HIV and ID Observations

Hey Bing AI, make a cartoon battle between cefepime and pip-tazo.

The secret to doing a great clinical trial is quite simple. Here, I’ll share it with you:

1. Come up with an important clinical question for which there’s true equipoise.
2. Choose primary and secondary endpoints that people care about.
3. Make the inclusion and exclusion criteria easy to understand and chosen so that they define a readily available and broadly applicable study population.
4. Create a system in which referrals, enrollment, and informed consent are as frictionless as possible.
5. Start the study, and collect the baseline and follow-up data.
6. Analyze and report the results.

There, piece of cake! Off you go, time to get started.

I hope your sarcasm meter is turned on because pulling off any of the above is in fact exceedingly difficult. That’s why the ACORN study comparing cefepime with piperacillin-tazobactam (pip-tazo) — presented last week at IDWeek, and simultaneously published in JAMA — is so wonderful. And it’s not just because of its clever name:  Effect of Antibiotic Choice On ReNal Outcomes.

Let’s go through the above recipe and see how they did.

Come up with an important clinical question for which there’s true equipoise. In adults hospitalized with acute infection, which antibiotic is better: cefepime or pip-tazo? Members of Team Cefepime would cite pip-tazo’s notorious association with kidney injury, especially when combined with vancomycin, and the broad anaerobe coverage, which may further disturb the microbiome.

Team Pip-Tazo would counter by reminding Team Cefepime about their drug’s neurotoxicity and say that pip-tazo’s nephrotoxicity may be artifactual — the result of inhibiting tubular secretion of creatinine, with no change in actual glomerular filtration rate (GFR) based on cystatin C measurements. Plus, Team Pip-Tazo will defend that extra anaerobic (and enterococcal) coverage as helpful for empiric therapy.

(Brief shout-out to Team Neither, made up of antibiotic stewards arguing that empiric coverage of Pseudomonas is not needed for all acutely ill people. You’re welcome.)

Choose primary and secondary endpoints that people care about. From the paper’s quick summary:

Does the choice between cefepime and piperacillin-tazobactam affect the risks of acute kidney injury or neurological dysfunction in adults hospitalized with acute infection?

That’s pretty clear. After all, the broad-spectrum coverage of these two drugs is similar, so adverse events are likely to dictate the best choice. If the study finds differences in efficacy, that would be quite interesting as secondary or exploratory endpoints. But common sense dictates that they’re probably not that different in efficacy, so setting the study up as a superiority trial on efficacy (“powering” for an endpoint of days of ICU care or death) would require a sample size so large as to make the study impractical.

Make the inclusion and exclusion criteria easy to understand, and chosen so that they define a readily available and broadly applicable study population. How about anyone in the emergency room or medical ICU for whom a clinician plans to start one of these two antibiotics? That’s pretty simple — and it must happen a zillion times a day in hospitals all over the world.

Create a system in which referrals, enrollment, and informed consent are as frictionless as possible. Here’s the true genius of the ACORN trial. After the doctor ordered either cefepime or pip-tazo in their medical center, the investigators leveraged the electronic medical record to screen automatically for critical exclusion criteria. With none found, their order triggered a prompt for the doctor to consider enrolling the patient in the trial.

During primary investigator Dr. Edward Qian’s excellent presentation at IDWeek, his slide with the image of the prompt (available on the study’s supplemental appendix) elicited gasps of amazement — at least it did from clinical trials geeks like me. This is just a brilliant way to encourage referral and enrollment. Frictionless clinical research at its best!

(Brief aside: I’m not sure about your hospital, but where I work, changes to EPIC, our electronic medical record, move forward about as fast as that slug you found under your garden’s flowerpot. What was this wizardry Dr. Qian and colleagues did to scoot this slug along?)

But what about informed consent? As discussed in detail on page 10 of the study’s appendix, it turns out informed consent to a study can be waived if a study is deemed minimal risk and the intervention is an emergency. Empiric cefepime or pip-tazo for acutely ill, hospitalized febrile adults certainly meets the criteria of standard of care (minimal risk), and the sooner antibiotics start the better (emergency). Check and check!

Start the study, and collect the baseline and follow-up data. From November 10, 2021 to October 7, 2022, 2634 participants enrolled in the study at a single center. (I told you this happens a zillion times a day — full enrollment in less than a year from just one hospital!) Patients enrolled were well-matched at baseline, with sepsis the initial diagnosis in just over half. The median duration of the initially assigned antibiotic was 3 days, but a substantial fraction received treatment for longer.

Clinicians added vancomycin in 83% of the cefepime group and 81% of the piperacillin-tazobactam group for a median of 2 days. Roughly equal proportions of both study arms had treatment modified via other antibiotic changes, including switching to the other drug (cefepime to pip-tazo in 19%, pip-tazo to cefepime in 17%), or broadening treatment by adding an aminoglycoside or switching to a carbapenem. In other words, the very common practice of antibiotic switches for critically ill patients — don’t just stand there, start meropenem! — was equally common in both arms.

Analyze and report the final results. For the primary endpoint, the highest stage of acute kidney injury or death by day 14 did not significantly differ between the cefepime group and the pip-tazo group. Even the addition of vancomycin didn’t have a differential effect. However, for the important secondary endpoint of delirium or coma, here the pip-tazo group had significantly fewer hospital days with this complication. Deaths were similar between arms, as were a bunch of other clinical and laboratory endpoints.

So Team Pip-Tazo wins, though it’s a really close call. And may the pip-tazo/vancomycin nephrotoxicity RIP, at least based on this short-term exposure.

Now no study is perfect, and the authors responsibly address the limitations in their discussion. For me, the biggest one is that as an open-label study, clinicians may have been more prone to make the diagnosis of encephalopathy with cefepime, and this turned out to be the only significant difference between the two strategies. Read the full paper and the accompanying editorial to see further discussion.

Limitations notwithstanding, the study remains an impressive feat of generating clinically relevant evidence on an important topic, and doing so quickly in a way that harmonizes with actual patient care. The novel strategy of enrolling eligible participants through the electronic medical record in such a seamless way is a model for other clinical questions in both ID and beyond.

What should we study next?

Because it can’t be as hard as doing these things on a treadmill …