Dr. Milne, the all cause mortality result you were having trouble finding is in figure 4 of the main text, page 20. It's called "Death from any cause," so if you were doing ctrl + f to look for "all cause" or "mortality" then that wouldn't have found it.
I tend to see the study more of a positive then was highlighted in the podcast. Some thoughts below. Comments were raised on 1) On the interpretation that the 5 point primary MACE was driven by re hospitalizations and revascularizations and therefore not objective outcomes 2) On patients not being on a background of high intensity statin 3) On some harm in terms of a small increase in afib. 4) the lack of recommendation for routine use at this time.
My comments are the following. 1) Lets refer to Figure 4 of the Reduce it study. 5 point MACE as well as all of the key primary and secondary endpoints were met and with statistical significance. 5 point MACE down 25 % (p<0.00000001), 3 point HARD MACE (CV death, nonfatal MI, nonfatal stroke) down 26% (0.0000006). NNT was 21. For comparison in the CV space, lipitor is around 47 and for repatha its 67, and zetia is 50. Icosapent ethyl primary endpoint was robust in all subgroups and I strongly disagree with the interpretation that it was driven by subjective endpoints.
Furthermore, in terms of the questionable efficacy points raised, this is not the only trial showing Icosapent ethyl to be effective. The Jelis trial in Japan used the same active ingredient with their drug EPADEL (essentially pure epa as well) and found a 19 percent RRR. This was met with criticism in the USA due to limitations (that I agree with) and thus Reduce it was more so confirmatory but in an unhealthier western population. Additionally, the CHERRY trial and the more recent (few months prior) EVAPORATE trial showed statistically significant reduction in plaque via CTA with pure EPA, icosapent ethyl. That is objective evidence of mechanism. Lastly, follow ups to the reduce it trial have noted significant reductions in revascularization including a reduction in CABG, which is rarely seen. See Below
(Need for revascularization procedures: First coronary revascularization for IPE vs. placebo: 9.2% vs. 13.3% (p < 0.0001), NNT = 24. PCI: 7.7% vs. 10.9% (p < 0.001), CABG: 1.9% vs. 3.0% (p = 0.0005). Reductions were noted in both elective and urgent/emergent revascularizations. Results were consistent across tertiles of TG levels (p for interaction = 0.07). Total revascularizations were similarly reduced with IPE vs. placebo.)
2) One limitation mentioned by the podcast is some patients were not on high intensity statins (See Table 1). However, if you look at the median LDL of 74 icosapent ethyl verses 76 placebo, most physicians would agree that is well controlled LDL by most standards. Interestingly, some cardiologists found this as a strength of the study (that patients were largely optimized).
3) In terms of harm, a small increase in afib was astutely pointed out. However the most important point was that stroke (fatal or nonfatal) was down 28 percent which is arguably the most important complication of afib. Therefore, people may have afib but have less strokes and 5 point MACE including death despite the afib. Additionally, the afib that was seen was largely in those who already had afib (However this is not disclosed in this paper). Benefits seem to far outweigh the risks here but nevertheless the podcast members appropriately raise this concern so that we can discuss with our patients.
4) Another final comment was on the suggestion against the routine use of this medication at this time. It is an opinion but understand the standard of care has well adopted this medication. There's a multitude of organizations that have written statements and or updated guidelines on it including, ADA, ESC, AHA, probably over 10 of them. It has only been approved for the CV indication for a year so it is still early and will take years to disseminate information to physicians but I do believe there to be benefit with this drug.
Thanks for listening to my rant. Proud member and supporter of Emrap
Thank you for the comments and I appreciate we can each interpret the study differently.
1) I should have been clearer and said the largest HR was of an objective outcome with smaller HR seen for more subjective outcomes.
Figure 4: The reductions you provided were relative not absolute. I prefer absolute reductions rather than relative. The reasons is that relative reductions can make a small absolute difference look impressive. The 25% relative reduction in the 5 Point MACE was a 4.2% absolute. This does not mean a 4.2% reduction cannot be clinically meaningful but it should illustrate the point. The key secondary endpoint was 3.6% absolute reduction (26% relative reduction).
It could be argued that nonfatal stroke is very objective. This is because stroke mimics are a well-recognized phenomenon in the literature. Stroke mimics introduce uncertainty into the data. https://pubmed.ncbi.nlm.nih.gov/29727304/
I was not sure what point you are trying to make with the p-values and perhaps you could clarify? There has been great debate about the use of the p-value in the world of epidemiology and biostatistics. https://www.nature.com/articles/d41586-019-00857-9
All-cause mortality had the smallest magnitude for HR and the 95% CI spanned 1.0.
I’m again not sure what you are trying to say with the comparison to other treatment modalities? These provide no information as to whether or not the claim about icosapent ethyl is “true” (best point estimate of observed effect size).
My revised conclusion would be that the more subjective outcomes (hospitalizations for unstable angina, urgent or emergency revascularization, fatal or nonfatal stroke) rather than objective outcomes (death) contributed to the magnitude of the reported HR.
