Rapid Responses to:

PAPERS: F C Taylor, H Cohen, and S Ebrahim Systematic review of long term anticoagulation or antiplatelet treatment in patients with non-rheumatic atrial fibrillation BMJ 2001; 322: 321-326 [Abstract] [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

The hidden risks of systemic anticoagulation

N G Mandal   (14 February 2001)

Systematic review of warfarin versus antiplatelet treatment for AF: garbage in equals garbage out

Gregory Peterson   (14 February 2001)

Inclusion criteria determine results of review

Jonathan Mant   (14 February 2001)

SIFA Study: Indobufen

Eduardo Stragliotto   (14 February 2001)

Are not the authors's conclusions biased?

Jose Maestre   (18 February 2001)

Could conclude: warfarin prevents more strokes than aspirin, without significantly increasing bleeds

Andrew Evans   (20 February 2001)

Risk stratification

V Baskar   (20 February 2001)

EDITORIAL IRRESPONSIBILITY?

Kevin S Channer   (20 February 2001)

Morbidity vs mortality

Kit Byatt   (23 February 2001)

warfarin use in the elderly with non-rheumatic atrial fibrillation

H Al-Qassab   (27 February 2001)

Thromboprophylaxis in atrial fibrillation: when a meta-analysis lays a finger on a weak spot

Giuseppe Bellelli, Angelo Bianchetti, Renzo Rozzini, Marco Trabucchi   (4 March 2001)

How do we decide between warfarin and aspirin?

Christopher Cates   (5 March 2001)

Systematic review of long term anticoagulation or antiplatelet treatment

M G Cripps   (6 April 2001)

Uncertainty of benefits and increasing risks in the elderly

V Adhiyaman   (10 April 2001)

Number of patients required

Agostino Colli   (27 August 2001)

