Anticoagulation influences long-term outcome in patients with nonvalvular atrial fibrillation and severe ischemic stroke

~Z},e'. Vem#gzoaer M

The A~eric~n Jonr#¢M o~ Ge;dzrric Ph~rm~corher~py

Case Series

Anticoagulation Influences Long-Term Outcome in Patients with Nonvalvular An-ial Fibrillation and Severe Ischemic Stroke Konstan~:inosN.Vemmos, MD, Georgios Tsivgoulis.MD, KonstantinosSpengos,MD. Efstathios Manios,MD, Savasioumanidis. MD, Nikolaos ZakopouJos,MD, and Spyridon D. Moulopoulos, MD, PhD Acute 5t£c,Pe Unit o( the Dep{/-'rr,ents oFC/inos/The;vpeutics cud Neu£c,/o£,/Un vermin/" o(k,'he;u. A/'edc
ABSTRACT Background: [ i u i t e d data exist regarding long term prognosis in patients with nonvalvular atriN fibrillation (AF) who have survived a severe, disabling stroke. Objective: The aim of tt'ds study was m assess long term ptognosis and its deter minants in a prospective case series of stroke survivors w{fl~ AF and moderate to severe handicap. M e t h o d s : From a consecutive series o f AF patients with Etst ever ischemic st*oke, we evaluated prospectively those with moderate to seveze disability (grade 4 5 on the modified Ra~kin Scale) who were t~eated during a 5 yea* fbllow up period w{fl~ either warfarin or aspirin. Death and tecta'tent vascula*' events were documenmd. Result: O a t of a pool of 438 AF patients, 191 were prospectively assessed. During a mean follow up of 50.4 months, the cumulative 5 yea* mortality was 76.7% (95% CI, 69.0 84.3) and fl~e 5 year recurrence rate was 33.79/0 (95% CI, 23.3 44.1). Cox regression analysis revealed that inaeasing age, increasing handicap, and aspirin versus wa*'fatin were independent predictors o f mortality. Prior transient ischemic attack and aspirin versus warfarin were predictors of vascular recurrence, Anticoagttlation was associated with a decreased risk o f death (haza*d ratio [HI<], 0.44; 95% CI, 0.27 0.70; P < 0.001) a*id recostent ttuomboembolism (HR, 0.36; 95% CI, 0 . 1 ~ 0 . 7 7 ; P < 0.01). Conclusion: O a t results suggest that ctlronic a*iticoagedation therapy may be eft"ective in lengthening survival and preventing rect~tent tt~omboembolism in AF patients who have surf?red a severely disabling ischemic stroke. ( A m J Ge~@~r Ph~rm~cother. 2004;2:265 273) Copyright (c" 2004 Excerpta Medica, Inc. Key words: ischemic stroke, atrial fibrillation, anticoagulants, long term outcome, mot tality, recurrence, T.is work ',~,,'4spresenr:xJ ir p r t at the ?gth lllterua*ionsJ Str ,uceCor*'s~ice, February S 7, 2004, San Diego, C s J i b r r i

kccupte~ tc~ p,xbiicotion ~ep en-~ber9, 2004 Pl-ilqted in the U~;A P\eplod~cion iu whole or pat t is !sot pe!-rdtLed