Thank you for bringing the JELIS 2009 trial to my attention. It too had a MACE outcome that consisted of: major coronary event, including sudden cardiac death, fatal and non-fatal myocardial infarction, and other non-fatal events including unstable angina pectoris, angioplasty, stenting, or coronary artery bypass grafting. Again, there is subjectivity to angioplasty, stenting or CABG. The RRR 19% and an ARR of 0.7%.
We record these PCMA episodes months in advance. I was not aware of the EVAPORATE trial when we recorded the January 2021 episode.
EVAPORATE was a trial n=80 showing a statistical difference in a disease-oriented outcome (DOO) not a patient-oriented outcome (POO).
CHERRY n=193 also showing a statistical difference in a DOO not a POO. We have been fooled before in medicine by DOOs and pathophysiologic explanations.
Did I make the comment about high-dose statins or was it Steve Brown? It if was Steve, I will prompt him to respond to that comment.
There was an increase reported in AF. The relative increase was around the same magnitude of the benefit (26%) but the absolute increase was 1.4%
It could be argued that the most important complication of AF is fatal and nonfatal stroke. However, I’m not as confident in the diagnosis because of the subjectivity in making that diagnosis (stroke mimics mentioned earlier). They were most likely using the new definition of stroke which includes imaging result based on pathophysiology. Do you know how many strokes were based on ischemic changes rather than functional outcome? I would suggest patients care about function more than their imaging test. This can complicate the interpretation of this type of outcome.
There are other patient-oriented outcomes with the diagnosis of stroke. People with a diagnosis of AF are more likely to be put on an antiplatelet drug or anticoagulated (DOAC/warfarin). This increases their risk of bleeding besides ICH. They are also more likely to be on a rate controlling drug (BB or CCB). This would put them at greater risk for falls. I did not see the granularity in the data presented in the trial we reviewed.
Perhaps the reason all-cause death was not statistically different was in part because the decrease in strokes was off-set by an increase in harm associated with the treatment of AF?
With regards to the “routine” statement it could come down to the definition of routine being used. https://www.merriam-webster.com/dictionary/routine
My use of “routine” was based on a review by CADTH. I could have been more specific or clearer in what I was trying to get across. Part of it was to avoid the possibility of indication creep. There is not enough compelling evidence for me at this time to be making a routine recommendation. I accept others may find the evidence compelling enough. https://www.cadth.ca/sites/default/files/cdr/complete/SR0619%20Vascepa%20-%20CDEC%20Final%20Recommendation%20July%2020,%202020_for%20posting.pdf I have not reviewed all the other guidelines you mentioned. Not a big fan of guidelines in general. They are there to guide care, not dictate care. The number of guidelines also does not necessarily make them correct. I tend to go back to the primary literature and critically appraise it myself (yes, I have a problem). There are quality issues with guidelines. The Institute of Medicine has a good resource on this called: Clinical Practice Guidelines We Can Trust. https://www.ncbi.nlm.nih.gov/books/NBK209539/
Thank you very much for your “rant”. I enjoy getting this type of feedback that challenges my positions. I freely admit my interpretation of the literature could be wrong, are they are subject to change. This has been very intellectually stimulating.
Keep on listening and sending us your thoughts. Ken
Thanks for the replies. Some really great points raised.
In terms of your comment on the reduction in strokes (both fatal and nonfatal) in the icosapent ethyl arm but your skepticism that the diagnosis was actually a stroke, I did pull the article you reference. Stroke mimics are out there, we all agree, however to discount this endpoint as merely subjective is almost cruel. The trial, as we know, was a randomized double blind placebo controlled trial that hit on all endpoints. I can see more skepticism in trials that were marginal or had a poor design etc. The efficacy is well supported. When you look at the CV space, you rarely see a therapy this conclusive. Theres no perfect medication and we only have what is available to us. However, I do agree with some reports that this may be the biggest advance to CV risk reduction since statins 20 to 30 years ago. Good to approach with skepticism (as you do best ha), but after a thorough review, continued skepticism on this medication has me a bit curious if you support another CV preventative medication other than a statin. It’s okay if you don’t because we need critical skeptics in medicine and therapies should live up to a high standard. However, that is why I brought up the other therapies with NNTs (worse) as a comparison.
We have Jelis (japanese trial with same medication but lower dose -2G instead of 4G), Reduce it (by far the strongest evidence) and two mechanistic studies to help clarify why the efficacy occurred. Cherry showed a reduction in plaque and now that finding was again demonstrated with Evaporate. Agree patient Oriented outcomes are what matters most but disease oriented outcomes do help support the validity of patient oriented outcomes when they all align as they do here. Furthermore the STRENGTH trial failed (similar CV outcome study but with a diff omega composition of less EPA but with DHA as well) likely due to lower on therapy EPA blood levels that didn’t even get near the levels in Reduce it. We can hypothesize that obtaining a high dose EPA blood level is what matters most to patients and here in Reduce it , this level correlated very closely with the MACE reductions seen in REDUCE IT, almost accounting for all of the benefit. I think measuring this EPA blood level will become standard , higher is better. TG reduction isn’t what matters as you know. Many other therapies reduce TGs better than icosapent ethyl and all of those therapies failed CV outcome trials. I think we are at the beginning of a new era in CV preventative medicine, which to me is very interesting to watch.
