Only the A-E-I-O-U’s to start RRT and now we know “Y”! The Start-AKI Trial

In residency and med school we are always taught the indications for emergent CRRT as the mnemonic: AEIOU

Acidosis
metabolic acidosis with a pH < 7.1 Electrolytes hyperkalemia > 6.5 mEq/L refractory to treatment or rapidly rising levels in potassium
Ingestions
with dialyzable drug, including salicylates, lithium, isopropanol, methanol, and ethylene glycol (SLIME)
Overload
volume overload that does not respond to diuresis
especially with increased oxygen requirements
Uremia
causing: uremic bleeding, encephalopathy, pericarditis, and neuropathy

However, in the ICU there has always been a question of how early should we start RRT in the setting of oliguria and BUN not causing symptoms. Initial thoughts were that earlier is better. Today’s infographic focuses on the latest RCT to determine the timing of CRRT. The Bottom Line here is that for the START-AKI TRIAL: Unless there is the emergent a-e-i-o-u’s to start CRRT doing it early does not appear to translate to a mortality benefit. There may be a signal of dialysis dependence at 90 days in this trial. Now we know “Y” we should wait until we have an indication other than the AEIOU’s.

What No COVID? Then at Least Lets Talk Steroids- in ARDS that is…

Reference: Villar. Dexamethasone treatment for the acute respiratory distress syndrome: a multicentre, randomised controlled trial. Lancet Respir Med 2020; 8: 267–76.

I know what you are thinking, “What gives? No COVID!” Well the two of you who read this are probably inundated with COVID info. So to give our brains a change I thought I would do a study that would DEFINITELY BE A HEADLINE if we weren’t COVID CRA. But just as DC gets overshadowed by Marvel so to this good (but slightly methodologically flawed) study may go unnoticed.

This study look at the utility of dexamethasone (dex) to treat MODERATE to SEVERE ARDS. I say this in bold because other trials didn’t have an entry criteria of a PaO2/FiO2 (“P/F” Ratio) of <200 (usually its worse like <150)…but this one did. Boy did they find a difference. The primary outcome was ventilator free days and they found that dex ALMOST halved this (7.5 vs 12.3)! Secondary outcomes (sloooowwww down those are hypothesis generating…quiet you, I want to be happy about something): All-cause mortality…halved!: (29 vs 50); ICU Mortality halved! (26 vs 43). Well not quite but pretty close. Now before we celebrate like its the end of social distancing* there are a TWO MAJOR FLAWS in the armor of this trial.

FLAWS

Flaw #1. Not only was it open label (the docs knew they were giving a steroid) but there WASN’T a placebo…what the what!!!! They state as follows:

“According to the ethical principles for medical research of the Declaration of Helsinki,19 the use of no placebo (no intervention) is acceptable when no proven intervention exists and when the patients who receive a placebo could be subjected to additional risks (eg, intravenous catheter-associated infections and interaction with other medications). The Spanish Agency of Drugs and Medical Devices and the referral Ethics Committee did not mandate a blinded design nor the administration of a placebo.”

And:
“Although dexamethasone was not administered in a masked manner, the risk of assessment bias is very low because one of the outcomes of interest (mortality) is objective.”

Any who has ever been a patient or spent time as learner (and we are all learners) knows NOTHING is OBJECTIVE. The decision to take someone off the vent earlier COULD absolutely have been influenced.

Flaw #2: The trial was stopped short. The trial was calculated to have 317 parties with 157 in each arm but only ended up with 139 (+1) in each.
The trial was stopped at 88% enrollment due to low enrollment. Stopping a trial short can OVERESTIMATE the effect size.

A minor flaw in this study is that they also took less sick patients than other trials. As is the mantra with research “If it didn’t work, you didn’t give it early enough”…once again this was “right”.

Two good things they did correctly is look at new infections and hyperglycemia in both groups and found no difference.

That being said there looks like a signal here. Im sure when it comes down to it the GIANT changes seen here, they won’t pan out to be as big in real life but I bet there is something to giving dex to ARDS…at least I can hope. As an ER/ICU doc its another reason for me to give steroids… and you can’t spell stERoids without ER!

*I REALLY don’t like the term social distancing/isolation…we should call it “physical distancing/isolation”. We can stay in touch, without the touching!

