Those who are writing a thesis, have just upgraded from one bibliographic manager to another, or have spend a week flying around your (ex) region collecting printed forms to tell a prospective employer you are not a danger to their staff, patients or cutlery may read the title one way.
Those who have been pondering the value of diagnostic test accuracy studies with imperfect “gold standards” may have another.
Take the situation of serious infection in children. We know someone’s got a serious infection because they a) look poorly and b) have a blood, urine or CSF culture that grows a Nasty Bug That Fits The Picture.
We also know that not all bugs that cause problems grown in culture bottles.
So how do we evaluate a new technology, such as PCR-based Bug Detectors; with what do we compare them?
If we’re comparing with blood cultures, then we’ve got three of the four squares of our usual ‘diagnostic table’ pretty much sewn up.
| Blood Cx +ve | Blood Cx -ve | ||
| PCR positive | A | B | |
| PCR negative | C | D |
Group A are those who are ‘real’ infections
Group C are those the PCR test failed to identify correctly
Group D are those who really aren’t infected
It’s group B that are most interesting – because we know that the ‘reference standard’ isn’t ideal – group B might contain
a) real false positives – failure of the PCR test
b) contaminants – bug was in the tested sample but not the patient
c) real ‘culture negative’ missed infections – bugs that were there in sufficient quantity to cause an infection but weren’t grown
d) low level bugs – too low to be a real infection but just floating by in the bloodstream (take as an analogy the ‘overdiagnosis’ issue of pulmonary embolism with better quality CT scans)
e) dead bugs – PCR don’t tell you if the DNA was growin’ or dyin’
What can we do when we hit this issue? As researchers, we can get all excited and start planning a series of studies to look at disentangling the elements. We can being in clever methodologists and multiple markers of infection to do latent class analyses. We can dream of pragmatic RCTs and Big Data.
As clinicians, we can weep softly into our generic-chain-store-coffee as the WI tea shop was closed down 8 years ago. And refuse to believe anything any more that damned scientists ever do. And just go with what we’ve been doing; look at the patient, argue with the microbiologists, talk to the family & keep treating everyone as an individual.
– Archi
(Thanks go to Dr Gray, Birmingham Children’s, for switching on the lightbulb in my brain about really why ‘cell B’ was a problem not just my murky discomfort.)