There’s a new RCT out in NEJM on amiodarone and lidocaine in cardiac arrest. It’s an interesting study we wrote on, but needed a less categorical take. In the unselected study population, amiodarone and lidocaine did little for the patients. In selected patients, it might help a little more. We’ll look at it in more detail.
The study’s analysis included over 3000 adult out-of-hospital recurrent or refractory VF/VT arrests treated by 55 EMS systems in US & Canada between 2012 and 2015. The patients were randomised to amiodarone, lidocaine or saline in a 1:1:1 fashion. The groups look well balanced and most numbers come out the same.
Recurrent or refractory VT/VF are exactly the patients that are the main targets for amiodarone in cardiac arrest. Too bad it didn’t seem to work as well as many hoped. The design of the study of a predefined effect size of 6.5% to generate the patient group size was probably too optimistic. But a smaller effect size might also very well be interesting.
One big problem with drugs in cardiac arrest is to get them to their site of action in this extreme low flow state. In the case of amiodarone and lidocaine, we want to get them to the heart. There has been many concerns with drugs in cardiac arrest, and atropine is out of the loop, adrenaline is still hanging in there for old time’s sake (and we might find better ways to use it than giving a megadose of 1mg every 3 minutes).
The results overall look pretty good in the study. Survival to discharge was just above 20%, with a modified Rankin of 3 or better in a bit under 20% of patients. That’s pretty good data on out-of-hospital prolonged arrests. And the different arms in the RCT look well balanced.
There’s a tendency toward better survival with the anti-arrythmic drugs of 2.2% in the amiodarone arm, but not reaching the predefined significance of 6.5% the trial was designed for. So maybe a small effect leading to a small advantage in survival in the unselected main pool – but the study finds no statistically significant effect.
One concern is that in cardiac arrest RCT studies, a small effect might be diminished because most of the included patients cannot be saved anyway. By selecting patients more likely to survive, we might hope to unmask the benefit of the intervention. In this case amiodarone and lidocaine. So, we turn to sub-analysis.
Subanalysis is a big box of pain. Even in this study, where the subgroups have been chosen apriori. They’re listed in the appendix. The more subgroups you have, the bigger the risk of a chance finding in your numbers. Due to pure statistical risk, background mismatch in the different arms even though the original main group looks well matched, and also because your subgroups often get very small, and chance difference in numbers can make a big, but not real, difference.
Let’s look at some of the groups where you might expect better survival and decent CPR circualtion, this should increase survival overall, and in theory also bring out the effect of amiodarone and lidocaine that might be hidden in the main data.
Witnessed arrest: This is a group you would expect to have better survival than the non-witnessed arrests, especially if trained health personnel witness it, as they get cracking straight away. And indeed, EMS witnessed arrest apparently show a huge win for amiodarone with a survival of 38% vs 16% in the placebo arm, and the p-value is nice and small, suggesting it is valid. But the lidocaine arm comes out the same in this subgroup as in the overall group, 23% survival. And look at the placebo arm. It should also show better survival with EMS witnessed arrest. But it doesn’t. It comes out with a survival of just 16% vs 21% in the unselected main group! These strange numbers are most probably due to the small number of patients in this group. There are only an odd 50 patients in each arm in the EMS witnessed arrests, and with the low survival rates, a chance save or loss of just one patient changes the percentage of survival in that arm dramatically. Low p-value or not – we know the p can’t be trusted.
Then there’s the next witnessed arrest group: Bystander witnessed. Here we have a quite a bit higher numbers, with 600 patients in each arm. And amiodarone comes out significantly better with a survival rate close to 28%, same as lidocaine in this group. And placebo comes in close to 23%. So, here there is a difference. The group is larger, so it’s easier to trust. But it’s still subanalysis.
In the unwitnessed arrests, as expected, there isn’t much difference. Although they were still in VF/VT at the time of inclusion, they show the importance of early CPR on survival. Their overall survival rate is around 16%, anti-arrhythmics or not.
Bystander CPR: This is also a group where you’d think survival might be better, and the anti-arrhythmic drugs could get a chance to shine. And they do a little bit better than placebo. Amiodarone survival 29%, placebo 25% – and lidocaine 26%. In the group with no bystander CPR, as expected, turns down survival to around 16%. But lidocain gets a surprise jump up to 20%!?! What’s going on here? I’d say a chance of numbers. The pitfall of subanalysis. Even though each arm has over 350 patients.
