If you have a heart, we have news for you.
On April 4, the Resuscitation Outcomes Consortium published the outcome of the Amio-Lido-Placebo Study in the New England Journal of Medicine. The results are another important step forward in the world of cardiac arrest response – and that means any one of us could be affected by them.
To find out more about what this means for our Medic One team (and for the average Joe), we sat down with Dr. Peter Kudenchuk, Professor of Medicine at the University of Washington and Medical Program Director for King County Medic One here at Public Health and lead author of the NEJM article on this subject.
PHI: What answers were you seeking from this study? What does “Amio-Lido-Placebo” mean?
Ultimately, we wanted to figure out the best way to treat cardiac arrest caused by ventricular fibrillation when the rhythm wasn’t responsive to shock treatment. For a very long time, cardiac arrest responders have used a couple of different drugs believed to regulate the heart’s rhythm. But, these drugs (lidocaine and amiodarone), had not been rigorously tested to see which one works better – or if they even work at all.
The study was called the Amiodarone-Lidocaine- or Placebo Study (ALPS for short) because it was a randomized controlled trial comparing both drugs to each other and to a placebo.
PHI: Okay, now we have to know. Which one worked best?
Well, as you might expect, it’s complicated.
In the immediate minutes after paramedics gave the study treatment in the field, there seemed to be different effects from active drugs compared to the placebo. Recipients of either amiodarone or lidocaine, as compared with placebo, required fewer shocks, fewer “hits” of study drug, and less need for ancillary rhythm control medications (like procainamide or magnesium). Active drug recipients were also most likely to be admitted alive to hospital, less likely to re-arrest in hospital and among amiodarone recipients, less likely to require additional rhythm controlling drugs in the hospital. All this speaks to immediate and potentially important antiarrhythmic benefits from these drugs.
PHI: What about the final outcome – getting folks with cardiac arrest out of the hospital and back home?
The primary endpoint of the trial was survival to hospital discharge. There are two ways you can look at this outcome. One is from the perspective of the overall group of patients who were treated, and the types of patients who comprised this overall group.
Two groups of patients were pre-specified by the study as likely to respond differently to treatment: those with a witnessed cardiac arrest and those with an unwitnessed arrest. When it was originally designed, the study predicted that because patients with witnessed cardiac arrest are recognized and treated sooner, they would more likely be responsive to effective treatments than unwitnessed arrests. When first discovered, patients with an unwitnessed arrest are more likely to have already sustained irreversible organ damage resulting from a longer “down time” and less likely to respond to any treatment. This is precisely what was seen in the study – a statistically significant 5% improvement in survival to hospital discharge in witnessed arrests, and no effect from the drugs in unwitnessed arrests.
This suggests that the biggest bang for the buck with antiarrhythmic drugs is when treatment is administered earlier rather than later. Moreover, when these two groups were combined for the overall results of the study, the survival benefit from amiodarone or lidocaine in witnessed arrests was probably muted by the absence of any effect in those with an unwitnessed arrest. Thus the overall outcome in this combined group of witnessed and unwitnessed cardiac arrests was a statistical trend that favored drug treatment (a 3% improvement in absolute survival to discharge) but now fell just short of statistical significance.
In my view, the trial proved that both amiodarone and lidocaine are effective in improving survival. But the reality is while we want everyone with cardiac arrest afforded the chance to be saved, not everyone can or will be saved no matter how effective a treatment might be. This is why it’s so important to not only look at the treatment, but by whom and when it was received.
PHI: Why was finding this answer so important?
Well, for starters, now we have a stronger evidence base for this critical work we do. As you’ve mentioned on this blog, the best place to have a cardiac arrest is in King County. We’re successful because everything we do is thoroughly tested. This study addressed a significant gap in knowledge and practice going back 70 years.
We went into this study knowing that the outcome would either change the game of cardiac arrest treatment or bolster the science behind what we’ve been doing.
PHI: How did the study work?
This large, multi-center study was the first of its kind. In 10 sites around North America (including here in King County), patients experiencing cardiac arrest were entered into the study. Responders delivered up to three syringes of medication – either amiodarone, lidocaine or a placebo of saline solution – and recorded results. The trial was blinded to everyone – meaning the study medication was labeled with a number that was entered into a database, and its identity was only disclosed after the trial was completed.