The hidden risks of systemic anticoagulation 14 February 2001
N G Mandal, Specialist Registrar in Anaesthetics Poole, UK Send response to journal: Re: The hidden risks of systemic anticoagulation	Editor - I was interested to read the article by Taylor and colleagues [1]. The authors have discussed the direct risks of systemic anticoagulation in terms of fatal and non-fatal bleeding. I would like to mention some of the indirect perioperative risks of systemic anticoagulation. Many patients with non-rheumatic atrial fibrillation are old and often with other systemic problems like hypertension, ischaemic heart disease, diabetes, chronic obstructive pulmonary disease etc. They are high risk for the development of perioperative complications. Many such patients get admitted to hospital with acute illness or trauma for an emergency operation. Quite often, the operation has to be delayed because of the abnormal clotting results. As a result of this, they do not get the benefit of an early operation and they have to stay in bed immobile for a longer period of time. Sometimes, these patients have to be taken to theatre with an abnormal clotting. Thus, they are more likely to bleed and also, more likely to have transfusions with blood and blood products. Although these patients could be provided with general anaesthesia for their operation but they can not have a regional (e.g. spinal or epidural) anaesthesia because of the clotting abnormalities. Thus, they do not get the benefit of good postoperative pain relief and comfort [2.3] provided by regional anaesthesia/analgesia techniques. It is known that a regional anaesthesia/analgesia technique improves the postoperative outcome in high risk surgical patients [2]. Thus, these patients are deprived of this benefit too. A delayed operation and a longer immobility in the bed together with the absence of regional anaesthesia/analgesia put these patients at higher risk of perioperative complications. Additional risks of blood loss and blood transfusion are always there. These added risks are the indirect results of systemic anticoagulation. Dr N G Mandal, Specialist Registrar in Anaesthetics, Poole Hospital NHS Trust, Longfleet Road, Poole, Dorset, BH15 2 JB, UK. n_mandal@hotmail.com References: 1. Taylor FC, Cohen H, Ebrahim S. Systematic review of long term anticoagulation or antiplatelet treatment in patients with non-rheumatic atrial fibrillation. BMJ 2001; 322: 321-6 (10 February 2001) 2. Grass JA. The role of epidural anaesthesia and analgesia in perioperative outcome. Anesthesiol Clin North America 2000; 18: 407-28 3. Michaloudis D, Petrou A, Bakos P, Chatzimichali A, Kafkalaki K, Papaioannou A, Zeaki M, Flossos A. Continuous spinal anaesthesia/analgesia for the perioperative management of high-risk patients. Eur J Anaesthesiol 2000; 17:239- 47
Systematic review of warfarin versus antiplatelet treatment for AF: garbage in equals garbage out 14 February 2001
Gregory Peterson, Professor of Pharmacy University of Tasmania Send response to journal: Re: Systematic review of warfarin versus antiplatelet treatment for AF: garbage in equals garbage out	Systematic review of warfarin versus antiplatelet treatment for AF: garbage in equals garbage out EDITOR - The systematic review by Taylor et al. (1) clearly illustrates how the usefulness of a meta-analysis is limited by the quality of the studies included in the analysis. The authors state that they assessed the quality of the reviewed trials based on the level of concealment of random allocation, degree of blinding used, and losses to follow-up. This is not good enough. These criteria have more to do with the statistical validity of the trials than the equally important issue of their clinical validity. A critical aspect is whether the trials approximated clinical practice with regard to the characteristics of patients with non-rheumatic atrial fibrillation. In this case, the decision to include the flawed PATAF study (2) severely weakens any findings on meta-analysis. The PATAF study has been extensively criticised on numerous grounds (3-6) - for including a high proportion of low-risk patients with lone atrial fibrillation, excluding patients with chronic heart failure, and arbitrarily excluding all patients aged 78 years or older from the standard anticoagulation arm of the study, and for a lack of statistical power. Furthermore, the PATAF study had a high drop-out rate, ranging from 20 to 32% for the three treatment arms - a fact which was not included by Taylor et al. in their Table 1. Clinicians should be wary of applying the implications drawn from the results of this imperfect meta-analysis to the care of their patients with atrial fibrillation. Gregory Peterson, professor. Shane Jackson, PhD student. School of Pharmacy, Faculty of Health Science, University of Tasmania, GPO Box 252-26, Hobart, Australia 7001 1. Taylor FC, Cohen H, Ebrahim S. Systematic review of long term anticoagulation or antiplatelet treatment in patients with non-rheumatic atrial fibrillation. BMJ 2001; 322: 321-326. 2. Hellemons BSP, Langenberg M, Lodder J, Vermeer F, Schouten HJA, Lemmens Th, et al. Primary prevention of arterial thromboembolism in non-rheumatic atrial fibrillation in primary care: randomised controlled trial comparing two intensities of coumarin with aspirin. BMJ 1999; 319: 958-964. 3. Ellis SJ, Hans R. Using anticoagulation or aspirin to prevent stroke. Research was methodologically flawed. BMJ 2000; 320: 1008-1009. 4. Mant J, Fitzmaurice D, Murray E, Hobbs R. Using anticoagulation or aspirin to prevent stroke. Study does not have the power to show that aspirin is as good as anticoagulation. BMJ 2000; 320: 1009. 5. Evans A, Perez I, Kalra L. Using anticoagulation or aspirin to prevent stroke. Anticoagulation has a major role in primary prevention of stroke in general practice. BMJ 2000; 320: 1009. 6. Rutten FH, Hak E, Hoes AW. Using anticoagulation or aspirin to prevent stroke. Results of the study cannot be generalised to the general practice population. BMJ 2000; 320: 1009-1010.
