Dogs and cats commonly present to veterinary hospitals with urinary bladder disease, but despite their clinical importance and comparative potential to human diseases, bladder diseases in Australian dogs and cats are under investigated. In veterinary pathology, insufficient levels of diagnostic agreement can occur, and this is influenced by sample quality as well as the pathologist’s own experience, training, and cognitive biases. Logistic regression is a statistical technique which, when applied to veterinary histopathology, could improve pathologist agreement. Thus, there were two overarching goals of this thesis - to investigate the pathology and comparative potential of canine and feline urinary bladder disease in Australia, and to explore the utility of logistic regression modelling in improving inter-pathologist agreement.This project conducted a retrospective evaluation of pathology cases of canine and feline urinary bladder tissue from the veterinary pathology archives of the University of Queensland School of Veterinary Science and Murdoch University’s School of Veterinary and Life Sciences, with prospective sampling from veterinary clinics and a commercial veterinary pathology service in South East Queensland. The demographics of the dataset were examined using proportionate morbidity and logistic regression to identify associations between animal factors and the diagnosis. Secondly, a comprehensive histological evaluation was undertaken of every sample, with logistic regression modelling performed to identify associations between histological variables and diagnosis. Thirdly, a subset of canine and feline diseased and normal bladder tissue samples was tested for biomarker expression using immunohistochemistry and polymerase chain reaction. This combined approach tested if retrospective samples were of sufficient quality, and when validated provided quantity as well as cellular location of the target biomarkers. To further investigate the comparative potential of feline idiopathic cystitis (FIC), a systematic review was conducted on biomarkers in bladder pain syndrome (BPS) compared to FIC, following the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines. Finally, to investigate the role of logistic regression modelling in veterinary pathologist agreement, the modelling of histological variables was used to formulate a predictive probability tool which we then tested on four pathologists evaluating the same set of twenty-five slides, with diagnostic agreement evaluated using the Fleiss kappa statistic.The main findings from the demographic analysis were a higher risk of bladder neoplasia in dogs compared to cats, increasing risk for bladder neoplasia with age, and decreased risk for cystitis in neutered animals. Next, logistic regression modelling on the histology dataset of canine and feline urinary bladder tissue from Eastern and Western Australia identified six significant variables that3were associated with the diagnosis – urothelial ulceration, urothelial inflammation, neutrophilic submucosal inflammation, submucosal lymphoid aggregates, amount of submucosal haemorrhage, and species. These six variables were used to create a predictive probability tool for bladder disease diagnosis. The pathologist agreement study revealed a good level of agreement between the four pathologists when diagnosing neoplastic lesions, but poor to fair agreement for cystitis, urolithiasis and normal bladder tissue. Agreement between pathologists did improve when signalment and clinical history was provided, with mixed results on inter-pathologist agreement when the predictive probability tool was used. However, the predictive tool did prove valuable in increasing the agreement of the study pathologists’ diagnosis with the reference diagnosis. There were multiple other confounders at play in this experiment such as variable digital slide quality and different interpretations of the study instructions. A systematic review on biomarkers in bladder pain syndrome revealed that nerve growth factor is the most likely urine biomarker to be useful in the diagnosis of human BPS. The aim of this review had been to compare biomarkers in BPS to those in FIC, however an unexpected variability in the study parameters meant we could not fulfil this goal. A final laboratory-based investigation of biomarkers of canine and feline bladder diseases revealed two findings - that archived formalin fixed, paraffin embedded tissues are not good samples for PCR experiments, and secondly that tight junction protein-1 may be a promising tissue biomarker for differentiating between some urinary bladder diseases in dogs and cats.In conclusion, this thesis has undertaken a comprehensive analysis of the pathogenesis and comparative potential of canine and feline bladder diseases and is the first to apply logistic regression modelling to veterinary histopathology diagnosis and to improving inter-pathologist agreement. Logistic regression modelling is a promising tool for veterinary pathology. Dogs and cats are potentially good comparative models for human bladder diseases; however, inconsistent case definitions in human research complicates veterinary and medical field alignment. Finally, a collaborative multicentre approach would be invaluable to collect high quality prospective samples of feline idiopathic cystitis cases to allow further investigation into this disease. In summary, the comprehensive approach utilised in this thesis has provided valuable information on bladder disease in cats and dogs and sets a foundation for further work in this field.
Read full abstract