All cause mortality was 274 (6.7) icospaent ethyl vs 310 (7.6), HR 0.87 (CI 0.74-1.02). Trend to benefit but not significant. I don’t think we can discount this completely as the curves began separating at a year and continued to separate (continue to obtain benefit) as time went on. Perhaps a slightly longer trial would begin to show all cause mortality. Perhaps not. You theorize possibly the harm with increased afib contributed to the lack of stat sig in all cause mortality in the treatment group. Interesting thought. They did post hoc handicap icospaent ethyl by including afib/serious bleeding as endpoints and the results still showed a statistical sig reduction in outcomes. If we want to discount this bc it is post hoc I am completely fine with that; just food for thought. See Table 15 https://www.fda.gov/media/132479/download
I agree with your points with not blindly following guidelines and appraising the lit yourself. Good point for all listeners. At the end of the day every therapy has a risk/benefit. I can be wrong on this and respect your opinion tremendously that is why I commented. I appreciate the criticism. Only way to fine tune my thought process.
Any specific comments on the use of relative vs absolute risk reduction or clarification on the use of p-values?
I could be more careful not to dichotomize the outcomes into objective v subjective. We can all agree that death is objective outcome. This had the smallest HR magnitude, and the 95% CI spanned the line of no difference. The other outcomes like stroke had greater magnitude of HRs and were more subjective. Or to say it another way, less objective (compared to death).
They change in definition of stroke to an imaging finding and not a functional outcome also is important. This change in 2009 makes strokes look better by included 30-50% of cases that would previously be diagnosed as a TIA. This is part of the evidence that increase my uncertainty about the magnitude and precision of this point estimate. https://www.ahajournals.org/doi/pdf/10.1161/STROKEAHA.108.192218 https://pubmed.ncbi.nlm.nih.gov/12444191/
If 1/3 to ½ of the stroke reduction in this trial are based on imaging and not functional outcome that would be a less patient-oriented outcome. I do not view this as being “almost cruel” but rather a reasonable concern.
Yes, this was a randomize-double blinded placebo controlled trial. I do have more skepticism for lower methodology. However, my skeptical radar does go off seeing this was a funded study by the maker of the drug and there was over ½ page of small print conflicts of interest. My increased skepticism is based on this SRMA published in Cochrane on RCT with COIs (Lundh et al 2017).
• Authors’ Conclusion: Sponsorship of drug and device studies by the manufacturing company leads to more favorable efficacy results and conclusions than sponsorship by other sources. Our analyses suggest the existence of an industry bias that cannot be explained by standard 'Risk of bias' assessments. https://pubmed.ncbi.nlm.nih.gov/28207928/
I agree there is no “perfect medication” and that is not my expectation. This would be a logical fallacy (Nirvana Fallacy). The available evidence is not strong enough at this time to recommend routine use of this medication but rather only for a narrow indication and after a discussion with the patient in a shared decision approach. https://www.logicallyfallacious.com/logicalfallacies/Nirvana-Fallacy
My healthy scientific skepticism is applied to each medical claim. The burned of proof is on those making the claim. We all will have different levels/amounts/quality of evidence to accept a claim. If the burden is met, we should reject the null hypothesis and accept the claim. If the burden is not met, we should accept the null hypothesis. My positions for any medical claim are tentative, subject to change and confidence is proportional to the strength of the available evidence.
The NNT can be another way to express data. There are some limitations to this metric which we discussed in a previous SGEM Xtra episode. https://www.thesgem.com/2019/12/sgem-xtra-nnt-wet-or-dri/
A good friend of mine Dr. Anthony Crocco put together a great video that explains some of the issues with the NNT/NNH. https://www.youtube.com/watch?v=SdtNJeB2i60 The magnitude of the NNT for other therapies does not provide information as to whether this therapy has a net patient-oriented outcome (potential benefits v potential harm). It would be good to remember when balancing out these two issues that the potential harms are systematically under-reported in RCTs (Hodkinson et al 2013). https://bmjopen.bmj.com/content/3/9/e003436
I am not discounting that most of the 95% CI favours treatment. With a small effect size and a wide 95% CI that spans 1.0 it decreases my confidence in this medication. I would encourage increased skepticism on post-hoc analyses. It does not mean it is incorrect but I’m more certain about a priori analyses than post-hoc ones.
Thanks for listening to PCMA, engaging with these very interesting and important comments/questions.
