Featured Post

Is the new professionalism and ACP's new ethics really just about following guidelines?

The Charter ( Medical Professionalism in the New Millennium.A Physician's Charter) did not deal with just the important relationship of ...

Thursday, December 22, 2016

Risk of stroke in patients with atrial fibrillation -with and without anticoagulation

Previously, I had commented on the risk of stroke in atrial fibrillation patients not treated with oral anticoagulants (OAC).In that commentary I quoted Dr. Overvad -"...current guidelines discrepancies also reflect the fact that the level of stroke risk among men with a score on 1 and women with a score of 2 is on the borderline of where the  impact of anticoagulation  treatment shifts from beneficial to harmful."  His comment seems reinforced by a recent publication in Circulation.

After that commentary I became aware of a very important -and perhaps dogma changing- article published in Circulation by GR Quinn. I added an addendum to that post but  the article is important enough to be highlighted in another commentary. I re-post that addendum here:

"Unfortunately I became aware of the 2016 Circulation article by GR Quinn et al after the above commentary was published.  That very important article  provides good reason to question the dogma that the CHA2DS2-VASc scores translate to fixed stoke rate. It is generally accepted that if a person's stroke rate is estimated to be 1-2% per year then treatment with an OAC offers a  net clinical benefit and that the risk score clearly relates to a quantitative stroke risk, e.g. a CHA2DS2-VASC score of 1 means the person has a risk of about 1 % per year and a score of 2 indicates a risk of 2%. 

However, analysis of 34 studies of patients not treated with anticoagulants demonstrated that the stroke rate varies widely in various cohorts. For example, with a risk score  of 2, 27% of the cohorts reported a stroke risk of less than 1% and 33% reported stroke risk greater than 2% per year. So the correlation between risk score and stroke risk varies with the cohort studied.The numbers from the Northern European studies formed the basis of the alleged relationship between the CHA2DS2-VASc score and annual stroke risk and the North American Cohort analyses indicate significantly lower ( about 1/3 of the European rate) stroke rates for untreated AF.

 Quoting from the authors conclusions: ' The majority of cohorts did not observe stroke rates that would indicate a clear expected net clinical benefit for anticoagulating AF patients with a CHA2DS2-VASc score of 1 or 2.' "


Prediction is difficult, especially the future. Neils Bohr 

The most that can be expected from any model is that it can supply a useful approximation
to reality .   George Box

Friday, December 16, 2016

The old and the new paroxysmal atrial fibrillation- new questions raised by new technology

Paroxysmal atrial fibrillation (PAF) has typically been diagnosed in a patient with cardiac symptoms seeking care for   episodic symptoms  which ultimately- by either  office EKG or some monitoring device shown to have atrial fibrillation ( AF).

The medical literature is conflicted in regard to the risk of PAF versus permanent or persistent AF. It is not surprising that expert opinion in that regard is conflicted as well, While opinions may differ current US guidelines recommend oral anticoagulation( OAC) for PAF patients based on their risk assessment score. Usually the CHA2DS2-VASc score is used and a value of 2  or more would lead to recommendation of OAC.In other words. the same recommendation for persistent or permanent AF applies to PAF.Note this refers to what I will call "clinical PAF"  subclinical PAF (SCAF) has become an  clinical management  issue as huge amounts of data has become available from the rate and rhythm recordings of pace maker patients.

A 2014 report by Vanassche which analyzed data from 6563 AF patients not treated with oral anticoagulants (OAC) determined that stroke risk in permanent AF was roughly twice that of PAF.
They reported that  AF pattern ( PAF versus non PAF )  was the second strongest predictor of stroke second only to previous history of stroke. If their data are  conclusive in this regard that does not necessarily mean that OAC  not be used . Even though the risk of stroke is lower, still the stroke risk may overwhelm the risk of  serious bleeding  from OAC leading  to a positive clinical benefit or trade-off and a reasonable recommendation for OAC.

A new and expanding data base has emerged from the experience of pacemaker (PM) patients. The ability of PMs to store and analyze large amounts of rate and rhythm data has lead to the realization that typically asymptomatic bursts or runs of subclinical atrial fibrillation  are very common in PM patients, values from several studies reporting  an incidence of 30-55%.