Inclusion criteria determine results of review 14 February 2001
Jonathan Mant, Senior Lecturer Department of Primary Care & General Practice, University of Birmingham Send response to journal: Re: Inclusion criteria determine results of review	Editor - We welcome Taylor et al's conclusions from their systematic review that questions remain unanswered over the relative risks and benefits of anticoagulation versus antiplatelet treatment in the treatment of non- rheumatic atrial fibrillation. [1] In particular, there is uncertainty over the optimum treatment of elderly patients, who were under-represented in all the trials of anticoagulation, and in whom there are both patho- physiological reasons and empirical evidence to suggest higher risk of haemorrhage on warfarin.[2] There is also uncertainty over the generalisability of the trial data to primary care. Our own MRC funded community based Birmingham Atrial Fibrillation Treatment of the Aged (BAFTA) study, which aims to randomise 1240 patients aged 75 or over to warfarin (target INR 2.5) or aspirin (75mg) and follow them up for an average of three years will address both these issues. However, we believe that the way the review has been conducted has led to conclusions that overstate the case against warfarin. In particular, two trials that included direct warfarin versus aspirin comparisons were excluded. The European Atrial Fibrillation study (EAFT) was excluded because it was 'difficult to interpret', even though the data for the 455 patients who were randomised to anticoagulation or aspirin are available both on the Cochrane library (odds ratio of serious vascular event 0.55 (95% CI 0.36-0.83 in favour of anticoagulation) and (albeit ambiguously) in the Lancet paper that they cite.[3] The SPAF III study, which was not cited, will have been excluded because it evaluated 'combined use of anticoagulation with antiplatelet drugs'. This study compared adjusted dose warfarin to aspirin plus fixed dose warfarin and reported a rate of ischaemic stroke, systemic embolism or vascular death of 6.4% in the adjusted dose warfarin group as compared to 11.8% in the group receiving combination therapy. [4] The decision to exclude these trials is likely to have made a substantial difference to the results of the review. For example, a systematic review published in 1999 that included the same trials as Taylor et al, but also included EAFT (but not SPAF III) reported a relative risk reduction of 36% (95% CI 14-52%) for stroke (ischaemic or haemorrhagic) for patients on warfarin as compared to aspirin. [5] Thus, inclusion criteria can have a substantial impact on the results of a systematic review. This creates a problem where the eligible studies are well known (as in this case) and the review is planned after the results are available, since the impact of different inclusion/ exclusion criteria can be predicted in advance. In hindsight, it is difficult to say what the 'correct' inclusion criteria should be, but where important studies are left out, these should be highlighted, since their results may influence how people choose to interpret the results of the review. Until more data are available from prospective randomised trials such as BAFTA, we would advocate caution in denying anticoagulation to high-risk patients with AF. Jonathan Mant, Senior Lecturer David Fitzmaurice, Senior Lecturer Ellen Murray, Research Fellow, FD Richard Hobbs, Professor Gregory YH Lip, Reader in Medicine* Department of Primary Care & General Practice, Division of Primary Care, Public & Occupational Health, University of Birmingham B15 2TT & (*) University Department of Medicine, City Hospital, Birmingham B18 7QH, UK Correspondence to: j.w.mant@bham.ac.uk 1. Taylor FC, Cohen H, Ebrahim S. Systematic review of long term anticoagulation or antiplatelet treatment in patients with non-rheumatic atrial fibrillation. BMJ 2001; 322:321-26. 2. The Stroke Prevention in Atrial Fibrillation Investigators. Bleeding during antithrombotic therapy in patients with atrial fibrillation. Arch Intern Med 1996;156:409-16. 3. Koudstaal PJ. Anticoagulants versus antiplatelet therapy for preventing stroke in patients with nonrheumatic atrial fibrillation and a history of stroke or transient ischaemic attacks (Cochrane Review). In: the Cochrane Library, 4, 2000. Oxford: Update Software (Date of last substantive amendment: 2/95). 4. Stroke Prevention in Atrial Fibrillation Investigators. Adjusted dose warfarin versus low-intensity, fixed dose warfarin plus aspirin for high- risk patients with atrial fibrillation: stroke prevention in atrial fibrillation III RCT. Lancet 1996; 348:633-38 5. Hart RG, Benavente O, McBride R, Pearce LA. Antithrombotic therapy to prevent stroke in patients with atrial fibrillation: a meta-analysis. Ann Intern Med 1999; 131:492-501.