Some more good points raised. I do think we are too far apart in our thoughts here to come to some perfect agreement. I am most intrigued not by the medication or trial we are discussing but the ability to interpret this as a no go in your podcast. It's almost impressive to me at how skeptical you are and probably you are right since you have been reviewing papers for quite some time. That is why I felt encouraged to comment on the thread in the first place. For me, this was ground breaking so I enjoyed a good discussion on your thoughts and to challenge mine. It seems you have a very high evidence standard here. I did ask to see if you endorse any other CV risk reduction medications. I am trying to appreciate a trial as robust or more in this space aside from statins. As the leading cause of death in the country, heart disease could use a medication to reduce MACE events as seen here on top of maximally tolerated statins. Please note I am no way trying to be rude as it may come across like this during an online comment section. This is just an EM doc with a niche in cardio enjoying the discussion. I have heard a really nice perspective from you and respect it and will apply some of your thoughts to other drugs in the future.
To respond to P value/RRR - I agree with the limitations of those that you raise. It is sometimes the norm to which articles report to boost the looks of the numbers but we both can see through that and value the data for ourselves. The p value article you referenced in my mind worked counter to the argument being raised. It seems that with borderline p values we should not dichotomize stat significant or not but rather look at the trends. It almost looks like that article favors looking for benefit amongst NON stat significant trials (maybe even accepting all cause mortality as a benefit since it trended that way? ha). This, on the other other hand, was highly statistically significant and I think we would both agree a p value to the multiple decimal places (as is here) is certainly better than one merely P<.049. Did I interpret this correctly?
-The stroke points you raise about it being a less objective outcome are thought provoking. It almost seems everything other than death is a less objective outcome. Silent stroke (imaging) is not a benign entity; However I do enjoy your points about skepticism into the diagnosis of stroke.
I think we do have more common ground and I do not expect perfect agreement. My position is not a “no go” but rather not to “routinely” prescribe this medication. There are narrow indications and even then, it would need to be discussed with the patient in a shared decision model.
My skepticism is informed by conducting medical research for ~37 years, practicing medicine for ~25 years and teaching critical appraisal for ~15 years. This does not make any of my positions correct but this is my area of expertise. We (myself included) have been fooled over the years. This article by Dr. Ioannidis also informs my position. https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.0020124 I do understand that he is a controversial figure lately.
Dr. Jerome Hoffman is one of my mentors. He taught me to consider just standing there as a therapeutic option. The value and art of deliberate clinical inertia is described in this article. https://onlinelibrary.wiley.com/doi/abs/10.1111/1742-6723.12922
I do have high standards. Patients deserve the best care, based on the best evidence. The level of evidence required is also proportional to the claim. Extraordinary claims require extraordinary evidence (Carl Sagan). So, when the claim that vitamin C cocktail cured sepsis I was very skeptical. https://www.thesgem.com/2017/04/sgem174-dont-believe-the-hype-vitamin-c-cocktail-for-sepsis/
I do support other CV risk reduction strategies. Each claim must stand on its own and my confidence will be proportional to the quality of the evidence. This includes the magnitude of effect size and precision around the point estimate. The goal is to try and find the least biased information There are a number of other quality checks lists which can help probe the literature for its validity. https://www.thesgem.com/2014/03/make-it-so-beem-appraisal-tools/
An example of CV risk reduction would be ASA for acute STEMI to decrease mortality. https://www.thennt.com/nnt/aspirin-for-major-heart-attack/
None of this has come across as rude. I agree tone can be hard in this form of communication. I have very much enjoyed this respectful dialog.
Yes, I tend to have more confidence in a low p-value than that of a higher p-value. However, it should not be viewed in isolation and I’m not suggesting you have asserted that point. It is just another data point that can be considered and interpreted. It has been dichotomized over the years and misused. A p-value lacks information on effect size, precision, clinically importance, cost, etc.
Glad you like the points about stroke. There is a range between objective and subjective. Most people I think would agree that death is a very objective outcome. There is still some controversy even around this idea. But even this statistic needs to be interpreted. There can be worse things than death. Survival with severe neurologic damage. Some people may prefer death over a very poor quality of life. An example of this interpretation would be the PARAMEDIC2 trial. Yes, they did have more survival with epinephrine in OHCA but it was mainly due to more people surviving. Severe neurologic impairment (mRS 4 or 5) was more common among survivors in the epinephrine groups compared to the placebo group (31.0% vs. 17.8%). https://www.thesgem.com/2018/12/sgem238-the-epi-dont-work-for-ohca/
Silent strokes may or may not be benign. How many people right now have a silent ischemia that would show up on MRI but are functionally perfectly and will die of something else? If you screened all those over 65yo looking for signs of silent ischemia how many would you find. Would knowing be a good thing for the person’s overall mental health? Would such knowledge change their behaviour for the better or worse? Would any intervention to treat the silent ischemia have a net benefit or net harm? As an example, stenting patients with stable CAD. https://www.thennt.com/nnt/stents-stable-coronary-artery-disease/
The point is it can be very complicated and there are always more layers. It is good to have these conversations to consider other valid points of views.
To join the conversation, you need to subscribe.
Sign up today for full access to all episodes and to join the conversation.
To earn CME for this chapter, you need to subscribe.
Sign up today for full access to all episodes and earn CME.
Andrew J. - January 12, 2021 8:55 AM
Dr. Milne, the all cause mortality result you were having trouble finding is in figure 4 of the main text, page 20. It's called "Death from any cause," so if you were doing ctrl + f to look for "all cause" or "mortality" then that wouldn't have found it.