The questions arises-is the risk of  these SCAFs equivalent to that of the classically diagnosed PAF. Further, is there value to quantitative the SCAF perhaps in terms of duration or frequency . The literature has adopted the concept of atrial fibrillation burden ( AFB) and several studies have attempted to determine if there is a correlation between PAF ( as in minutes per day) and stroke risk). More importantly  is it possible to find an AFB threshold associated with a clinically important risk of stroke or how much SCAF is necessary to warrant OAC.

In general ,reports  of these attempts  indicate  there is a correlation between PM detected  AFB and stroke risk , however the reports differ in regards to  finding  burden levels that demonstrate an increase hazard ratio- usually in the range of 2 or under (and  sometimes statistically significant and sometimes not). The small number of events of interest-stroke and peripheral emboli- in various duration based subgroups lead to wide confidence intervals and HR values that fluctuate as burden levels increase-sometime being statistically significant at lower levels but  surprisingly not at higher which was the case, for example in the SOS-AF project.

Some of the of  AF burden levels at which a increased hazard ratio (HR) for stroke or other embolic events) have been reported from various studies are shown here: ( note with one exception runs of SCAF less than five or six minutes-depending of the particular study-are not analyzed.) So to date with one recent  exception there is essentially no data on the risk of SCAF events of less than 5 minutes

ASSERT six minutes
TRENDS -5.5 hours
SOS AF - one hour
CARELINK/VA 5.5 hours

These studies were reviewed by Chen-Scarabelli et al in 2015 and also by Camm et al in 2016 and they reached opposing conclusions. Camm's review concluded that data were insufficient to recommend OAC in SCAF while Chen-Scarabelli  and Kenneth  Ellenbogen said that OAC should be initiated on the basis of stroke  risk assessment using the CHA2DS2-VASc score regardless of the mechanism of detection of the atrial fibrillation.

 In the observational TRENDS project, data from 3035 PM patients were analyzed. Those characterized as a "low" AF burden (AFB) ( low is  defined as less than or equal to 5.5 hours on an single day) gave a risk estimate similar to having zero  AF.  An AF burden of greater than 5.5 hours doubled the thromboembolism risk.

Also a sub-study of the TRENDS project showed that 29 of the 40 patients who had a embolic event had no AF detected in the 30 day period  preceding the event suggesting that the relationship between AF and stroke may not be a simple as the former invariably causing the latter.  


As Glotzer and his co-authors in the TRENDS project emphasized, the available data-even combining data from several studies as  was done in the SOS AF project- do not necessarily make it possible define a "safe "AFB that confers no risk greater than observed with no AFB. The event rates ( i.e stroke and other embolism events) are low in all of these studies, lower than predicted based on a widely accepted 4% per year rate in atrial fibrillation patients) and this  limits statistical precision and produces wide confidence limits as well as limits the number of threshold ranges that can be analyzed with the expectation of statistical significance.

 Small numbers in each  duration subgroup lead to results that sometimes seem counter intuitive- for example : In the SOS AF a statistically increased HR was found for the five minutes cutoff and the one hour cutoff but not the six hours or 12 hour or 23 hour. A long duration would have been thought to impair a higher HR.

The RATE RHYTHM study was designed to determine if there is a definable burden of AFIB that will be predictive of thromboembolic events. At this writing I have only access to an abstract of the findings. I believe that the investigators found there were no risk for bursts of atrial fibrillation less than 20 seconds. . 


The data indicate  that the risk of PM detected SCAF is lower than the clinically detected classical PAF but there is a  correlation between SCAF and stroke risk and SCAF is an independent risk factor for or  predictor of stroke. ( An independent risk factor is not necessarily causal- see here) .  Of course  the question is with the variable  coarse- grain risk values derived from any of  these studies is anticoagulation for these SCAF warranted and in that regard there is not unanimity of expert opinion.