SIFA Study: Indobufen 14 February 2001
Eduardo Stragliotto, Head of Cardiovascular therapeutic area Pharmacia Italy Send response to journal: Re: SIFA Study: Indobufen	With reference to the article, the drug used in the SIFA study is INDOBUFEN and not Indoprofen. Indobufen is a reversible cyclo-oxigenase inhibitor. Currently, a new large-sized trial is comparing Indobufen Vs aspirin in primary and secondary prevention in patients affected by NRAF (SIFA II: Studio Italiano Fibrillazione Atriale II).
Are not the authors's conclusions biased? 18 February 2001
Jose Maestre, Service of Neurology. Cerebrovascular Unit H. Virgen de las Nieves. Granada. Spain Send response to journal: Re: Are not the authors's conclusions biased?	Dear Sir, this systematic review has been deeply discussed in the Neurolist: http://www.neurolist.com and in the Spanish language List of Neurology: http://listserv.rediris.es/archives/neurologia.html After all these discussions I think that a more accurate conclusions should read (stress added): "Conclusions: The heterogeneity between the trials and the limited data result in considerable uncertainty about the value of long term anticoagulation; *we remain uncertain even if our own meta-analysis has detected a significative 32% risk reduction for non-fatal stroke favouring anticoagulation* compared with antiplatelet treatment. The risks of bleeding, *albeit imaginary since we could not find a significative difference between both treatments in this regard*, and *overall* the higher cost of anticoagulation make it an even less convincing treatment option for us, so far" Accordingly, the e-BMJ Editor's statement: "Long term anticoagulation is no better than antiplatelet treatment in non-rheumatic atrial fibrillation" also has a biased flavour. Current available evidence seems to favour anticoagulation vs antiagregation even after this systematic review and meta-analysis: both treatments are similar regarding safety, and anticoagulation is better for prevention of non-fatal stroke in non- rheumatic a-fib; all in all the most part of strokes are non-fatal; thus, in my opinion, this advantage should not be dismissed. Jose Maestre MD Neurologist
Could conclude: warfarin prevents more strokes than aspirin, without significantly increasing bleeds 20 February 2001
Andrew Evans, Clinical Lecturer Guy's, King's & St Thomas' School of Medicine Send response to journal: Re: Could conclude: warfarin prevents more strokes than aspirin, without significantly increasing bleeds	Taylor et al have produced a thorough analysis of head-to-head studies of the relative benefits and risks of anticoagulation and antiplatelet agents (1). However, we feel that their conclusions are strongly influenced by their own prior hypotheses, potentially endangering patients who would benefit from adjusted-dose warfarin. The major reason to anticoagulate patients with atrial fibrillation is not to increase life expectancy, it is to prevent stroke. As several guidelines suggest, it should not be normal practice to treat all patients with atrial fibrillation with long-term warfarin (2-4). Low-risk patients are better served with aspirin. Despite the inclusion of a substantial proportion of such low-risk patients, they still demonstrate a significant benefit in favour of anticoagulation for stroke prevention. They dismiss this as modest, but then highlight a non-significant increase in major bleeding as an important harm. Reasons could be postulated for the exclusion of many of the trials, and not just for the one that weakens their argument. They also raise the question of cost of anticoagulation services, while not mentioning the large hospital, community and social costs of stroke, particularly as the large cortical infarcts associated with atrial fibrillation tend to be particularly severe and disabling (5). When we see patients with atrial fibrillation, we assess their risk of stroke and of bleeding, clinically and by the judicious use of echocardiography. We explore the potential for cardioversion, ablation or surgical treatment. If their stroke risk is high, we would still advise them to take warfarin. Andy Evans, Clinical Lecturer Lalit Kalra, Professor of Stroke Medicine Guy's King's & St Thomas' School of Medicine, London References (1) Taylor FC, Cohen H, Ebrahim S. BMJ 2001;322:321-6 (2) Anonymous. Arch Intern Med 1994;154:1449-57 (3) Laupacis A, Albers G, Dalen J et al. Chest 1998;114:579S-589S (4) Anonymous. JAMA 1998:279:1273-7 (5) Jorgensen HS, Nakayama H, Reith J et al. Stroke 1996;27:1765-9
Risk stratification 20 February 2001