ESETT: A Randomized Trial of 3 AEDs for Status Epilepticus

Kapur, J et al. Randomized Trial of Three Anticonvulsants Medications for Status Epilepticus. NEJM 2019;381;2103-13

This was a randomized, double-blinded, national multicenter pragmatic study to determine comparative effectiveness of traditional anti-epileptic medications in status epilepticus. The purpose of this study is to see if there is a difference between levetiracetem (Keppra), fosphenytoin (Cerebyx), and valproate (Depakote) in treatment of status epilepticus. By definition, these are second line treatments as first line treatment remains benzodiazepines (BZD). This is a well designed trial that was developed by the NIH and FDA then conducted by Neurological Emergencies Treatment Trials (NETT) and Pediatric Emergency Care Applied Research Network (PECARN). It utilized a response adaptive design to attempt to find the “best” AED (see below). Clear primary endpoint was established of seizure resolution and improving mental status one hour after starting trial medication. The protocol was guideline based as patients received cumulative BZD doses of 10 mg diazepam, 4 mg of lorazepam, or 10 mg of midazolam or weight based for children. Leaving the discussion of adequacy of those doses for another time, patients then went on to receive a weight based, unmarked vial of medication no more than 30 minutes after last dose of BZD. Dosing for these trial drugs were based off the Established Status Epilepticus Treatment Trial (ESETT) using 60 mg/kg of Levetiracetam (big doses of >4g !), 20mg/kg fosphenytoinand 40mg/kg of valproate. Protocol deviation was marked as unmasking trial drug in less than 60 minutes. 348 adults and children were enrolled into the study with interval analysis reaching a 1:1:1 trial drug administration. The primary endpoint was reached in less than half of all patients (46%) with absolutely no difference in efficacy. No difference in secondary end points such as ICU admission, median ICU stay, median ICU stay or median time to seizure termination after drug infusion. This study was not powered for to detect the rare adverse side effects but more patients that received fosphenytoin developed hypotension.

SAFETY OUTCOMES

Safety Outcome*LevitiracetumfosphenytoinValproic Acid
Hypotension0.7%3.2%1.6%
Intubation20%26%16.8%
Composite of
Hypotension
and arrhythmia
1.3%3.2%1.6%
Adverse Outcomes among the 3 AEDs
*Not powered to detect significance

Interestingly, pseudoseizure patients (~10%) were included which is clinically practical as there are times you simply do not know. 

JC Learning point: 

This trial did not have any statistical juijitsu as the primary endpoint was clear and there was no difference in efficacy. I speculate this will lead to use of levetiracetam as the go to second line agent as there is no difference to the other traditional agents which have their own respective drawbacks such as drug level monitoring, hypotension and arrhythmias as seen in rapid administration of fosphenytoin (which can easily happen in stressful situations like status epilepticus). The bigger take away for me here is 1 hour of status epilepticus is way too long for less than half the patients to improve from status epilepticus. The drawbacks of neurological damage, rhabdo and aspiration while waiting for these medications to take effect over an hour (which occurred in less than half of these patients) would be hard to stomach. As a newer emergency medicine physician, I have a tough time being patient for 15 minutes let alone 1 hour! Not to mention the longer these interventions take to have effect the more likely patients can develop super-refractory status which does not sound like fun. Bottom line, whichever second line anti-epileptic you go with, it likely does not make much of a difference in the short term as you continue to push BZD and move onto giving more BZD, intubation and propofol vs barbitutes.

Response Adaptive design

Adaptive designs allow for the review during the trial and then “adapting” the trial to allow to change the study for such changes as:

  • Abandoning treatments or doses
  • changing the allocation ratio of patients to trial arms
  • identifying patients most likely to benefit and focusing recruitment efforts on them
  • stopping the whole trial at an early stage for success or lack of efficacy.
  • Refining the sample size

In the ESETT trial the adaptive part was that at interim analysis points the drugs were evaluated and probabilities to determine if one was better (via Bayesian methods) were reviewed. If one drug was found to be worse then that arm could be dropped and the two remaining ones would stay. Thus you would get a larger sample size for the better treatment arms. Surprisingly in all the interim analyses THIS DID NOT HAPPEN. The three drugs were all found to be equally probably by the end of the trial. 