Ken M. - January 12, 2021 10:18 AM
Andrew: Great to have eagle eye audience helping us find some of these details.
Joseph B. - February 6, 2021 7:28 AM
I tend to see the study more of a positive then was highlighted in the podcast. Some thoughts below.
Comments were raised on
1) On the interpretation that the 5 point primary MACE was driven by re hospitalizations and revascularizations and therefore not objective outcomes
2) On patients not being on a background of high intensity statin
3) On some harm in terms of a small increase in afib.
4) the lack of recommendation for routine use at this time.
My comments are the following.
1) Lets refer to Figure 4 of the Reduce it study. 5 point MACE as well as all of the key primary and secondary endpoints were met and with statistical significance. 5 point MACE down 25 % (p<0.00000001), 3 point HARD MACE (CV death, nonfatal MI, nonfatal stroke) down 26% (0.0000006). NNT was 21. For comparison in the CV space, lipitor is around 47 and for repatha its 67, and zetia is 50. Icosapent ethyl primary endpoint was robust in all subgroups and I strongly disagree with the interpretation that it was driven by subjective endpoints.
Furthermore, in terms of the questionable efficacy points raised, this is not the only trial showing Icosapent ethyl to be effective. The Jelis trial in Japan used the same active ingredient with their drug EPADEL (essentially pure epa as well) and found a 19 percent RRR. This was met with criticism in the USA due to limitations (that I agree with) and thus Reduce it was more so confirmatory but in an unhealthier western population. Additionally, the CHERRY trial and the more recent (few months prior) EVAPORATE trial showed statistically significant reduction in plaque via CTA with pure EPA, icosapent ethyl. That is objective evidence of mechanism. Lastly, follow ups to the reduce it trial have noted significant reductions in revascularization including a reduction in CABG, which is rarely seen. See Below
(Need for revascularization procedures: First coronary revascularization for IPE vs. placebo: 9.2% vs. 13.3% (p < 0.0001), NNT = 24. PCI: 7.7% vs. 10.9% (p < 0.001), CABG: 1.9% vs. 3.0% (p = 0.0005). Reductions were noted in both elective and urgent/emergent revascularizations. Results were consistent across tertiles of TG levels (p for interaction = 0.07). Total revascularizations were similarly reduced with IPE vs. placebo.)
2) One limitation mentioned by the podcast is some patients were not on high intensity statins (See Table 1). However, if you look at the median LDL of 74 icosapent ethyl verses 76 placebo, most physicians would agree that is well controlled LDL by most standards. Interestingly, some cardiologists found this as a strength of the study (that patients were largely optimized).
3) In terms of harm, a small increase in afib was astutely pointed out. However the most important point was that stroke (fatal or nonfatal) was down 28 percent which is arguably the most important complication of afib. Therefore, people may have afib but have less strokes and 5 point MACE including death despite the afib. Additionally, the afib that was seen was largely in those who already had afib (However this is not disclosed in this paper). Benefits seem to far outweigh the risks here but nevertheless the podcast members appropriately raise this concern so that we can discuss with our patients.
4) Another final comment was on the suggestion against the routine use of this medication at this time. It is an opinion but understand the standard of care has well adopted this medication. There's a multitude of organizations that have written statements and or updated guidelines on it including, ADA, ESC, AHA, probably over 10 of them. It has only been approved for the CV indication for a year so it is still early and will take years to disseminate information to physicians but I do believe there to be benefit with this drug.
Thanks for listening to my rant. Proud member and supporter of Emrap
Ken M. - February 8, 2021 11:04 AM
Thank you for the comments and I appreciate we can each interpret the study differently.
1) I should have been clearer and said the largest HR was of an objective outcome with smaller HR seen for more subjective outcomes.
Figure 4: The reductions you provided were relative not absolute. I prefer absolute reductions rather than relative. The reasons is that relative reductions can make a small absolute difference look impressive. The 25% relative reduction in the 5 Point MACE was a 4.2% absolute. This does not mean a 4.2% reduction cannot be clinically meaningful but it should illustrate the point. The key secondary endpoint was 3.6% absolute reduction (26% relative reduction).
It could be argued that nonfatal stroke is very objective. This is because stroke mimics are a well-recognized phenomenon in the literature. Stroke mimics introduce uncertainty into the data. https://pubmed.ncbi.nlm.nih.gov/29727304/
I was not sure what point you are trying to make with the p-values and perhaps you could clarify? There has been great debate about the use of the p-value in the world of epidemiology and biostatistics. https://www.nature.com/articles/d41586-019-00857-9
All-cause mortality had the smallest magnitude for HR and the 95% CI spanned 1.0.
I’m again not sure what you are trying to say with the comparison to other treatment modalities? These provide no information as to whether or not the claim about icosapent ethyl is “true” (best point estimate of observed effect size).
My revised conclusion would be that the more subjective outcomes (hospitalizations for unstable angina, urgent or emergency revascularization, fatal or nonfatal stroke) rather than objective outcomes (death) contributed to the magnitude of the reported HR.
