Does cefepime cause encephalopathy and death?

[Image source]
Encephalopathy is a pretty typical finding in the ICU setting, and often its etiology is clear: even setting aside iatrogenic causes (i.e. a midazolam drip), reasons to be sleepy or goofy during critical illness abound.

However, a not-uncommon scenario looks like this: a patient comes in seriously sick. Maybe they’re septic, or a multi-system trauma, or a major post-op case, or whatever you please. They are understandably altered and most likely sedated. You resuscitate them and they get better. A few days later, with things on the upswing, you wean their sedation and let them wake up.

Except they don’t. You try to be patient, figuring the drugs have built up and they were nearly dead 48 hours ago and they have a right to be a bit sluggish, but hours turn into days and they’re still largely unresponsive, with a RASS of -5 and little more than brainstem reflexes.

Eventually, as yet another day dawns with the patient still deeply obtunded despite every indication that you have fixed their underlying problem, you start to wonder if you’re missing something. Maybe you CT their head. Someone always suggests an LP. In the end, usually they wake up on their own, mere victims (presumably) of a prolonged stupor from their extreme metabolic derangements. At least, that’s the hand-waving I prefer.

However, there could be another culprit: it’s their cefepime.

Cefepime as a cause of encephalopathy

The idea that cefepime could be neurotoxic has been documented in a number of case reports that temporally associated its use—and its discontinuation—with the appearance and then resolution of alterations in consciousness, or most concerningly with seizure activity. This was described particularly, although not exclusively, among the renally impaired (Jallon 2000Chatellier 2002, Lam 2006, Fernandez-Torre 2006, Kim 2013).

However, the only large-ish study was by Fugate et al in 2013, reporting their experiences from the Mayo Clinic. In a retrospective chart review enrolling everyone who received 3+ days of cefepime in their ICU over a three-year period, they asked a panel of experts (with a modified Delphi format, aka “what do y’all think?”) whether the patient had experienced any neurotoxicity attributable to their cefepime. This was based upon temporal association of CNS derangement with both initiation and termination of the therapy, with no other plausible cause appreciated; notably, if the encephalopathy resolved spontaneously (i.e. with cefepime still on board), it was not counted.

In their population of 100 cefepimeurs, the drug was felt to have caused neurological sequelae in 15 cases. Seven were “definite,” five were “probable,” three were “possible.” Even with a conservative reading, that would suggest an incidence of at least 7%, which would be quite striking.

What were these complications? “Impaired consciousness,” presumably meaning some degree of obtundation, occurred in 13; myoclonus occurred in 11; disorientation in six; and nonconvulsive status epilepticus (NCSE) in a single case. The authors note that renally adjusting the dose seemed to decrease the risk, but was not reliably protective.

Myoclonus was either diffuse or in many cases localized to the facial muscles, particularly periocular. Unfortunately, only 60% of the “toxic” patients had EEGs, so NCSE may have been underreported. However, some EEG perturbations (diffuse background slowing, triphasic waves, multifocal sharp waves) were not uncommon.

Cefepime causes death?

It’s hard to discuss cefepime toxicity without mentioning another elephant in the room, one with an interesting history.

In 2007, Yahav et al. performed a systematic review and meta-analysis, pooling 57 studies involving head-to-head comparisons of cefepime versus other beta-lactam antibiotics. Despite reportedly similar baseline levels of risk, and no obvious differences in treatment failure or complications, 30-day all-cause mortality was higher in patients receiving cefepime (risk ratio 1.26, 95% CI 1.08–1.49, making an NNT of 50). Mortality was essentially higher in every subgroup analyzed, except patients with UTI, who had no mortality.

This being a rather alarming report, and despite some questions regarding the data quality, the FDA reexamined the issue, releasing their findings in 2009. After throwing in some additional studies Yahav had not included, and using some different statistical methods, they agreed that there was a small signal of hazard, but according to their number-crunching, it did not reach statistical significance. (Only a few subgroups were significant, most of which seem arbitrary and probably spurious, e.g. skin infections). In short, they were not impressed, and cefepime remains on the market.

Since then, one additional study (Al-Hasan 2011) has retrospectively compared cefepime against other beta-lactams for use in monomicrobial, aerobic Gram negative bacteremia, finding no difference in mortality after the usual efforts at statistical adjustment.

What to make of all this? Tough to say. The overall constellation of data seems to suggest that there may be some signal there, but certainly not a large one, and it is unclear which subgroups give rise to it. My intuitive guess is that the danger may lay in empirically using cefepime as the “smallest big gun,” and occasionally undertreating cases (such as anaerobic infections) that ought to instead have received piperacillin-tazobactam or a carbapenem.

Conclusions

Where does that leave us?

My thoughts:

  1. Cefepime remains a valuable tool in the antimicrobial armamentarium, especially as a way to deescalate or otherwise avoid the use of increasingly-common ubiquicillins such as pip-tazo or meropenem.
  2. In the persistently altered or obtunded patient on cefepime, consider its discontinuation and/or obtaining an EEG. (Of course, EEG should probably be considered in any patient with persistent encephalopathy.)
  3. Cefepime probably is not killing great swaths of people. Probably.

 

Leave a Reply

Your email address will not be published. Required fields are marked *