Background

Ischaemic stroke is typically either thrombotic (clotting within a cerebral vessel) or embolic (passage of clot material from a more proximal vessel to become lodged in a cerebral vessel). A proportion of embolic stroke events will arise from arterial vessels such as the carotid in the neck, while others will arise from the heart. In atrial fibrillation, the chamber wall does not contract properly and this “stagnant” blood is more inclined to development of thrombus (clot), which may embolise up the arterial tree to the brain (or indeed anywhere else in the body).

While embolisation from high flow vessels may be reduced by antiplatelet agents, such as aspirin, that from low flow vessels (veins and the atria) may be reduced by anticoagulant agents, such as warfarin.

It can be seen, therefore, that some causes of stroke may be reduced by anticoagulant therapy. It will also be readily seen that such therapy is not going to make any difference at all to the statistical majority of causes of stroke, and in fact by its very nature may increase the likelihood of non-ischaemic stroke, namely brain haemorrhage. Nevertheless, there is a rather long-held view that on balance anticoagulation statistically reduces the risk of stroke in many patients who have atrial fibrillation (AF). In fact, this applies not only to patients who have already had a stroke or transient ischaemic attack, or who have structural heart disease making them even more susceptible to thrombus formation, but to patients who have AF as an isolated finding – so-called lone atrial fibrillation.

Two factors have made the issue of anticoagulation for AF topical:

• Recent evidence has emphasised the view that lone AF is worth treating, as demonstrated in the UK by a recent National Institute of Clinical Evidence (NICE) Guidance document (CG 180).
• As well as warfarin, there are four (at the time of writing) new anticoagulant agents to choose from and these do not require tedious weekly to monthly blood monitoring.

However, two factors have made the issue of anticoagulation for AF controversial:

• The new drugs are much more expensive than warfarin
• Anticoagulation will kill some people and harm others. The very nature of anticoagulants means that they will increase haemorrhage from the bowel or at the time of trauma or emergency surgery and, since some strokes are in fact the result of brain haemorrhage rather than ischaemia, they will even increase the risk of  this type of stroke!

It is not surprising therefore that a meta-analysis of the major studies comparing risks and benefits of warfarin versus novel oral anticoagulants (NOACs) was published recently and has become the subject of much debate. This study, “Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with atrial fibrillation: a meta-analysis of randomised trials (Lancet 2014)” by Ruff et al., looks at the four major trials on this subject, all run by the drug companies manufacturing the NOACs.

Before describing the paper, it is worth mentioning wider controversies surrounding these studies.

First, as reported by the British Medical Journal, one of the studies (Rocket-AF) used a defective device to record the clotting effectiveness (INR) of the patients in the warfarin arm. Therefore, the patients using warfarin may have had an artificially bad outcome, not only potentially harming them but compromising the study’s findings. There is debate over when the drug company first knew about this fact.

Second, there was an issue where it was found that blood monitoring of NOACs improves outcome in terms of efficacy and reduction in bleeding complications. Obviously, there is not the same level of risk associated with not monitoring NOACs versus not monitoring warfarin, but there is an argument that some patients statistically might come to avoidable harm through not monitoring. Nevertheless, because the US Food and Drug Administration (FDA) approved the drugs before this was known about, the prescribing recommendation, and the major selling point of the new drugs, need not be changed.

Study Design and Findings

The meta-analysis looked at four studies

RE-LY, comparing dabigatran in two doses versus warfarin

ROCKET-AF, comparing rivaroxaban and warfarin

ARISTOTLE, comparing apixaban and warfarin

ENGAGE AF-TIMI 48, comparing edoxaban and warfarin.

A meta-analysis was felt to be justified on the basis that the class effect of the NOACs is similar; they all have a direct antithrombotic effect, while warfarin acts via inhibition of vitamin K, a precursor for different components of the thrombotic pathway. This results in similar shared benefits of faster onset and offset of action,  more predictable action and fewer drug interactions. Any intra-class differences would therefore be outweighed by differences between subgroup populations. Pooling the data would increase the chance of finding subgroup differences in the balance between efficacy and safety, the main stated purpose of the meta-analysis. In all there were around 72,000 non-valvular AF participants!

Median follow up ranged from 1.8 to 2.8 years and the outcome measures were occurrences of ischaemic stroke, haemorrhagic stroke, myocardial infarction, all cause mortality, intracranial haemorrhage, gastrointestinal bleeding and other major bleeding events.

Some studies had higher baseline risks than others, as expected because they had different CHADS2 scores ( a scale measuring risk of ischaemic stroke from AF).