Thank you for bringing the JELIS 2009 trial to my attention. It too had a MACE outcome that consisted of: major coronary event, including sudden cardiac death, fatal and non-fatal myocardial infarction, and other non-fatal events including unstable angina pectoris, angioplasty, stenting, or coronary artery bypass grafting. Again, there is subjectivity to angioplasty, stenting or CABG. The RRR 19% and an ARR of 0.7%.
We record these PCMA episodes months in advance. I was not aware of the EVAPORATE trial when we recorded the January 2021 episode.
EVAPORATE was a trial n=80 showing a statistical difference in a disease-oriented outcome (DOO) not a patient-oriented outcome (POO).
CHERRY n=193 also showing a statistical difference in a DOO not a POO. We have been fooled before in medicine by DOOs and pathophysiologic explanations.
Did I make the comment about high-dose statins or was it Steve Brown? It if was Steve, I will prompt him to respond to that comment.
There was an increase reported in AF. The relative increase was around the same magnitude of the benefit (26%) but the absolute increase was 1.4%
It could be argued that the most important complication of AF is fatal and nonfatal stroke. However, I’m not as confident in the diagnosis because of the subjectivity in making that diagnosis (stroke mimics mentioned earlier). They were most likely using the new definition of stroke which includes imaging result based on pathophysiology. Do you know how many strokes were based on ischemic changes rather than functional outcome? I would suggest patients care about function more than their imaging test. This can complicate the interpretation of this type of outcome.
There are other patient-oriented outcomes with the diagnosis of stroke. People with a diagnosis of AF are more likely to be put on an antiplatelet drug or anticoagulated (DOAC/warfarin). This increases their risk of bleeding besides ICH. They are also more likely to be on a rate controlling drug (BB or CCB). This would put them at greater risk for falls. I did not see the granularity in the data presented in the trial we reviewed.
Perhaps the reason all-cause death was not statistically different was in part because the decrease in strokes was off-set by an increase in harm associated with the treatment of AF?
With regards to the “routine” statement it could come down to the definition of routine being used. https://www.merriam-webster.com/dictionary/routine
My use of “routine” was based on a review by CADTH. I could have been more specific or clearer in what I was trying to get across. Part of it was to avoid the possibility of indication creep. There is not enough compelling evidence for me at this time to be making a routine recommendation. I accept others may find the evidence compelling enough.
https://www.cadth.ca/sites/default/files/cdr/complete/SR0619%20Vascepa%20-%20CDEC%20Final%20Recommendation%20July%2020,%202020_for%20posting.pdf
I have not reviewed all the other guidelines you mentioned. Not a big fan of guidelines in general. They are there to guide care, not dictate care. The number of guidelines also does not necessarily make them correct. I tend to go back to the primary literature and critically appraise it myself (yes, I have a problem). There are quality issues with guidelines. The Institute of Medicine has a good resource on this called: Clinical Practice Guidelines We Can Trust. https://www.ncbi.nlm.nih.gov/books/NBK209539/
Thank you very much for your “rant”. I enjoy getting this type of feedback that challenges my positions. I freely admit my interpretation of the literature could be wrong, are they are subject to change. This has been very intellectually stimulating.
Keep on listening and sending us your thoughts.
Ken
Joseph B. - February 8, 2021 12:49 PM
Thanks for the replies. Some really great points raised.
In terms of your comment on the reduction in strokes (both fatal and nonfatal) in the icosapent ethyl arm but your skepticism that the diagnosis was actually a stroke, I did pull the article you reference. Stroke mimics are out there, we all agree, however to discount this endpoint as merely subjective is almost cruel. The trial, as we know, was a randomized double blind placebo controlled trial that hit on all endpoints. I can see more skepticism in trials that were marginal or had a poor design etc. The efficacy is well supported. When you look at the CV space, you rarely see a therapy this conclusive. Theres no perfect medication and we only have what is available to us. However, I do agree with some reports that this may be the biggest advance to CV risk reduction since statins 20 to 30 years ago. Good to approach with skepticism (as you do best ha), but after a thorough review, continued skepticism on this medication has me a bit curious if you support another CV preventative medication other than a statin. It’s okay if you don’t because we need critical skeptics in medicine and therapies should live up to a high standard. However, that is why I brought up the other therapies with NNTs (worse) as a comparison.
We have Jelis (japanese trial with same medication but lower dose -2G instead of 4G), Reduce it (by far the strongest evidence) and two mechanistic studies to help clarify why the efficacy occurred. Cherry showed a reduction in plaque and now that finding was again demonstrated with Evaporate. Agree patient Oriented outcomes are what matters most but disease oriented outcomes do help support the validity of patient oriented outcomes when they all align as they do here. Furthermore the STRENGTH trial failed (similar CV outcome study but with a diff omega composition of less EPA but with DHA as well) likely due to lower on therapy EPA blood levels that didn’t even get near the levels in Reduce it. We can hypothesize that obtaining a high dose EPA blood level is what matters most to patients and here in Reduce it , this level correlated very closely with the MACE reductions seen in REDUCE IT, almost accounting for all of the benefit. I think measuring this EPA blood level will become standard , higher is better. TG reduction isn’t what matters as you know. Many other therapies reduce TGs better than icosapent ethyl and all of those therapies failed CV outcome trials. I think we are at the beginning of a new era in CV preventative medicine, which to me is very interesting to watch.