Because cardiac arrest happens without warning and consent can’t be given ahead of time, patients were unable to give their permission to take part in this study. This required multiple levels of regulatory review and scrutiny of the study to make sure it was conducted with complete safety. For example, an independently appointed Data Safety and Monitoring Board monitored the trial as it was ongoing and had the authority to stop or change how the trial was conducted at any time if any safety issues were seen. No such problems occurred. In addition, the trial required approved by Institutional Review Boards in every community where it was performed.
PHI: What’s next?
I’m proud to be part of a team that values constant evaluation and assessment. We have a long history of research participation, and we continue to strive to improve cardiac arrest outcomes using lessons learned from this and other studies currently being conducted around the world.
The results of the study are clear, there was no significant differences in the results. None.
What Dr. Kudenchuk is stating is very misleading.
Dr. Kudenchuk states, “This is precisely what was seen in the study – a statistically significant 5% improvement in survival to hospital discharge in witnessed arrests, and no effect from the drugs in unwitnessed arrests.”
Dr. Kudenchuk is referring to a secondary endpoint that is not statistically significant. The published paper specifies why this is not significant, so Dr. Kudenchuk should know that he is wrong.
Here is the relevant section from the paper.
“We observed an interaction of treatment with the witnessed status of out-of-hospital cardiac arrest, which is often taken as a surrogate for early recognition of cardiac arrest, a short interval between the patient’s collapse from cardiac arrest and the initiation of treatment, and a greater likelihood of therapeutic responsiveness. Though prespecified, this subgroup analysis was performed in the context of an insignificant difference for the overall analysis, and the P value for heterogeneity in this subgroup analysis was not adjusted for the number of subgroup comparisons. Nonetheless, the suggestion that survival was improved by drug treatment in patients with witnessed out-of-hospital cardiac arrest, without evidence of harm in those with unwitnessed arrest, merits thoughtful consideration.”
The secondary outcomes (results in patients with witnessed arrests) “merits thoughtful consideration.”
That is not statistically significant.
Dr. Kudenchuk should know better.
Dr. Kudenchuk stop misleading people about the results of the study.
This is not surprising after Dr. Kudenchuk and the Resuscitation Outcomes Consortium decided to study which type of ventilation is best in cardiac arrest without any evidence that any ventilation is safe or effective in cardiac arrest. There is no evidence that ventilation improves outcomes for adult cardiac arrest of cardiac origin.
.
The following response is from Dr. Kudenchuk:
There will always be those who look at a half-filled glass as half-empty rather than half-full. In ALPS there was an approximate 3% absolute improvement in survival to hospital discharge favoring antiarrhythmic drugs. Overall, these findings closely approached but did not quite meet statistical significance (p=0.08 for amiodarone which narrowly missed the official p≤0.05). For the glass half-empty folks the story ends here. Abandon ship on drugs! Ignore all other findings! However, when the numbers are this close even statisticians recognize that a conclusion doesn’t automatically become “true” on one side of the p=0.05 divide and “false” on the other. To do so commits what’s called “statistical tunnel vision” – where all a person can see is see a single p value and stops thinking about the big picture.
Before ALPS began, it was recognized that no therapy will save every patient from cardiac arrest. This includes defibrillation and it certainly includes drugs as well. In particular, because of their often prolonged down-time, patients with unwitnessed arrest are less likely to respond to any therapy regardless of its potential effectiveness. Rather, effective therapies are going to be most effective in those with still enough life left in them to respond – whether that be defibrillation, or when defibrillation fails, antiarrhythmic drug therapy.
Patients with a bystander-witnessed arrest are recognized sooner, treated sooner and more likely to be responsive to treatment because they’re often still more alive than dead when treatment is started. In short, they have enough life left in them to respond to effective treatments. So before ALPS even began, the study investigators predicted that this group would be the one to more likely show benefit. This was not a secondary endpoint. Rather this was a pre-specified analysis of the primary endpoint in this group of patients that was planned long before the trial even began. It was based on both common sense and previous studies suggesting that if cardiac arrest were to be divided into two camps, witnessed and unwitnessed, witnessed arrests are the ones more likely to benefit from effective treatments.
