Saturday, May 25, 2019

The Test is Not the Truth: One Week in the Lonely Life of a Bayesian Clinician

If there is one thing you should remember about clinical decision making it is this:  the test is not the truth.  A diagnostic test raises or decreases the prior or pre-test probability (PTP) of the disease under consideration.  The amount of increase or decrease in probability with a positive or negative test depends on the starting probability and the likelihood ratio of the test.  (LR+ = sensitivity/1-specificity; LR- = 1-sensitivity/specificity).  If we don't attend to the PTP of disease, serious diagnostic errors and therapeutic misadventures may result.  This is especially true when a low PTP disease is diagnosed on the basis of a test with poor sensitivity and specificity (and a LR not much greater than 1 or 2 or even 4 or 5).  Several examples of this came up a while back.

A woman presented with thunderclap headache and had recurrent seizures during initial evaluation.  A differential diagnosis was formulated and it included PRES (posterior reversible encephalopathy syndrome) with a PTP of about 20%.  Subarachnoid hemorrhage was excluded with CT and LP and the PTP of PRES rose to about 40% (since it occupied some of the probability space previously occupied by SAH once the latter was excluded.)  The subsequent MRI images were consistent with PRES.  Nonetheless, a vascular MRI was ordered to "exclude the possibility of cerebral vasculitis".  The problems are twofold.  First, the probability of PRES is now on the order of 70% if the sensitivity and specificity of MRI are on the order of 80%, and it is 85% if sensitivity and specificity are each 90%.  (Go ahead and plug some numbers into the calculator on the sidebar of the blog.)  This probability meets or exceeds the probability threshold to both consider the diagnosis made, and to take action based on it.  In this case inaction and supportive care are indicated.  Even if a vascular MRI were consistent with cerebral vasculitis, which has a PTP an order of magnitude or more less than PRES, the diagnosis is still PRES.  The truth is not in the test, the truth is in the rationally considered diagnostic process of which the test is one part.

A 60ish man with a family history of cirrhosis and a questionable history of alcohol use was admitted with sepsis and bacteremia from a confirmed source.  His platelets fell to a nadir of 15K and his bilirubin rose to 25, in the absence of imaging abnormalities other than fatty liver.  Sepsis can cause low platelets but it is not common for them to fall this low.  So, the possibility of TTP was raised and an "ADAM-ANTS-13" (my facetious name for it) level and other testing was ordered.  Nonetheless, the probability of sepsis is 100%, so even if it causes very low platelets only 5% of the time, this probability still swamps that of TTP, with an annual incidence of approximately 4 per 100,000 (that's rare).  For the patient to have TTP, we would have to build a case for it, and for it being coincident with sepsis.  Worse, if one of the tests for TTP is positive, it is going to be a false positive, but I can guarantee you there will be hand-ringing, and PLEXing before the dust settles.  Next, ceruloplasmin and a raft of serologies were ordered to explain the hepatic failure.  His MCV was 100 and GGT 250.  Alcoholic hepatitis did not completely fit the picture, but this "uncommon presentation of a common illness is more common than a common presentation of an uncommon illness."  The annual incidence of Wilson's disease is about 3 per 100,000, quite rare.  Nevermind that he's the wrong age and his alkaline phosphatase was elevated, etc.

Out of the MICU and off to consults.  A woman with Wegener's granulomatosis who is on dialysis is having hemoptysis and opacities on CXR.  We are asked to do a bronchoscopy and BAL to evaluate for "diffuse alveolar hemorrhage" (DAH) which would prompt plasma exchange and other immunosuppressive therapies.  The problem is that the alveolar hemorrhage that we discover is just as likely to be from renal failure and volume overload as it is from an inflammatory process, so we learn nothing of use from finding it.  We suggested repeating imaging after several days of aggressive dialysis to allow us to make some potentially strong inferences about the "volume overload causing alveolar hemorrhage" hypothesis, but wound up doing the BAL anyway to exclude infection.  She's going to get intensive dialysis either way, and if immunosuppression is started concurrent with that and she improves, we have the classic "true, true, and unrelated" problem - we can't know which therapy is responsible for any benefit that is observed and we may commit her to prolonged immunosuppression with all risk and no benefit.  The take-home?  Response to therapy will provide a more specific test of the hemorrhage hypothesis than a finding of hemorrhage.

Next up, a pancytopenic patient with a nodular opacity in a dependent a segment of the lung, distal to a pre-existing lesion in a feeding airway.  A bronchoscopy is requested to exclude fungal infection and obviate antifungal therapy.  The problem?  Both the sensitivity and specificity.  The sensitivity of the BAL for a dependent lesion in a potentially obstructed feeding airway will not be very high, so not finding fungus will not exclude fungus.  Finding aspergillus may not indicate infection with this organism either, especially since the patient has a dental abscess which may explain the lesion (even though it does not explain its location).  One potential solution - since the patient is stable, treat with antibiotics for several days and re-image.  If the lesion improves, a strong inference can be made that it is bacterial in nature.  If it or he worsens, a strong inference can be made that it is fungal.

The worrisome common thread in these cases is that rather than a logical analysis of how to proceed in the face of diagnostic uncertainty, testing is ordered and the results treated as if they have 100% sensitivity and specificity (and infinite or undefined LRs) that swamp the PTP.  The test is not the truth.  The truth must be sorted out by a careful evaluation of the evidence in the case with testing as an adjunct.  Finally, the Bayesian can use clinical reasoning to arrive at tentative conclusions about diagnoses in the absence of testing by attending carefully to the PTP of disease and acknowledging the limited likelihood ratios of the available tests.

1 comment:

  1. Alas the DAH samples tested positive for PJP on PCR but not on DFA, and the 1,3-Beta-D was negative. This is very likely to be a false positive PCR result, but who could resist treating it?


Note: Only a member of this blog may post a comment.