All cause mortality was 274 (6.7) icospaent ethyl vs 310 (7.6), HR 0.87 (CI 0.74-1.02). Trend to benefit but not significant. I don’t think we can discount this completely as the curves began separating at a year and continued to separate (continue to obtain benefit) as time went on. Perhaps a slightly longer trial would begin to show all cause mortality. Perhaps not. You theorize possibly the harm with increased afib contributed to the lack of stat sig in all cause mortality in the treatment group. Interesting thought. They did post hoc handicap icospaent ethyl by including afib/serious bleeding as endpoints and the results still showed a statistical sig reduction in outcomes. If we want to discount this bc it is post hoc I am completely fine with that; just food for thought.
See Table 15 https://www.fda.gov/media/132479/download
I agree with your points with not blindly following guidelines and appraising the lit yourself. Good point for all listeners. At the end of the day every therapy has a risk/benefit. I can be wrong on this and respect your opinion tremendously that is why I commented. I appreciate the criticism. Only way to fine tune my thought process.
Ken M. - February 9, 2021 6:25 AM
No problem. I always enjoy talking nerdy.
Any specific comments on the use of relative vs absolute risk reduction or clarification on the use of p-values?
I could be more careful not to dichotomize the outcomes into objective v subjective. We can all agree that death is objective outcome. This had the smallest HR magnitude, and the 95% CI spanned the line of no difference. The other outcomes like stroke had greater magnitude of HRs and were more subjective. Or to say it another way, less objective (compared to death).
They change in definition of stroke to an imaging finding and not a functional outcome also is important. This change in 2009 makes strokes look better by included 30-50% of cases that would previously be diagnosed as a TIA. This is part of the evidence that increase my uncertainty about the magnitude and precision of this point estimate.
https://www.ahajournals.org/doi/pdf/10.1161/STROKEAHA.108.192218
https://pubmed.ncbi.nlm.nih.gov/12444191/
If 1/3 to ½ of the stroke reduction in this trial are based on imaging and not functional outcome that would be a less patient-oriented outcome. I do not view this as being “almost cruel” but rather a reasonable concern.
Yes, this was a randomize-double blinded placebo controlled trial. I do have more skepticism for lower methodology. However, my skeptical radar does go off seeing this was a funded study by the maker of the drug and there was over ½ page of small print conflicts of interest. My increased skepticism is based on this SRMA published in Cochrane on RCT with COIs (Lundh et al 2017).
• Authors’ Conclusion: Sponsorship of drug and device studies by the manufacturing company leads to more favorable efficacy results and conclusions than sponsorship by other sources. Our analyses suggest the existence of an industry bias that cannot be explained by standard 'Risk of bias' assessments. https://pubmed.ncbi.nlm.nih.gov/28207928/
I agree there is no “perfect medication” and that is not my expectation. This would be a logical fallacy (Nirvana Fallacy). The available evidence is not strong enough at this time to recommend routine use of this medication but rather only for a narrow indication and after a discussion with the patient in a shared decision approach. https://www.logicallyfallacious.com/logicalfallacies/Nirvana-Fallacy
My healthy scientific skepticism is applied to each medical claim. The burned of proof is on those making the claim. We all will have different levels/amounts/quality of evidence to accept a claim. If the burden is met, we should reject the null hypothesis and accept the claim. If the burden is not met, we should accept the null hypothesis. My positions for any medical claim are tentative, subject to change and confidence is proportional to the strength of the available evidence.
The NNT can be another way to express data. There are some limitations to this metric which we discussed in a previous SGEM Xtra episode.
https://www.thesgem.com/2019/12/sgem-xtra-nnt-wet-or-dri/
A good friend of mine Dr. Anthony Crocco put together a great video that explains some of the issues with the NNT/NNH. https://www.youtube.com/watch?v=SdtNJeB2i60
The magnitude of the NNT for other therapies does not provide information as to whether this therapy has a net patient-oriented outcome (potential benefits v potential harm). It would be good to remember when balancing out these two issues that the potential harms are systematically under-reported in RCTs (Hodkinson et al 2013). https://bmjopen.bmj.com/content/3/9/e003436
I am not discounting that most of the 95% CI favours treatment. With a small effect size and a wide 95% CI that spans 1.0 it decreases my confidence in this medication. I would encourage increased skepticism on post-hoc analyses. It does not mean it is incorrect but I’m more certain about a priori analyses than post-hoc ones.
Thanks for listening to PCMA, engaging with these very interesting and important comments/questions.