As predicted, patients with bystander-witnessed cardiac arrest (numbering more than 1900 in all) were more responsive to treatment, and achieved a statistically significant 5% absolute improvement in survival to hospital discharge (p≤ 0.05). By comparison, unwitnessed arrests (many of whom were already arguably beyond help from any treatment) had no response to drugs whatsoever. When the two groups were combined, and the 5% improvement in witnessed arrests (p≤ 0.05) was partially diluted out by the complete absence of any response in unwitnessed arrests, the net yield was about a 3% benefit (p=0.08). Other factors could also have played a role in the smaller net benefit in the total group of treated patients. While acknowledging that this degree of improvement in the overall population of treated patients was not in itself statistically significant, the New England Journal of Medicine article still recognized the life-saving potential of the drugs with these words: “Nonetheless, the suggestion that survival was improved by drug treatment in patients witnessed out-of-hospital cardiac arrest, without evidence of harm in those with unwitnessed arrest, merits thoughtful consideration.” The independently-written editorial by outside experts that accompanied the ALPS study in the New England Journal of Medicine also supports this view and is worth a read.
So for those seeing the glass as “half empty,” and fixated on the overall p value of 0.08, it’s permissible statistically to conclude that antiarrhythmic drugs should be abandoned entirely in shock-refractory cardiac arrest. But for those seeing the glass as “half-full” – there’s the bigger picture of what ALPS actually proved clinically: Antiarrhythmic drugs resulted in the need for fewer shocks to stabilize rhythm, less need for other rhythm control drugs both before and after hospitalization, a higher rate of admission alive to hospital, a lower risk for re-arrest in the hospital, and among witnessed cardiac arrests a higher survival to hospital discharge –each and every one of which was statistically significant and coupled with few adverse drug effects and no evidence of major harm. So glass half-empty, or half-full? If it’s my life or yours (and the more than 1800 additional persons these drugs could potentially save each year from shock-refractory cardiac arrest), I’d choose carefully.
Dr. Kudenchuk,
First a question. The intention to treat group included 22.8% of the patients because of “Non-shockable initial rhythm.” Does that mean that they were initially in a non-shockable rhythm and were treated appropriately for the non-shockable rhythm and the rhythm later changed to shockable? Or does that mean that they were inappropriately treated as if they were in shockable rhythms, but analysis of the rhythms showed that these rhythms were non-shockable. I could not tell from the documentation.
As for the rest, here is the short answer, followed by a more detailed answer.
Our patients lives are very important. Too much of what we do is based on bad evidence. Our patients deserve better research. None of the treatments, no matter how you describe them, showed improvement in neurologically intact survival to discharge.
–
–
“Overall, these findings closely approached but did not quite meet statistical significance”
You designed the study (ALPS – Amiodarone, Lidocaine, or Placebo in Out-of-Hospital Cardiac Arrest) with the expectation that the study size would easily demonstrate that the drugs work.
You were overoptimistic.
I think that your interpretation of the subgroups is a repetition of this overoptimism.
–
“This was not a secondary endpoint. Rather this was a pre-specified analysis of the primary endpoint in this group of patients that was planned long before the trial even began.”
This was certainly not the primary endpoint, since you only get one primary endpoint in a study. The primary endpoint was clear – no benefit. Even the subgroup analysis shows no benefit of anything that matters.
The subgroup endpoint for witnessed arrest survival does not examine the neurological function of the patients.
Is neurological function not important for these patients?
–
“As predicted, patients with bystander-witnessed cardiac arrest (numbering more than 1900 in all) were more responsive to treatment, and achieved a statistically significant 5% absolute improvement in survival to hospital discharge (p≤ 0.05).”
When it comes to significant, I am only concerned with neurologically intact survival. I think patients, and their family members, are interested in neurologically intact survival.
Survival to discharge in a coma, only to die in the next few months, never having awoken, is not meaningful survival. How many families are paying a fortune for the medical bills for this lack of benefit? Are the glasses of these families half-full?
–
In the ALPS paper, you admit that you did not adjust your calculation of the subgroup p values, which I quoted in my initial comment, but will repeat here.
“Though prespecified, this subgroup analysis was performed in the context of an insignificant difference for the overall analysis, and the P value for heterogeneity in this subgroup analysis was not adjusted for the number of subgroup comparisons. Nonetheless, the suggestion that survival was improved by drug treatment in patients with witnessed out-of-hospital cardiac arrest, without evidence of harm in those with unwitnessed arrest, merits thoughtful consideration.”
“merits thoughtful consideration.” is not even close to “statistically significant improvement in outcomes that matter.”
Of the patients with improved survival to discharge, how many of the patients are capable of “thoughtful consideration”?
Was there any detectable difference between groups in neurologic function?
Was the study large enough to answer these questions?
–
“The independently-written editorial by outside experts that accompanied the ALPS study in the New England Journal of Medicine also supports this view and is worth a read.”