Route of access: IO gets similar numbers for both anti-arrhythmics and placebo, but small numbers, and placebo actually comes out on top here. IV access shows a difference again, with amiodarone near 26% survival, lidocaine 25% and placebo 21%, and more partients in this group suggesting we might be able to trust it. A little bit. Maybe.
Phew. There are even more subgroups in there, but I’ve covered the most interesting ones, I think. And made the point that although there are interesting numbers there, it’s hard to trust them.
Take-home from the sub-groups
The first take-home is never trust the subgroup! Their main use is hypothesis generation. But then again, they might shed some interesting light on the study that gets hidden in the main patient pool. As shown in the paragraph above, it’s a little hit and miss. You will often end up trusting the numbers you want to believe. The ones you think make sense. And ignore the others. And believable results are important. If they make sense, it’s actually a better chance that they are true. Bayes sez.
But we need to be careful in interpreting them. Still, they shouldn’t be ignored. I’ll go over three groups I think might make sense:
Bystander witnessed arrest: Both amiodarone and lidocaine come out at around 28% survival compared to 23% in the placebo group.
Bystander CPR: Amiodarone gets a 29% survival, lidocaine 26% and placebo 25%. But the CI’s aren’t that convincing.
The numbers from bystander witnessed arrests and bystander CPR are in the same ballpark, and probably should be, but there is fluctuations between the groups – again hinting that these numbers cannot be fully trusted.
IO or IV access: An interesting number, as IO shows the same survival across all intervention arms, hinting that if the anti-arrhythmic drugs should have an effect, they’re not reaching their target. IV access shows a difference similar to the witnessed arrests and bystander CPR groups, with amiodarone at 26% lidocaine at 25% – and placebo convincingly at the base rate for the big pool: 21%.
Early CPR: What really stands out in the subanalysis is the effect of early CPR! There’s a big difference in the placebo group between witnessed and unwitnessed arrests, and even bigger between bystander CPR and no bystander CPR, going up close to 10 percentage points! So the big winner in the subanalysis game is the good old and basic CPR!
But these are all still subgroups.
The main focus in cardiac arrest is to get early and good CPR and early defib. Drugs take the back seat. That’s also what this study shows. We were a bit too categorical and provokative in our first take on this study, and were challenged by excellent comments from fan2physio and Bøtker that helped us get more in line. We very much appreciate the comments, and feel they show that peer review works – even online and post publication!
Our conclusion stands, that amiodarone and lidocaine show little real effect in an unselected pool of refractory/recurrent VF/VT patients. There is a small, but statistically insignificant survival advantage of amiodarone of 3.2% absolute increase in survival, which shrinks to 2.2% for survival with a modified Rankin score of 3 or more.
In patients with an apriori better chance of having kept their circulation going due to early and good CPR, and having a good access route to give the drug, amiodarone and lidocaine show a better effect – but remember the numbers are from subgroups, and they are fluctuating. If you believe in the effect of amiodarone, you’ll also have to buy-in on the effect of lidocaine, as their numbers are very similar in this study.
Also, is IO not as good as we think it is? From the subgroups, it seems as only IV access delivers the drugs (if you believe they work in an arrest situation). But the numbers here are small, so I’ll continue to use IO.
One big point from this study is that there seems to be no harm from either amiodarone or lidocaine. And with a patient in cardiac arrest, I would easily give a drug with no harm and a possible small effect.
Finally, across the board of the subgroups, the main difference is in patients that had the chance to get early and good CPR.
Worry about good CPR and getting the flow going. The drugs take the back seat. Add drugs once you’ve got control over your CPR flow and got hands to spare. Access and drugs should not interfere with your CPR.
Amiodarone and lidocaine in this study show no significant effect over placebo for out-of-hospital unselected refractory arrest patients. But we can speculate: Anti-arrhythmic drugs might work better in selected patients: in a good CPR setting with early CPR and a good EtCO2 good number, indicating some circulation. Here the larger subgroups suggest the drugs will work, and give your patient a slight survival advantage. And in a refractory/recurrent arrest situation, you have nothing to lose, and take all you can get.
As expected, drugs will add nothing far into a bad cardiac arrest case. This study adds to the evidence that you don’t have to give drugs before you call a bad arrest.
For cardiac arrests in general, and also in this study and all its subanalysis, focus on pushing fast and hard – and electricity. Those are the things that make a big difference.