Joseph B. - February 9, 2021 10:39 AM
Some more good points raised. I do think we are too far apart in our thoughts here to come to some perfect agreement. I am most intrigued not by the medication or trial we are discussing but the ability to interpret this as a no go in your podcast. It's almost impressive to me at how skeptical you are and probably you are right since you have been reviewing papers for quite some time. That is why I felt encouraged to comment on the thread in the first place. For me, this was ground breaking so I enjoyed a good discussion on your thoughts and to challenge mine. It seems you have a very high evidence standard here. I did ask to see if you endorse any other CV risk reduction medications. I am trying to appreciate a trial as robust or more in this space aside from statins. As the leading cause of death in the country, heart disease could use a medication to reduce MACE events as seen here on top of maximally tolerated statins. Please note I am no way trying to be rude as it may come across like this during an online comment section. This is just an EM doc with a niche in cardio enjoying the discussion. I have heard a really nice perspective from you and respect it and will apply some of your thoughts to other drugs in the future.
To respond to P value/RRR - I agree with the limitations of those that you raise. It is sometimes the norm to which articles report to boost the looks of the numbers but we both can see through that and value the data for ourselves. The p value article you referenced in my mind worked counter to the argument being raised. It seems that with borderline p values we should not dichotomize stat significant or not but rather look at the trends. It almost looks like that article favors looking for benefit amongst NON stat significant trials (maybe even accepting all cause mortality as a benefit since it trended that way? ha). This, on the other other hand, was highly statistically significant and I think we would both agree a p value to the multiple decimal places (as is here) is certainly better than one merely P<.049. Did I interpret this correctly?
-The stroke points you raise about it being a less objective outcome are thought provoking. It almost seems everything other than death is a less objective outcome. Silent stroke (imaging) is not a benign entity; However I do enjoy your points about skepticism into the diagnosis of stroke.
Ken M. - February 10, 2021 11:39 AM
I think we do have more common ground and I do not expect perfect agreement. My position is not a “no go” but rather not to “routinely” prescribe this medication. There are narrow indications and even then, it would need to be discussed with the patient in a shared decision model.
My skepticism is informed by conducting medical research for ~37 years, practicing medicine for ~25 years and teaching critical appraisal for ~15 years. This does not make any of my positions correct but this is my area of expertise. We (myself included) have been fooled over the years. This article by Dr. Ioannidis also informs my position. https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.0020124
I do understand that he is a controversial figure lately.
Dr. Jerome Hoffman is one of my mentors. He taught me to consider just standing there as a therapeutic option. The value and art of deliberate clinical inertia is described in this article. https://onlinelibrary.wiley.com/doi/abs/10.1111/1742-6723.12922
I do have high standards. Patients deserve the best care, based on the best evidence. The level of evidence required is also proportional to the claim. Extraordinary claims require extraordinary evidence (Carl Sagan). So, when the claim that vitamin C cocktail cured sepsis I was very skeptical. https://www.thesgem.com/2017/04/sgem174-dont-believe-the-hype-vitamin-c-cocktail-for-sepsis/
I do support other CV risk reduction strategies. Each claim must stand on its own and my confidence will be proportional to the quality of the evidence. This includes the magnitude of effect size and precision around the point estimate. The goal is to try and find the least biased information There are a number of other quality checks lists which can help probe the literature for its validity. https://www.thesgem.com/2014/03/make-it-so-beem-appraisal-tools/
An example of CV risk reduction would be ASA for acute STEMI to decrease mortality. https://www.thennt.com/nnt/aspirin-for-major-heart-attack/
None of this has come across as rude. I agree tone can be hard in this form of communication. I have very much enjoyed this respectful dialog.
Yes, I tend to have more confidence in a low p-value than that of a higher p-value. However, it should not be viewed in isolation and I’m not suggesting you have asserted that point. It is just another data point that can be considered and interpreted. It has been dichotomized over the years and misused. A p-value lacks information on effect size, precision, clinically importance, cost, etc.
Glad you like the points about stroke. There is a range between objective and subjective. Most people I think would agree that death is a very objective outcome. There is still some controversy even around this idea. But even this statistic needs to be interpreted. There can be worse things than death. Survival with severe neurologic damage. Some people may prefer death over a very poor quality of life. An example of this interpretation would be the PARAMEDIC2 trial. Yes, they did have more survival with epinephrine in OHCA but it was mainly due to more people surviving. Severe neurologic impairment (mRS 4 or 5) was more common among survivors in the epinephrine groups compared to the placebo group (31.0% vs. 17.8%). https://www.thesgem.com/2018/12/sgem238-the-epi-dont-work-for-ohca/
Silent strokes may or may not be benign. How many people right now have a silent ischemia that would show up on MRI but are functionally perfectly and will die of something else? If you screened all those over 65yo looking for signs of silent ischemia how many would you find. Would knowing be a good thing for the person’s overall mental health? Would such knowledge change their behaviour for the better or worse? Would any intervention to treat the silent ischemia have a net benefit or net harm? As an example, stenting patients with stable CAD. https://www.thennt.com/nnt/stents-stable-coronary-artery-disease/
The point is it can be very complicated and there are always more layers. It is good to have these conversations to consider other valid points of views.