V Baskar, Clinical Lecturer, Diabetic Medicine New Cross Hospital Send response to journal: Re: Risk stratification	Sir, I read with interest the meta-analysis by Taylor et al, of data comparing warfarin and aspirin in the reduction of fatal and non-fatal cardiovascular events in non-rheumatic atrial fibrillation (AF). Despite consistent differences in efficacy when compared with placebo, direct head to head comparison failed to show any significant difference between the two agents. Not all patients with non-rheumatic AF have the same risk of stroke and stratification of their risk into high, medium and low are easily made on clinical grounds (1). The differences in the risk of stroke and the benefit of anti-thrombotic prophylaxis among the various risk groups are shown in the table (2). When used in patients with high risk of stroke, especially when used for secondary prevention, warfarin clearly appears much superior to aspirin. I conclude that careful selection based on risk factors may identify a group of patients with non-rheumatic AF who would benefit from warfarin more than aspirin and comparison of the two agents in the meta- analysis after risk stratification may well prove this. Table – Risk of stroke in patients with non-valvular atrial fibrillation with and without anti-thrombotic prophylaxis. (Values are % per patient per year) Risk No Prophylaxis Aspirin Warfarin High Secondary 12 10 4-5 prevention Primary 8 4-5 1-2 prevention Moderate 4 1-2 1-2 Low 1 <1 <1 References: 1. Lip GYH. Thromboprophylaxis for atrial fibrillation, Lancet 1999;353:4 -6 2. Lip GYH, Lowe DGO. Antithrombotic prophylaxis treatment for atrial fibrillation, BMJ 1996; 312:45-49
EDITORIAL IRRESPONSIBILITY? 20 February 2001
Kevin S Channer, Consultant Cardiologist and Physician Royal Hallamshire Hospital, Sheffield Send response to journal: Re: EDITORIAL IRRESPONSIBILITY?	Dear Sir RE: EDITORIAL IRRESPONSIBILITY? Previous editorials have highlighted the reluctance of physicians to treat patients with chronic atrial fibrillation with warfarin[1,2]. The paper by Taylor et al[3] now validates that approach and in so doing may reduce the potential benefits of oral anticoagulation in reducing stroke in the population. The broad brush assessment in the systematic review over simplifies the issues. The purpose of clinical trials is to attempt to guide patient management by helping the clinician with decision-making when there is uncertainty. However, as all readers of the BMJ know all clinical trials have limitations. For the application of trial results into clinical practice it is necessary to customise decision-making for the individual patient. A good example of this is the introduction of thrombolysis into clinical practice. In ISIS II[4], which demonstrated unequivocally that Aspirin and Streptokinase reduce mortality in acute myocardial infarction (AMI), all patients with a suspicious history of AMI were recruited whatever the appearances of the electrocardiogram (ECG). Today we do not treat all patients but only those with ECG evidence of left bundle branch block or ST segment elevation. This is because an analysis of the results of ISIS II demonstrated that there was no benefit in treating other groups. This clinical interpretation of ISIS II has never been independently tested but it is established clinical practice and in all guidelines. In the same way an analysis of the clinical trials of thrombo- prophylaxis in chronic atrial fibrillation has identified high and low risk groups[5]. High risk patients are likely to benefit more from warfarin than low risk groups for whom aspirin should be used. The heterogeneity between trials as identified by Taylor reflects the patient population studied. For example, in the PATAF[6], study used by Taylor in their review, high risk cases were excluded from entry to the trial and the majority of the study population was low risk (paroxysmal atrial fibrillation in 17% and lone atrial fibrillation in 40% with only 12% of patients having a left atrial diameter above the normal limit). In deciding on the use of aspirin or warfarin for thrombo-prophylaxis it is necessary to make a clinical judgement as to the risk:benefit ratio. This risk stratification requires a more sophisticated approach than that used by Taylor. In these days of re-validation and continuing professional development the BMJ has a great potential for educating and informing practising doctors. Let us hope that by publishing this paper the editorial committee have not undermined the messages of early editorials. Yours sincerely Dr K S Channer MD FRCP Consultant Cardiologist & Physician Dr Tom Downes MBBS MRCP Specialist Registrar in Geriatric Medicine References 1. English K.M. Channer K.S. Managing atrial fibrillation in elderly people. BMJ 1999; 318: 1088-9 2. Connolly S.J. Anticoagulation for patients with atrial fibrillation and risk factors for stroke. BMJ 2000; 320: 1219-20 3. Taylor F.C. Cohen H. Ebrahim S. Systematic review of long term anticoagulation or antiplatelet treatment in patients with non-rheumatic atrial fibrillation. BMJ 2001; 322: 321-6. 4. ISIS-2 Collaborative group. Randomised trial of intravenous streptokinase, oral aspirin both or neither among 17,187 cases of suspected acute myocardial infarction: ISIS-2 Lancet 1988; ii:349-60. 5. Lip G.Y. Thromboprophylaxis for atrial fibrillation Lancet 1999; 353: 4-6 6. Hellemons B.S.P. Langenberg M. Lodder J.Vermeer F. Schouten H.J.A et al Primary prevention of arterial thromboembolism in non-rheumatic atrial fibrillation in primary care: randomised controlled trial comparing two intensities of coumarin with aspirin. BMJ 1999; 319: 958-64.
Morbidity vs mortality 23 February 2001
