There are a couple schools of thought regarding the need for a CT before an LP in the setting of infectious cerebral disease. The traditionalist school of thought: herniation. The pragmatist school: no big deal.
This article falls on the side of “no big deal”, which was probably the bias of the authors prior to its conception. These authors looked at comatose children in Malawi with suspected malaria. They analyzed the mortality outcomes of 1,827 patients, including 1,470 who received an LP and 357 who did not. Unadjusted mortality was higher in those who did not receive an LP, for which the authors attempted to adjust using propensity-based analyses, or by directly comparing those who had documented brain swelling on MRI or with papilledema. Using their admittedly small numbers in their retrospective cohort, they did not find any signals of harm relating to overall mortality or herniation precipitated by LP within 12 hours of procedure.
We probably will only ever get this level of evidence regarding the safety of LP in the critically ill with elevated ICP secondary to infection. Adverse events are rare, regardless, and it will always be difficult to shake out the confounding features of the malignant infection. I tend to agree with these authors that LP is safe in a stable patient without localizing neurologic signs, but it is entirely reasonable to take the opposite view.
“Safety of lumbar puncture in comatose children with clinical features of cerebral malaria”
http://www.neurology.org/content/early/2016/10/28/WNL.0000000000003372