The headline finding of the meta-analysis was that NOACS had a significantly (19%) reduced stroke risk (i.e. better efficacy) and a significantly  (48%) reduced intracranial haemorrhage risk (i.e. better safety). There was reduced all-cause mortality (10%) but increased gastrointestinal bleeding (25%). The relative efficacy and safety was consistent across a wide range of patients.

Opinion

Some of the cautionary notes on this study have already been publicised.

First, the study uses one parameter both as a measure of efficacy and safety – intracranial haemorrhage is counted twice! The study does say that the effects of stroke reduction were largely because of reduced haemorrhage; If we are looking at the effects of anticoagulation on stroke, we should be looking at its specific biological action, namely to reduce embolic stroke from the heart in comparison to warfarin, not at reduction in all-cause stroke.

Second, the way data are publicised can shift emphasis, especially for patients. The 0.48 relative risk reduction means a clinician could say that there is less than half the chance of a brain haemorrhage on NOACs compared to warfarin. But this actually equates to a 0.58% risk versus 1.24% risk. The actual risk reduction is therefore 0.66%.  (The absolute rise in gastrointestinal bleed was almost as much at 0.5%, but the latter would still be preferable on balance to intracranial bleed.)

The absolute risks are not annualised risks so they illustrate a point rather than being something useful to quote to patients. In its guidance to UK clinicians, NICE recommends that patients have a choice whether or not to receive anticoagulation. They provide a chart for clinicians so that they can explain the particular annual risk of a stroke based on the CHADS2 score versus a bleeding complication based on the VASC score. These risks are described in terms of “out of 1000 patients, x would be saved from having a stroke each year by taking anticoagulation”.  Converting the figures of this meta-analysis into an annualised form using the same language, about 5 patients per year would have a brain haemorrhage on warfarin, and 3 on NOACs.

In a rationed health economy, we have also to look at costs. In the UK, annual costs per patient for Rivaroxaban and Apixaban are around £700 to £800, and that for warfarin including the clinics for regular monitoring are£283 (NICE CG180).

NICE did a costing exercise (June 2014) based on the CG180 guidance and based on the increased uptake of anticoagulation for AF in general but they assumed that warfarin and NOAC use would increase in parallel (warfarin from 34% to 47% use in AF, and NOACs each from 4.7% to 11.7%). NICE stipulates that patient and clinician choice should determine whether patients go on warfarin or NOAC. But if the figure of 50% reduction in brain haemorrhage is quoted, and it is explained that patients do not have to attend a clinic for a blood test every fortnight, we can all guess what patients who are not paying for their medication will choose!

However, health providers might choose differently. In the UK, some have taken the view that since the excess risk of warfarin was more in a subgroup who had brittle control, warfarin should be given first line in patients for whom vitamin K antagonists are not contraindicated, and switched to a NOAC only if they prove to have brittle control.

Some may find all these economic arguments  laboured and somewhat distasteful – can one put a price on human life? But, again in a rationed economy, some person (or unwieldy committee) has to decide whether a limited amount of money goes on NOACs, or on life-prolonging cancer treatment, or on running blades for childhood amputees. The way these decisions are (supposed to be) made is on the basis of quality adjusted life years (QALYs). The UK Department of Health might have a guide of, for sake of argument, maximum £50,000 cost per QALY gained. The QALY loss of a brain haemorrhage might be 0.5 per year for 5 years, i.e. 2.5. (Some will be trivial, some will be fatal while factoring in a certain life expectancy, some will recover over time, some will die after a finite time.) The prevalence of AF is 1.6% and for 100,000 population, around 1000 should be on anticoagulation, so NOACS would save 2 brain haemorrhages per year in that population. The excess annual cost per 1000 patients is about £500,000. So a rough estimate of cost per QALY is £100,000. One could argue about additional increased efficacy and reduced mortality of NOACs, but perhaps that is subsumed by the haemorrhage reduction anyway, and there would be nearly as many increased GI bleeds as reduced intracranial bleeds.

The other reason why this topic is an emotive issue, and why it has received considerable attention with NICE invoking “patient choice”, is a controversy not related to NOACs versus warfarin as such, but to the knowledge that in a few patients we will be doing potentially terminal harm by starting either treatment. This goes against the physician’s mantra, “First do no harm” and follows the alternative mantra, “The needs of the many outweigh those of the few”. Statistically we know anticoagulation will help more AF patients than it harms, but no-one wants to be one of the few who suffer the brain haemorrhage, or to be the physician who gave them the drug that caused it.