Kit Byatt, Consultant Physician (Care of the Elderly) QEH, King's Lynn Send response to journal: Re: Morbidity vs mortality	Sir Pretty well every patient (all over 65) I see in the clinic with atrial fibrillation has absolutely no hesitation in saying when we discuss treatment issues that they are not nearly so much worried about dying as being left incapacitated by a stroke. Both of these end points are much more common in the elderly population than in younger patients. The data available are difficult to squeeze into a convenient consensus because the different trials were methodologically different as discussed extensively in the paper, and in the rapid responses. However, simple eyeballing of all the trials suggests greater gains for anticoagulation over antiaggregation and the benefit increases with the a priori risk (as summarised in Clinical Evidence Issue 4, Dec 2000 p 135 et seq.). We need data from real clinical practice in this country to show if the various trial protocols implemented in different health care settings translate into useful clinical gain (not just significant statistical difference) for the patient in our health care system.
warfarin use in the elderly with non-rheumatic atrial fibrillation 27 February 2001
H Al-Qassab, Consultant Physician/Geriatrician King George Hospital, Goodmayes, Essex Send response to journal: Re: warfarin use in the elderly with non-rheumatic atrial fibrillation	Dear Sir, Taylor et al [1] systematic review of long term anticoagulation /antiplatelet treatment in non-rheumatic atrial fibrillation [ NRAF] would be welcomed by may clinicians especially geriatricians. Many of these studies didn't include patients above the age of 75 [ in one study only 30 out of 420 patients were above the age of 80 ] [2] . The exclusion criteria in many of these studies were quite strict [excluding among other things those with creatinine above 300 mg/l, those with dementia ] . However falls or tendency to falls have not been mentioned neither in the these studies nor in the BNF. All patients in these trials , as Taylor et al mentioned , were followed up intensively by phone and/or by frequent clinic visits . In effect , not surprisingly, extrapolation from these trials to the general elderly population and to the common clinical practice in the UK is fraught with many dangers [ to put it mildly] Some clinicians feel that many patients who should be on warfarin are not offered it [3] . However as others pointed out [4] , in many patients warfarin would not be appropriate . Many high risk elderly patients have many other co-morbid conditions and elements of cognitive decline. Anticoagulation in these circumstances can be dangerous and difficult . Nevertheless the message should be that it is essential to assess the risk and individualise treatment in all patients regardless of age and offer anticoagulation or antiplatelet unless there are strong clinical reasons not to or unless it is the patient's choice. It is always good practice to document in the notes the explanation given to patients and the reasons for any decisions. Hisham Al-Qassab Consultant Physician /Geriatrician, King George Hospital, Barley Lane, Goodmayes, Essex IG3 8YB e-mail : alquassab@zetnet.co.uk no competing interests 1.Taylor F C, Cohen H, Ebrahim S. Systematic review of long term anticoagulation or antiplatelet treatment in patients with non-rheumatic atrial fibrillation BMJ 2001;322; 321-325 2. Al-Qassab H.K. Anticoagulation in atrial fibrillation in those over 80. Age & Ageing; 2000, 29 : S1-38. 3.Sudlow M, Rodgers H, Kenny R A, Thomas R. Population based study of use of anticoagulation among patients with atrial fibrillation in the community. BMJ 1997; 314: 1529-1530. 4. Roderick E, Cox J. Identification of patients with atrial fibrillation in general practice . BMJ 1999; 318; 191.
Thromboprophylaxis in atrial fibrillation: when a meta-analysis lays a finger on a weak spot 4 March 2001
Giuseppe Bellelli, Rehabilitation and Geriatric Units, Ancelle della Carità Hospitals Cremona and Brescia, Italy, Angelo Bianchetti, Renzo Rozzini, Marco Trabucchi Send response to journal: Re: Thromboprophylaxis in atrial fibrillation: when a meta-analysis lays a finger on a weak spot	Dear sir, The excessive enthusiasm following the results of randomized controlled trials on thromboprophylaxis in non-rheumatic atrial fibrillation [NRAF] should be defined with the aphorism "let well alone". In fact, while seeking for the best pharmacological treatment, researchers of the entire world loose touch with the daily problems of the clinical practice, castling on radicalism about the inadequacy of physicians’ prescriptions. We welcome the Taylor and colleagues review [1]. While stating the lower efficacy of anticoagulants [AC] in comparison to previous studies, they indirectly voice a general feeling about the uncertainty to start AC in all clinical conditions. The question have an high relevance especially for elderly patients. We do not forget that NRAF prevalence increases steadily with age, from 0.5% for patients aged 50 through 59 to 8.8% for patients aged 80 through 89 years [2]. In fact, in these subjects, the “real world” use of warfarin is suboptimal [3]. Reasons for this low use are not conclusively clear, but a number of factors, including the patient prognosis, the cognitive and functional status, the fear of drug-interactions, and the physician perception about the compliance to the therapy, have a significant impact on the decision to start anticoagulation [4,5]. For these reasons, we are persuaded that the true question is not “which drug is more effective?” but “within the proven efficacy of single drugs, which of them is really feasible for elderly patients with chronic NRAF?”