Guidelines ,with the exception of the Canadian Cardiovascular Society guidelines, do not recommend  specifically to give OAC for device-detected AF. The Canadians suggested that OAC could be given to patients 64 and older with a CHADS score one or more who have SCAF episodes lasting a day or more  or for those with a shorter interval if they have recently had a cryptogenic stroke. ( ref 1)

A number of researchers have argued that randomized trials are needed to determine the clinical  benefit for OAC in PM detected   PAF. Two such trials are now in progress The ARTESIA project aims to enroll 4000 patients and randomize patients with at least one episode equal to or greater than 6 minutes of AF  to either apixaban or aspirin. 2019 is the projected end date.  Note even after this trial is complete there will still not be any RCT level evidence about what to do with patients  with AF burdens of less than 6 minutes per day.  The NOAH project will compare edoxaban or aspirin/placebo in patients over 65 years of age with one additional CHA2DS2VASc risk factor.

RCTs do seem to be needed, I believe we have learned all we can from the type the observational trials mentioned above and expert opinion differs regarding the management of SCAAF detected by PM recordings. An optimistic view is that these 2 RCTS may allow recommendations regarding OAC and device detected to move past battling expert opinion and trying to tease clinical truth from conflicting observational data.


1.Verma A et al 2014 Focused update of the Canadian Cardiovascular  Society Guidelines for the Management of Atrial fibrillation. Can J Cardiol. 2014 30 114-1130

update 12/16/16 A paragraph inadvertently omitted was added. 
addendum 1/18/17 and 2/4/17  several minor changes made in syntax and typos corrected.

Friday, December 09, 2016

How do patients with atrial fibrillation do when not treated with anticoagulants?

Clinical guidelines and conventional medical wisdom all recommend oral anticoagulation (OAC) for patients with atrial fibrillation who are determined to have an increased risk of stroke according to widely used risk indicator tools. Usually the CHA2DS2-VASc risk assessment tool is used and for   patients with a risk score of greater  than  or equal 2 ,OAC is generally recommended.European and Canadian and U.S.  guidelines differ  in regard to recommendations for risk score of 1. For risk categories 0 and 1 there are no randomized clinical trials demonstrating the net clinical benefit of OAC . Those recommendations are driven simply observational trials and  expert option seems to vary geographically.


 According to a 2016 article by Overvad et al :. "current guidelines discrepancies also reflect the fact that the level of stroke risk among men with a score of 1 and women with a score of 2 is on the borderline of where the impact of anticoagulation treatment shifts from beneficial to harmful."( ref 2)
In simple terms-it is a close call. 

 The clinical issue is the trade off between OAC reduction in ischemic stroke risk and the risk of serious bleeding complications the worse of which is intracranial hemorrhage. Importantly and perhaps ironically the higher the CHA2DS2 score , the higher the risk of major hemorrhage. Analysis (ref 1)  of over 44 thousand in the Military Health System patients with non-valvular atrial fibrillation over a 2.5 year period treated with rivaroxaban demonstrated a strong relationship between the risk score and the risk of major bleeding.So the patients at the highest risk of stroke are  the same people who have the highest risk of major hemorrhage.,especially those with the high vascular disease components of the risk score score.The data indicated that while the risk of hemorrhage increases with the risk score the stroke risk increases more so it is generally believed that there is net clinical benefit even at the highest hemorrhage risk levels.

Hess et al published a study focused on data from a outcomes registry for patients with atrial fibrillation in part to determine if patients with afib were being treated according to  accepted guidelines.(American Journal of Medicine 2016 , Nov 22).See here for link to abstract.

There were 9555 patients with afib and 1846 of those were not being treated with OAC even though they had a CHA2DS2-Vasc score greater than 2.

Interestingly the stroke risk in those not treated compared with those who were treated was not statistically significantly elevated.. The adjusted HR was 1.18 with a 95% Confidence interval of 0.91--1.54.

It is often stated that risk of stroke in afib patients is increased by a factor of 3- 5 (based on Framingham data) and that OAC may decrease that risk by 60% or more.Yet in this large data base OAC did not seem to bring about that degree of benefit. However, absence of OAC was associated with a statistically significant increase in risk of death. ( HR 1..22 (1.05--1.41).

So how do patients do  with atrial fibrillation  not treated  with OAC ? Based on this observational data- better than you might expect. On the other hand if we look at older data from a randomized clinical trial published in 1991 ( SPAF trial) the stroke risk per year for the placebo arm was 6.3% versus the warfarin  arm which was 2.3% and surprisingly the aspirin  arm was 3.6%.Later studies never confirmed the value of aspirin  in stroke prevention in afib but the SPAF numbers conform with the broad brush comments that the risk of stroke is about  three to five times and the risk reduction from OAC may be in the 40-60% range. The BAFTA study not only failed to show the value of aspirin in stroke reduction in afib patients but demonstrated that in older patient s ( over 75 years of age) aspirin caused a similar risk of intracranial bleeding as warfarin.