By Sagar Dave and Jesse Shriki

We’re Back Baby!! The Quick hit article: For ESBL, Just Say NO to PIP/TAZO

Effect of Piperacillin-Tazobactam vs Meropenem on 30-Day Mortality for Patients With E coli or Klebsiella pneumoniae Bloodstream Infection and Ceftriaxone Resistance
A Randomized Clinical Trial

This was a randomized, double-blinded, international multicenter pragmatic (phew! That’s a mouthful!) study to determine if piperacillin/tazobactam (pip-taz) is non-inferior to meropenem. The reason for this study is to see if there is a way to reduce carbapenems use and prevent further Carbapenem resistance acinetobacter (CRAB, not the Maryland kind). ESBL drugs are those that are by definition resistant to ceftriaxone or ceftazidime. This is a really well done trial with the trial being registered at clinicaltrials.gov and meeting all the important qualities of an RCT. The only missing item (which is probably a small one given how rigorous this trial was) is the fact that of all the 9 countries the US wasn’t one of them (Isn’t Canada close enough? Probably, still too close if you aren’t a fan of Trump). The study used a non-inferiority protocol with 30 day all cause mortality as the end point. Also a home run here with a very good patient oriented outcome. They randomized 378 patients and found an impressive 12.3% vs 3.7% mortality rate for pip-taz vs meropenem. Yikes! That is an 8.6% absolute risk reduction!!! In classic statistical phrasing using double negative: This trial found pip-tazo not non-inferior. They powered the study to find 454 patients but stopped at a predefined interim-analysis of 340 patients. Its bad form to stop a trial for “benefit” when you are looking for non-inferiority but not for harm. As such the trial enrolled 391 pts. In this cohort the patients had about 86% E. Coli and 14% Klebsiella. Most of the patients came from Singapore and Australia with only 1 patient coming from Canada (C’mon Canada!). Most of the infections were either urinary (~60%) or intra-abdominal infections (~15%). Interestingly, 60% of patients were deemed to have been treated with “appropriate” empiric therapy and EVEN MORE INTERESTING is the fact that almost 45% of patients had community associated acquisition. This is probably true world wide but maybe less so in the US? But we are likely heading that way. So to sum up in patients that have ESBL need meropenem not pip-tazo! For a number needed to treat version: If you will have one death for every 12 patients you treat if you use pip-tazo and not meropenem!!

The JC learning point
This trial was performed as a “non-inferiority” trial. These trials should be looked at with a scrutinizing eye! Drug companies love to use this because they don’t have to prove the drug is better, they just have to prove its not worse. The critical point here is to make sure that the correct significance (α level) is set. Non-inferiority is established at the α significance level if a confidence interval for the difference in efficacies (new – current) is contained within a safety margin interval. In this study they used a margin of +5%; so if the studies showed a mortality within 5% of each other then we would have said that pip-tazo is non-inferior to meropenem (but it didn’t). Normally when we compare to things we can use a α of 0.05. However, in non-inferiority testing we have to use an α of 0.025 because we are really performing TWO ONE-sided tests (think of it as needing a p-value for the +margin and the -margin so you cut the p-value in half ). Lastly, interpreting a non-inferiority trial as a superiority trial is OK and doesn’t require a multiple testing correction IF the 95% confidence interval for the treatment benefit excludes the non-inferiority margin AND zero is not in the confidence interval. However, the opposite approach is not true. If a superiority trial fails to reject the null hypothesis but the trial data appear to suggest treatment non-inferiority, you cannot default to non-inferiority

References

  1. Effect of Piperacillin-Tazobactam vs Meropenem on 30-Day Mortality for Patients With E coli or Klebsiella pneumoniae Bloodstream Infection and Ceftriaxone Resistance A Randomized Clinical Trial. JAMA. 2018;320(10):984-994. doi:10.1001/jama.2018.12163
  2. Understanding noninferiority trials. Korean J Pediatr 2012;55(11):403-407. http://dx.doi.org/10.3345/kjp.2012.55.11.403

Quick Hit Article: Difficult vascular access: You know the drill… but do you really know the drill?

Kawano et al. Intraosseous Vascular Access Is Associated With Lower Survival and Neurologic Recovery Among Patients With Out-of-Hospital Cardiac Arrest. Ann Emerg Med. 2018;71:588-596.