. We hope that future trials will be designed trying to encompass this ambiguity. Giuseppe Bellelli, MD Rehabilitation Unit Ancelle della Carità hospital, Cremona, Italy Angelo Bianchetti, MD Medicine Unit Ancelle della Carità hospital, Cremona, Italy Renzo Rozzini, MD Geriatric Unit Poliambulanza hospital, Brescia, Italy Marco Trabucchi, MD, Professor, University Tor Vergata, Rome and Geriatric Research Group, Brescia, Italy References 1. Taylor FC, Cohen H, Ebrahim S. Systematic review of long term anticoagulation or antiplatelet treatment in patients with non-rheumatic atrial fibrillation. BMJ. 2001;322:321-6. 2. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham study. Stroke.1991;22:983-988. 3. Bungard TJ, Ghali WA, Teo KK, McAlister FA, Tsuyuki RT. Why do patients with atrial fibrillation not receive warfarin? Arch Intern Med. 2000;160:41-6. 4. Bellelli G., Barbisoni P., Gusmeri A, Sabatini T, Rozzini R, Trabucchi M. Underuse of anticoagulation in older patients with chronic atrial fibrillation: malpractice or accuracy? J Am Geriatr Soc. 1999;47:1034-5. 5. Bellelli G, Rozzini R, Barbisoni P, Sabatini T, Trabucchi M. Geriatric assessment and anticoagulation in elderly patients with chronic atrial fibrillation. Arch Intern Med. 2000; 160:2402.
How do we decide between warfarin and aspirin? 5 March 2001
Christopher Cates, General Practitioner Bushey Health Centre, Hertfordshire WD23 2NN Send response to journal: Re: How do we decide between warfarin and aspirin?	EDITOR – The authors of this review seek to persuade us that ‘there is little to chose between the two treatment options (antiplatelet treatment and anticoagulation in atrial fibrillation), except cost’. Rather different conclusions are drawn from analysis of a systematic review of a very similar data set recently published on the Cochrane library. In the latter case the reviewers conclude that ‘the evidence strongly supports warfarin in AF for patients at average or greater risk of stroke, although clearly there is a risk of haemorrhage. Although not definitively supported by the evidence, aspirin may prove to be useful for stroke prevention in sub-groups with a low risk of stroke, with less risk of haemorrhage than with warfarin.’ (1) In these reviews it has not been possible to prove beyond reasonable doubt that aspirin is more efficacious than placebo or that aspirin is less efficacious than anticoagulation. The disadvantages of using a 5% significance level to decide if we can be sure about results was highlighted earlier this year in the BMJ (2). Non-significant trends are open to subjective interpretation when results are handled dichotomously in this way. Moreover whilst aspirin is certainly more convenient than anticoagulation, the cost argument employed by Taylor et al is flawed as the costs of caring for stroke sufferers (or those with major bleeds) has not been considered (3). The directions of the differences found in trials randomising patients to warfarin or aspirin are the same as those found in the placebo -controlled trials. If non-fatal strokes are compared to major bleeds the pooled odds ratios are almost reciprocal from the meta-analysis of the head to head trials. In practice therefore the trade off for an individual patient depends on their assessed risk of having a stroke or a major bleed. In the majority of trials included non-fatal strokes are roughly twice as common as bleeds, and therefore since both outcomes are rare the odds ratio behaves like a risk ratio. This means that in comparison with antiplatelet treatment, if 100 such patients are given anticoagulation for two years, roughly two non-fatal strokes will be prevented and one extra major bleed will occur. In practice therefore the decision to prescribe anticoagulation or antiplatelet treatment needs to be individually assessed and discussed with each patient. Some may well choose aspirin, but this needs to be on the basis of the risks that they face of having a stroke or bleeding, not on whether the pooled results of a meta-analysis reach 5% significance. References: 1. Segal JB, McNamara RL, Miller MR, Powe NR, Goodman SN, Robinson KA, Bass EB. Anticoagulants or antiplatelet therapy for non-rheumatic atrial fibrillation and flutter (Cochrane Review). In: The Cochrane Library, Issue 1, 2001. Oxford: Update Software. 2. Sterne JA, Smith GD. Sifting the evidence-what's wrong with significance tests? BMJ 2001;322(7280):226-231 3. Cates CJ. Care is required with cost effectiveness approach. BMJ. 2000;321:449.
Systematic review of long term anticoagulation or antiplatelet treatment 6 April 2001
