract clinical signs, score 2 represented upper severe respiratory tract clinical signs, and score 4 represented lower respiratory tract clinical signs and pneumonia.
The explanatory variables considered in the analysis included age (categorical variable), sex (binary variable), season (binary variable), and six pathogens (binary variables). Supplementary Table S2 displays the frequencies of demographics, temporal characteristics, and the presence and absence of six pathogens. Moreover, this analysis involved the consideration of ten co-infections consisting of two pathogens as explanatory variables. The frequencies of the presence and absence of these 10 combinations are presented in Supplementary Table S3, with influenza consistently showing a negative result. Furthermore, this analysis included consideration of ten co-infections involving three pathogens as explanatory variables (Supplementary Table 4). Variables with a proportion of binary results lower than 5% were excluded due to the substantial standard errors produced (Supplementary Tables S2-4).