A Bayesian analysis of variables causally associated with hair cortisol concentration in dogs with obesity

To identify clinical variables causally associated with hair cortisol concentration (HCC) in dogs with obesity using a Bayesian analysis.

A retrospective analysis of clinical data and samples gathered from a cohort of dogs with obesity undergoing therapeutic weight reduction.

Hair was clipped from two sites (jugular groove, antebrachium), from dogs attending a specialist obesity care clinic, and combined before storage at -20 °C. Hair cortisol concentration was measured by liquid chromatography mass spectrometry. Causal associations between HCC and different clinical variables were assessed, informed by a directed acyclic graph. Variables assessed included age, sex, breed group, coat colour, body fat mass, weight reduction and the presence of comorbidities. Statistical analyses involved Bayesian multi-level modelling, with the magnitude of causal effects estimated using simulation from the posterior probability distributions.

In total, 73 hair samples were collected from 52 dogs, with 31 providing single (before weight reduction) and 21 providing paired samples (before and after weight reduction). Dogs were of different ages, sexes and breeds, with most (44/52) having one or more comorbidities including orthopaedic, skin, cardiorespiratory, dental and neoplastic diseases. Mean HCC was 10.4 (standard deviation 19.52) pg/mg (logHCC 1.3, standard deviation 1.36). Bayesian multi-level models provided strong evidence that greater body fat percentage (98% probability) and presence of one or more comorbidities (>99% probability) were causality associated with increased HCC. Causal associations with other variables including, age, breed, sex, coat colour and season were less convincing.

Greater adiposity and having at least one comorbidity are causally associated with hypothalamic-pituitary-adrenal axis upregulation in dogs with obesity. Mechanisms warrant further investigation.