M G Cripps Send response to journal: Re: Systematic review of long term anticoagulation or antiplatelet treatment	Dear Sir Taylor et al, BMJ Volume 322, page 321, comment that the heterogeneity between the trials of a limited data result in considerable uncertainty about the value of long term anticoagulation compared with antiplatelet treatment. They state that they identified five relevant randomised control trials of which only AFASAK 1 reported greater benefits from long term anticoagulation than antiplatelet treatment. It should be noted that the dose of Aspirin used in AFASAK 1 was 75mg whilst in the other three out of four trials using Aspirin the dose was higher (150mg, 300mg and 325mg). A conclusion from looking at all these trials might be that "the risks of a bleeding and the higher costs of anticoagulation make it an even less convincing treatment option". Another conclusion is that in atrial fibrillation it is important to use a dose of Aspirin at higher than 75mg as at 75mg AFASAK 1 found the benefits of long term anticoagulation greater than antiplatelet treatment but at the higher doses this difference was not found. In 1994/95 I chaired a sub-group of the Dorset Clinical Guidelines Working Party which produced clinical guidelines for the management of anticoagulation in atrial fibrillation concluding that as the AFASAK 1 study using 75mg of Aspirin did not demonstrate a significant benefit with Aspirin but the SPAF 2 study using 325mg of Aspirin suggested a 42% reduction in the ischaemic stroke and embolism rate in the Aspirin group, we therefore recommended that a dose of 300mg of Aspirin daily be used where Aspirin was used in atrial fibrillation. Perhaps the Aspirin/Warfarin debate in atrial fibrillation boils down to sub-groups, i.e. Aspirin 300mg is effective whereas 75mg is not. Perhaps the SPAF investigators showed that Warfarin produced absolute benefit in patients under 75 with other risk factors but that Warfarin was no more effective than Aspirin which produced a low incidence of thromboembolism in patients under 75 with no other risk factors. Yours faithfully Dr M G Cripps STURMINSTER NEWTON MEDICAL CENTRE, BARNES CLOSE, STURMINSTER NEWTON, DORSET DT10 1BN
Uncertainty of benefits and increasing risks in the elderly 10 April 2001
V Adhiyaman, Specialist Registrar Glan Clwyd Hospital Send response to journal: Re: Uncertainty of benefits and increasing risks in the elderly	Editor - I read the interesting meta-analysis by Taylor and colleagues [1]. It comes at a time when physicians are convinced about the benefits of warfarin in high-risk elderly patients with non-rheumatic atrial fibrillation (AF). Anticoagulation in AF is a very popular audit topic across the hospitals in the UK and the need for it is greatly emphasised. The two important issues are efficacy and safety of anticoagulation in the elderly. It has been shown that high-risk patients are older and many of them are unsuitable for anticoagulation [2]. The target International Normalised Ratio (INR) in clinical practice is between 2 and 3 and only one study used this range [3]. In this study, even though there was a marginal benefit in stroke prevention, it was offset by the higher fatal and non-fatal outcome in the elderly in the warfarin group. The safety of anticoagulation is hard to predict in the elderly. A safe anticoagulation could rapidly become unsafe because any acute illness could lead to falls, collapse and confusion. It may be necessary to stop the anticoagulation in many patients in the future because of substantial risks outweighing the benefits. So these patients should be monitored and reviewed regularly long term, which is not always possible to do, once they are discharged into the community. The varied responses to the article point to the fact that the role of anticoagulation in the elderly is not established beyond any doubt. I agree with the authors that more evidence is needed in the elderly. When elderly patients are anticoagulated, serious consideration should be given for long term safety and provisions should be made to review and reassess the risks and benefits. Competing interests: none V. Adhiyaman References 1. Taylor FC, Cohen H, Ebrahim S. Systematic review of long term anticoagulation or antiplatelet treatment in patients with non-rheumatic atrial fibrillation. BMJ 2001; 322: 321-6. 2. Adhiyaman V, Kamalakannan D, Oke A, Shah IU, White AD. Underutilization of antithrombotic therapy in atrial fibrillation. J R Soc Med 2000; 93: 138-40. 3. Gullov A, Koefoed B, Peterson P, Sander Pederson T, Anderson E, Godtfredson J, et al. Fixed minidose warfarin and aspirin alone and in combination vs adjusted dose warfarin for stroke prevention in atrial fibrillation. Arch Intern Med 1998; 158: 1513-21.
Number of patients required 27 August 2001
Agostino Colli, head of medical department Ospedale "A. Manzoni" Lecco Italy Send response to journal: Re: Number of patients required	In the systematic review ( BMJ 2001; 322:321-7) on anticoagulation or antiplateled treatment in patients with non-rheumatic atrial fibrillation, the authors stated that trials would require 4920 patients for each treatment group to detect a 25% superiority of anticoagulation over antiplateled treatment with an event rate of 10%, a power of 80% and a significance of 5%. According to Dl Sackett, R. B. Haynes, G.H. Guyatt and P Tugwell(Clinical Epidemiology. A basic science for clinical medicine, second edition. Little, Brown and Company, Boston 1991)we obtained 1979 as the number of patients required for each treatment group, assuming an event rate of 10%, alpha = 0.05 and beta =0.20. Thus the number of patients reported in the paper from the trials would be enough. According to our method, 4920 should be the number required assuming an event rate of 5%. We are looking forward your explanation. Thank you Sara Massironi and Agostino Colli Divisione di Medicina II Ospedale “ A.Manzoni” Lecco Italy