Blast from the "past"

Reuben posted this on his site in September...


September 21st, 2010

Traumatic LP for Meningitis

by seth in .fevermeningitis/encephalitis
leave a comment
Special thanks to Reuben for inviting me to post this review from my teaching resident rotation.
Question:
Is a “corrected” CSF WBC count accurate for diagnosing meningitis for a traumatic LP?

Background:
Traumatic lumbar punctures may obscure accurate diagnoses. Many authors suggest correcting the WBC count by various methods — the most popular seem to be either 700 RBC = 1 WBC, or by using the actual patient’s RBC:WBC ratio in the blood. While this seems intuitive, does it work?
Answer:
Probably not.
Basically, no; the calculations are not helpful. But if the WBC count is MUCH higher than expected, it’s probably a positive tap.
Key points:
  • The sources I could find simply assert that correction is a viable method; I could not find any actual evidence that these corrections are valid.
  • Multiple small studies show that corrections are generally not accurate (including ref. 1), with ROC curves equivalent regardless of how — or if! — correction is applied
  • However, a few small studies also show that bacterial meningitis may be obvious despite a traumatic tap (refs 2 & 3):
If the “observed:predicted” ratio of CSF WBCs is >10, then some authors conclude that it indicates bacterial meningitis. Sensitivity & specificity are both around 80-90% with this method.
I think a higher threshold is probably better (ratio >100) — see images below.
Example:
CBC:
5 RBC (Hgb 15; Hct 45)
5 WBC
This is a predicted ratio of 1000:1 (RBCs are reported as 10^6/mcL and WBCs are 10^3/mcL)
A purely traumatic tap in this patient would be expected to look like this:
CSF
2000 RBC
2 WBC
If the CSF looked like this:
2000 RBC
20 WBC
than it is “likely” to be bacterial meningitis (Observed:Predicted = 10)
Looking at the data, I think we can all agree that this CSF is infected:
2000 RBC
200 WBC
(Observed:Predicted=100)
Here are the results from the Bonadio paper:

Bonadio data
Looking at their raw data, the ratio of 100 looks like a much better diagnostic cutoff, although it is probably best to still treat (i.e. antibiose & admit) pending more accurate tests (i.e. culture) if the picture is less clear.
Here is a ROC curve for their data, which looks pretty good altogether:
References:
  1. Greenberg RG, Smith PB, Cotten CM, Moody MA, Clark RH, Benjamin DK Jr. Traumatic lumbar punctures in neonates: test performance of the cerebrospinal fluid white blood cell count. Pediatr Infect Dis J. 2008 Dec;27(12):1047-51.
    There a number of similar small studies that all agree that adjustments are not useful.
  2. Bonadio WA, Smith DS, Goddard S, Burroughs J and G Khaja. Distinguishing cerebrospinal fluid abnormalities in children with bacterial meningitis and traumatic lumbar puncture. The Journal of Infectious Diseases. July 1990: 162(1): 251-254.
  3. Mayefsky, JH. Determination of leukocytosis in traumatic spinal tap specimens. The American Journal of Medicine. June 1987: 82(6): 1175.
NB I didn’t put references for any of the textbooks or papers (most of which refer to the same 2-3 textbooks) that simply assert that calculations are helpful.

Comments