I have read that editorial several times. It does not appear to state what you think it does, but the same is true of your ALPS paper.
The editorial states – “The data do not support the use of amiodarone or lidocaine for all patients, but, although they are not absolutely conclusive, the data suggest that EMS personnel should consider these agents when the arrest is witnessed.”
http://www.nejm.org/doi/full/10.1056/NEJMe1602790
The data “suggest” that EMS personnel should “consider” these agents when the arrest is witnessed.
First, “consider” is all that the current ACLS guidelines “suggest.”
“2015 Recommendations—Updated
Amiodarone may be considered for VF/pVT that is unresponsive to CPR, defibrillation, and a vasopressor therapy (Class IIb, LOE B-R).
Lidocaine may be considered as an alternative to amiodarone for VF/pVT that is unresponsive to CPR, defibrillation, and vasopressor therapy (Class IIb, LOE C-LD).”
http://circ.ahajournals.org/content/132/18_suppl_2/S444.full#sec-43
The authors of the editorial are stating that we should limit the “consideration” of amiodarone and lidocaine to witnessed arrests. This is a recognition that your study did demonstrate a lack of effectiveness of amiodarone and lidocaine for unwitnessed arrests, but these authors appear to recognize that your study does not demonstrate a clear benefit from amiodarone or lidocaine.
If there is a clear benefit, why do the authors only “suggest” that EMS merely “consider” the use of the drugs? I apologize for the redundancy, but you appear to be missing their point.
Why do the authors “suggest” that EMS “consider” the use of the drugs less often than in the current guidelines?
Don’t even think about amiodarone or lidocaine for unwitnessed arrests, but if you tend to overestimate the effects of your treatments, you can think about drugs for witnessed arrests.
–
“Though prespecified, this subgroup analysis was performed in the context of an insignificant difference for the overall analysis, and the P value for heterogeneity in this subgroup analysis was not adjusted for the number of subgroup comparisons.”
This is from your paper. As you can see, it admits that there were no adjustments made to the subgroups to determine what p value would be statistically significant. This is a recognition that adjusting the p value of subgroups is the right thing to do. Less that 0.05 does not automatically mean statistically significant. There is nothing magical about less than 0.05, which is a less than 5% chance that your results are accidental. This is why you will have so many people remind you that one study should rarely change the way we treat patients.
This is explained in “The Handbook of Biological Statistics” –
http://www.biostathandbook.com/multiplecomparisons.html
We need to protect patients from both type I errors and type II errors – what you apparently see as just different opinions about the same level of liquid in a glass.
If you want to claim that lidocaine and amiodarone improve survival to discharge in witnessed cardiac arrests, but that only amiodarone improves survival to discharge from EMS witnessed cardiac arrest, set up the study to test that hypothesis and calculate the p value that is necessary. However, survival to discharge is not an important outcome. Neurologically intact survival to discharge is the important outcome.
Less than 0.05 does not necessarily mean statistically significant. There are even several homeopathy papers that show p values of less than 0.05, but homeopathy is still just another placebo.
–
As the ALPS paper states, “Nonetheless, the suggestion that survival was improved by drug treatment in patients witnessed out-of-hospital cardiac arrest, without evidence of harm in those with unwitnessed arrest, merits thoughtful consideration.”
Many things merit thoughtful consideration, but that does not mean that we should subject patients to them without valid evidence of efficacy and safety.
An important part of study design is setting the study up with the ability to demonstrate a statistically significant result, if the treatment truly is safe and efficacious.
Why did only amiodarone appear to improve EMS witnessed survival to discharge, while both lidocaine and amiodarone appeared to improve witnessed survival to discharge? And, if we are going to be encouraged to exaggerate what we learn from inadequate data, why does lidocaine appear to have no effect for EMS witnessed arrests, but appear to have slightly more of an effect than amiodarone appears to have for witnessed cardiac arrests?
That apparent increased effect of lidocaine vs. amiodarone for witnessed arrest is insignificant. For the EMS witnessed arrests “these findings closely approached but did not quite meet statistical significance (p=0.08” – I would need to change that to p=0.09, but as long as we are blurring the lines, why stop the blurring at p=0.08. These are merely details that can easily be blurred and this seems to be about the interpretation of the level of water in a glass, not about what we do to patients.
Our patients lives are of the utmost importance. Too much of what we do is based on bad evidence. Our patients deserve better research.
None of the treatments, no matter how you describe them, showed improvement in neurologically intact survival to discharge. Can you show otherwise?
.