So the question remains why did the patients without treatment do so well in Hess's data or maybe the real life treatment of afib with OAC is much less impressive than it  is in randomized clinical trials or maybe there are so many potential biases in observational big data exercises  that solid "take home" messages are difficult to find and/or rely upon..

And things get even murkier or perhaps more clear when Dr. H. Kamel  discusses challenges to the notion that the risk of stroke in afib is in fact  solely due to   thrombi in the left atrium and proposes  that things are much more complicated including the notion that at least sometimes a stroke can cause an atrial thrombus. It is well known that stroke can precede the onset of afib.He and coauthors discuss a new model for stroke and afib in which both afib and emboli are downstream  effect of abnormal atria substrate (an atrial myopathy).This model might explain why stroke  risk is not eliminated by rhythm control strategies and why stroke can predate afib and generally the poor temporal relationship that exist between afib onset and stroke and why some reports do not show a dose response relationship between afib duration and stroke(.However, other reports such as the TRENDS data do show  dose response relationship .)

Drs Akar and Marieb make similar comments and say in part "...AF may simply be a marker of underlying conditions that causes stroke as opposed to an active participant in the stroke  pathogenesis"

 In short, they are saying  there   is reason to believe there is much  more to it than simply that afib cause thrombi in the left atrium and the clot embolizes  to the brain. although no one is saying that does not happen (" Atrial Fibrillation and Thrombogenesis:Innocent bystander or guilty accomplice", JACEEP 2015, 2015;1(13) p 218.

Another article  (ref 3) tends to support the notion that there is more to afib and stroke than the century old model that the former simply  leads to the later.Boriani published a study evaluating the risk of stroke in men and women with pacemakers and found that atrial fibrillation of at least five minutes occurred in 44% of 2398 patients and that a higher atrial fibrillation burden was not associated with a higher stroke risk. Confirming other studies this analysis found the stroke and TIA risk in women was twice that on men.

 references

1.Peacock,WF et al CHA2DS2-VASc Scores and Major Bleeding in Patients with Nonvalvular Atrial Fibrillation Who are Receiving Rivaroxaban.Annals of Emergency Medicine, published online, accessed  12/4/16

2. Overvad TF et al. "Treatment thresholds for stroke prevention in atrial fibrillation :observations on the CHA2DS2-VASC Score" Euro Heart Journal-Cardiovascular pharmacology Published on line August 2016

3. Boriani, G et al " The increased risk of stroke/transient ischemic attack in women with a cardiac implantable electronic device is not associated with a higher atrial fibrillation burden. Europace: 2016, dec 28.

"Truth is much too complicated to allow anything but approximations" John Von Neumann 


Addendum: 12/22/2016. Unfortunately I became aware of the 2016 Circulation article by GR Quinn et al after the above commentary was published.  That very important article  provides good reason to question the dogma that the CHA2DS2-VASc score translate to fixed stoke rate. It is generally accepted that if a person's stroke rate is estimated to be 1-2% per year then treatment with an OAC offers a a net clinical benefit and that the risk score clearly relates to a quantitative stroke risk, e.g. a CHA2DS2-VASC score of 1 means the person has a risk of about 1 % per year and a score of 2 indicates a risk of 2%. 
 However, analysis of 34 studies of patients not treated with anticoagulants demonstrated that the stroke rate varies widely in various cohorts. For example, with a risk score  of 2, 27% of the cohorts reported a stroke risk of less than 1% and 33% reported stroke risk greater than 2% per year. So the correlation between risk score and stroke risk varies with the cohort studied.The numbers from the Northern European studies formed the basis of the alleged relationship between the CHA2DS2-VASc score and annual stroke risk and the North American Cohort analyses indicate significantly lower ( about 1/3 of the European rate) stroke rates for untreated AF.
 Quoting from the authors conclusions: " The majority of cohorts did not observe stroke rate that would indicate a clear expected net clinical benefit for anticoagulating AF patients with a CHA2DS2-VASc score of 1 or 2."