Laurie's Blogs.


Oct 2022

What the Heck is Bayesian Clinical Reasoning?

Laurie Edge-Hughes, BScPT, MAnimSt, CAFCI, CCRT

TRUST ME, YOU SHOULD READ THIS... despite the title!!

I was scrolling around looking to see what was new with the Cambridge IVDD Research Group, and I came across a study published by one of the investigators.


Khan S, Freeman P. Bayesian clinical reasoning in the first opinion approach to a dog with suspected thoracolumbar pain. J Small Anim Pract. 2022 Jun 25. 


I speculate at WHY this research group MIGHT have published such an article… but more to the point, is to answer the question I posed I the title.  What the heck is Bayesian clinical reasoning?


“Bayesian clinical reasoning is a way of taking into account the known pretest probability of a disorder (based on a mix of published reviews, studies and case reports, and expert opinion), and using this information to interpret the probability of that disorder following the results of a diagnostic test. In other words, it involves using our prioritised differential diagnosis list to inform and interpret our use of the diagnostic investigations we undertake.”


Essentially, this is something we do day in and day out in clinical practice (human or animal).  Where was the paper going with this?  


If a case walks through your door with suspected thoracolumbar pain, it might be difficult to differentiate true spinal pain from abdominal pain.  (I might personally disagree, but my manual assessment skills are pretty good, and I’ve sent dogs back to vets when I’ve suspected abdominal pain over spinal pain.)  According to the paper, realistically, if a dog comes in with thoracolumbar pain, you are most likely thinking intervertebral disc extrusion, pancreatitis, or other spine-related disorders, such as discospondylitis, neoplasia (of the vertebrae, meninges and spinal cord), and sterile meningomyelitis (MUO).  (On a personal note, I have sent dogs back to veterinarians for ALL of these things because their clinical findings didn’t add up.)


“Whilst many other causes have been reported for thoracolumbar pain, these are extremely rare and generally limited to isolated case reports. Our purpose is to aid in the diagnosis of those conditions that are either common or require early diagnosis in order to avoid inappropriate management. So, what is the most appropriate next step when presented with a dog with suspected thoracolumbar pain, but when logistical or financial constraints are an impediment to referral for a complete diagnostic investigation?”


How does Bayesian clinical reasoning affect our approach to the potential usefulness of these diagnostic tests?  We must first consider the prevalence of each possible differential diagnosis in our patients. In order to estimate this, it is important to take into account signalment, including particularly breed predisposition. For example, the middle-aged chondrodystrophic dog is far more likely to be suffering from an IVDE than any other possible cause.

“Perhaps the most useful information in terms of arriving at a likely diagnosis will be the progression of the clinical signs with empirical therapy. A dog who progresses to gastrointestinal signs is inevitably more likely to have pancreatitis, whilst a dog who progresses to pelvic limb ataxia and proprioceptive deficits will usually have suffered an IVDE. It is also likely that with rest and appropriate analgesia, a significant proportion of individuals will simply improve, and a definitive diagnosis will not have proved necessary (also helping to exclude discospondylitis and neoplasia).”


“If we simultaneously consider the signalment we can refine this process even further. Whilst specific prevalence data for each breed and age are lacking, clinical experience tells us that the 5-year-old dachshund will most likely have a disc extrusion, whilst the middle-aged miniature schnauzer more often has pancreatitis. In essence, clinical progression combined with signalment can almost be considered to be a “test” which will not only meet many of the criteria previously described but also, in the authors' experience, lead to many dogs recovering without the requirement for further testing.”


From the Summary, I choose to share this wise nugget of advice.


“Based upon the requirements of Bayesian clinical reasoning and the existing evidence, our conclusion is that no test meets all the requirements for a perfect Bayesian test. Instead at first presentation the signalment and progression of clinical signs are likely to be the most useful pieces of information diagnostically, but should signs recur or fail to improve with empirical therapy, survey radiography comes closest to satisfying the requirements for a suitable empirical (and relatively inexpensive and widely available) test according to the tenets of Bayesian reasoning.”


Ooh!  I can’t wait until they publish all of their data on Conservative Management of IVDD!!!!

I like where this is going, especially for those of us in rehab!!


Until next time,