THE BOTTOM LINE
This study was a secondary analysis of a large EMS trial conducted in 7 sites in the US and Canada. This is a really interesting study that looks at the association between outcome and IO access. They looked at 12,500 pts with IV access and 660 pts with IO access. The primary outcome measure was favorable neurologic outcome at hospital discharge with a mRS of 0 to 3. In the IO group 24% had rosc, 4% survived to hospital discharge, and 1.5% had a favorable neurological outcome. In the IV group 38% had ROSC, 10% survived, 8% had a favorable neurological outcomes. In the UNADJUSTED propensity matched model IO vs IV access was associated decreased probability of ROSC (25% vs 34%,OR=0.66), survival to discharge (4.5% vs 8.6%, OR=0.5), and favorable neurologic outcome (1.8% vs 5.9%, OR=0.29). The two groups differed in interventions very noticeabley in that the IO group were an almost 15% LESS therapeutic hypothermia and 9% less interventional catheterization. This is JUST AN ASSOCIATION and maybe IO is just a marker of sicker patients, HOWEVER, I think this is enough to allow for an RCT looking directly at this topic.

THE DETAILS
Difficult IV access in cardiac arrest? Just drill it! Utilize intraosseous vascular access that is. The IO has become the go to for rapid IV access in cardiac arrest patients. We can deliver drugs quickly and not muck around looking for a vein either peripherally or centrally. But are there down stream consequences? Is it REALLY better? This study takes a look at outcomes of non-traumatic cardiac arrest patients who had an IO placed.

This study was a secondary analysis of a publicly available registry conducted in 7 sites in the US and Canada. The original trial looked at the timing of the first analysis for defibrillation and application of an impedance threshold device. The original data was prospectively collected. Exclusion criteria were the typical ones including trauma, prisoners, pregnant, dnr, and hypovolemia. To the study’s benefit the obtained a cohort whose access route was the INTENDED PRIMARY CHOICE and excluded failed attempts at either route of access. The importance of this is that it MIGHT equal out the bias of a sicker patient or delays in care where a failed access attempt delays the care. The primary outcome measure was favorable neurologic outcome at hospital discharge with a mRS of 0 to 3. This is an interesting outcome change point because normally mRS is 0 to 2 is the cutoff in stroke studies. Because out-of-hospital vascular access was not randomly assigned across this cohort, they applied propensity matching to decrease selection bias and unmeasured confounder.
They found after exclusions 13,155 patients with 660 pts (5%) and 12,495 (95%) pts. Note that patients with IO access had a higher proportion of nonshockable initial rhythms, fewer public location and shorter times from call to first provider. In the IO group 24% had rosc, 3.8% survived to hospital discharge, and 1.5% had a favorable neurological outcome. In the IV group 38.3% had ROSC, 10% survived, 7.6% had a favorable neurological outcomes. They found after multiple regression with multivariate analysis that IO access was ASSOCIATED with a DECREASED probability of ROSC [OR 0.6 (CI: 0.49 to 0.74)] and DECREASED favorable neurological outcome [OR 0.24 (CI: 0.13to 0.46)].
They found 1116 patients in the propensity match. In the UNADJUSTED model IO vs IV access was associated decreased probability of ROSC (25% vs 34%,OR=0.66), survival to discharge (4.5% vs 8.6%, OR=0.5), and favorable neurologic outcome (1.8% vs 5.9%, OR=0.29). Looking at the propensity matching these groups for the
Baseline characteristics are quite well matched and markers of their propensity matching were well correlated. They did differ in some ways in the interventions. The very noticeable differences in the IO group were an almost 15% LESS therapeutic hypothermia, 5% less fibrinolytics, and 9% less interventional catheterization.

Although the propensity scoring appears quite well and there were multiple calculations to correct for confounders there are still significant biases. I think foremost the discrepancy between the number of of IO vs IV’s performed is hard to overcome (660 vs 12500). Also there were some big differences in the patients receiving hypothermia and cardiac catheterizations. I still think the data is not definitive for those two interventions but certainly this may be a marker of quality of care and attention to care. Still the OR for survival with favorable neurological outcome is impressively in favor of IV access. It is interesting that some EMS systems used IO as their first line and it would be interesting to know why IO was first line. I also wonder WHY the IO patients got less interventions. Was the IV access the reason or were they initially deemed to be sicker? Also it would be nice to know how long the IO was in and what was the time frame to when it was swapped out for IV access. Lastly, I can’t get away from the thought that sicker patients are going to get IO’s and are going to do worse.

So what’s the bottom line? IO access is critical to establishing vascular access to care for patients (depending on what you think the validity of ACLS drugs are, oh HEY OPALS TRIAL). Maybe IO access is a marker for sicker patients or maybe worse IO access biases providers against the same level of care as those with an IV…We all know the drill… but does it make our patients better? I think this is enough to allow for an RCT looking directly at this topic.