Piperacillin-tazobactam is an antibiotic that a lot of people love to hate. Combine this with the conservatism that often (rightfully) surrounds drug safety, and it creates a lot of volatility.

Before going further, we should review some recent history regarding piptazo. In the early 2010's, a series of retrospective correlational studies found that patients treated with piptazo developed elevated creatinine. There was ample evidence that this didn't reflect actual nephrotoxicity, but rather might reflect pseudo-nephrotoxicity related to impaired creatinine secretion (reviewed on the blog in 2016 and 2022). In fact, there was never any actual scientific rationale for why piptazo should be nephrotoxic, or why it should be nephrotoxic only when specifically combined with vancomycin. The whole thing was a little bizarre. Regardless, this myth persisted for years until the ACORN trial finally put this issue to bed.

Incidentally, this is an illustration of the Bullshit Asymmetry Principle. Fears regarding piptazo were stirred up by some retrospective correlational trials. In order to settle the issue, a multicenter RCT was required.

A lesson that we should take from that whole episode is that antibiotic selection shouldn't be based on retrospective correlational data – especially if that data doesn't line up with other sources of data.

Well, I guess that leads us to… the newest retrospective correlational study linking piptazo to increased mortality. These authors studied a piptazo shortage that occured from 2015-2016, causing a huge shift in antibiotic use at their institution:

The authors used this as a natural experiment to evaluate whether the shift in antibiotic use affected outcomes. They found that patients treated with piptazo had a 5% higher 90-day mortality, as compared to patients treated with cefepime:

I'm not going to go through this paper in detail, but instead I'm going to highlight five reasons that this paper should not be practice-changing.

(#1) The raw data is negative

The whole point of a natural experiment is that medical practice is forced to change, and this has a notable effect on some outcome variable.

For example, Alston et al. evaluated the effect of a piptazo shortage on the rate of C. diff infection (figure below). You can see with your own eyeballs that when the rate of piptazo decreases, the rate of C. diff infection increases. This is what a positive natural experiment looks like:

Now let's get back to the current paper. What happens to mortality during the piptazo shortage?

Nothing happened! This curve is a flat line.

If you like numbers rather then graphs, then here are the unadjusted mortality numbers before, during, and after the piptazo shortage:

Crunching the raw numbers:

• Mortality during piptazo shortage = 417/2109 = 19.8%
• Mortality before & after shortage = 1137/5460 = 20.8%

There is no statistical difference here (p=0.3).

So the raw data is negative.

Yeah, I know, that's not a really fancy statistical analysis. But if the data looks positive only after you've manipulated it heavily – that's suspicious. Especially for a retrospective study, which is a landmine field full of unidentified confounding variables.

(#2) Purported mortality differences emerged only after 75 days

The mortality rate among patients treated with piptazo and cefepime only diverges at the 75- and 90-day timepoints. After 30 and 50 days, the point difference in mortality is not statistically significant (p = 0.2 after 50 days).

This is weird. Generally, a treatment would have the greatest effect early on – and that effect would be diluted over time.

For example, if piptazo were causing harm by disrupting the gut microbiome, this harm should be greatest initially. Over time, the microbiome will recover – so it wouldn't make mechanistic sense that a delayed effect would be seen.

(#3) Mortality relates to illness severity, not anaerobic activity

Look at an analysis of various different antibiotics and their relationship to 90-day mortality:

Clindamycin isn't correlated with mortality at all. Ampicillin-sulbactam correlated with a reduced risk of mortality. This isn't what we would expect if anti-anaerobic therapies were harmful.

The antibiotics that look really good here are amoxicillin and cefpodoxime – probably because their prescription is a marker for patients who are less ill.

Overall, one gets the sense that antibiotic use is largely a correlate of how sick the patient is and what their diagnosis is (the technical term for this is confounding by indication).

(#4) Inconsistency with the ACORN trial

The ACORN trial was a multicenter RCT comparing piperacillin-tazobactam to cefepime. It found no difference in mortality between the two arms.

The authors of the current study explain away this difference because the ACORN trial looked at 14-day mortality, whereas the current trial evaluated 90-day mortality. However, the ACORN trial did include 28-day mortality in the supplemental data section. Let's compare this to the 25-day mortality in the current trial:

• ACORN trial 28-d mortality: 8.2% in piptazo group vs. 8.6% in cefepime group.
• Current trial 25-d mortality: 13.3% in piptazo group vs. 12.5% in cefepime group.

So the ACORN trial was showing a trend towards reduced mortality in the piptazo group at the 28-day mark.

The ACORN trial didn't publish data regarding mortality beyond that. But do we really think that the two mortality curves would flip-flop such that there would be a much, much higher mortality in the piptazo group at 90 days? I deeply doubt that.

(#5) Inconsistency with other RCT-level evidence

Piptazo isn't a new drug. If it actually led to a 5% greater mortality rate than cefepime, someone probably would have figured it out by now.

The largest source of RCT-type evidence we have involving head-to-head trials between cefepime and piptazo seems to be among patients with neutropenic fever. A Cochrane Database Systematic Review found that piptazo actually caused a lower mortality, whereas cefepime led to a higher mortality:

The primary result of the Cochrane analysis evaluates all-cause mortality among RCTs that compared cefepime to various other antibiotics (including piptazo). Cefepime increased mortality across a variety of different RCTs:

• There are reasons to doubt the conclusions of this study: no differences in the crude mortality rate over time, confounding by indication, and the late emergence of mortality differences in adjusted analyses (only after 75 days).
• The current study suggests that piptazo causes a substantial increase in mortality. However, this is contradicted by best available RCT-level evidence (both the ACORN trial and a Cochrane meta-analysis of RCTs among neutropenic fever patients).
• Piptazo isn't a new drug – its safety and efficacy have been established over decades. At this point, changing practice on the basis of a single retrospective study would be foolish.

ADDENDUM: Subsequent to posting this blog, a paper was posted on medRxiv evaluating this further by Fergus Hamilton, Todd C Lee, and Guillaume Butler-Laporte here. This study uses simulation analysis to demonstrate precisely how the JAMA paper may have been confounded.

Opening image credit: Photo by Faris Mohammed on Unsplash

• Author
• Recent Posts

Social Me

Josh Farkas

Josh is the creator of PulmCrit.org. He is an associate professor of Pulmonary and Critical Care Medicine at the University of Vermont.

Social Me

Latest posts by Josh Farkas (see all)

• PulmCrit: ADAPT and SCREEN trials are full of sound and fury, signifying little - December 13, 2024
• PulmCrit: How to quickly create a useful professional account in BlueSky - November 28, 2024
• PulmCrit Wee: Why MedTwitter should move to Bluesky - November 15, 2024