Copyright @ 2004 Excerr~zaNedica, h,c

doi: J0 10 i0 i 6/~ arnjopha! n~ )00 ~ 12 00i 1543 594 6/04/$19 00

voJume 2 * J',umber4

Dece~sbee2004

265

The Amerirc#~ ]oe¢r~a.I of Ger~e:~He Pgr#rma.co~er~py

$g b~'. Vem~,geoae~ ~t

iNTRODUCTiON

Nonvalvular atria[ fibrillation (AF) predisposes patients to left atria[ tt~ombus Formation and increases the risk of" suoke 6 fiold
This study was based on 1421 first ever stroke patients consecutively included in the Athens Stroke Registry bem~een June 1992 and November 2000. The Athens Suoke Registry is a computerized, prospective, observational data bank gathered in a university teach ing hospital that provides tertiary care services to the mban population of Athens. Derails of the st*oke popu lation evaluated over the first 5 years have been pub lished previously, i4 All patients were adnfitred from the emergency department and hospitalized in a 5 bed acute stsoke unit or a general medical ward. At the time of admis sion, an internist specializing in stroke or a neurologist examined all patients. The initial evaluation consisted o f a noncontrast computed tomography (CT) scan o f "die brain, standard blood tests, a chest x ra); and a 12 lead elecu'oca~diogram (ECG). A second C T scan or magnetic resonance imaging scan of the brain was perfiormed in most cases (4 14 days after stroke onset). During hospitalization, the fiollox~ing risk Factors were documented: hypertension, diabetes mellims, AF, hyper cholesterolemia, smoking, coronary artery disease, congestive heart failure, and tnstory of" TIA. T L 4 was defined as a temporazy and focal neurologic deficit last ing <24 hours, documented by a neurologist bef'ore the index event, t4 AF was classified as itater**~it~e~x~t if ECG

266

showed evidence o f sinus rhythm between ~2 episodes of AF during the preceding 12 months; AF was classi fled as cotasta~t if ECG showed e~dence o f sustained AF for >3 weeks ,aithout intervening sinus rhythm in the preceding 12 months, qs A[I patients ~ f l l o u t evidence of int~acerebral hemor rhage on the first CT scan received antiplatelet agents (100 325 m g of aspirin) on admission. Anticoagu]ation was initiated 1 m 4 weeks after ictus in selected patients vvidl AF, according m the tEeating physician's decision in t~e stroke uuit and dm general medical wards. Patients were required m be medically and neu*ologically stable, wid~out hemorrhagic infasct t~ansformation on the sec ond CT scan. The sdection of'patients tot anticoag~a tion therapy also rdied hea~ily on their compfiance, the presence of a motivated and supportive family emiron ment, and their access m anticoagulation services. The target range riot anticoaguiation was an international nom~alized ratio of 2.0 m 3.0 using adjusted warfarin doses. ~s* The remaining patients were t~eated with aspirin "after hospital discharge. In both groups, handicap was assessed with use o f rile mILS at hospital dischazge. :t6

Follow-Up All surviving patients were followed up prospectively by a study investigator and a t~ained nusse at 1, 3, and 6 months, and every 6 months thereafter until 5 y e a s afte¢ the index event. Follow up was routinely per fbzmed in the om~patient cffuic of our department or at the patients' usual place of residence in cases of severe handicap. Patients were assessed for neu*olog[c status, and all changes in drags were documented. The outcome events of interest were death and recuzrent vascala* e~'eu~s (ie, srioke and systemic embofism). To determine tecu*rent vascalm events aud causes of death, we evaluated all of the available in~brmation, including death certificates, hospital records, physidar~s' notes in private practice, necropsy findings, and the patients' cliuical presentation at the reg ula~ follow up assessments. R e c u r r e n t stroke was defined as a cerebrovascuiar event ~ith sudden onset and >24 hours' duration sub sequent to the initial stroke that clearly resulted in a new neurologic deficit or an increase in an existing deficit. 7,q3 The diagnosis o f .~y~temic e*~bolis'n~ was cliui cally defined as abrupt vascuiar insufficiency o f the fimbs or internal organs associated with cfiuical or radio logic evidence o f arterial occlusion in the absence of other likely mechanisms (eg, atherosclerosis). >/ Causes *Trademark:Conmadm® (Bristol Myers Squibb Company, Prklcerom New jrerscy)

5~ },e'. Vem#geos er M

o f death were divided into the %llowing 4 subgroups: (1) death due to infectious (eg, pulmonary infections, minaty infectious, septic shock); (2) death due m vas cular causes (eg, myocardial in[b_Ection; sudden death due to catdiac atthythmia, acute pulmonaty edema, ot heatt fallme; systemic embolism; recurrent stroke); (3) deattl due to other causes (eg, renal ratline, cancel); and (4) deattl due to unknown causes. The occurrence of" bleeding complications (eg, intractauial hemorrhage, uoncerebral major bleeding event) was recorded at each follow up. Noncerebral bleeding events were considered major if they tequited hospital admission, blood transfusion, or smgery, or if they caused permanent increase in disability. 7

Study Population We prospectively evaluated patients who fiflfilled all o f the fiollowing criteria: (1) electr'ocardiogtaptncally pt oven AF in the !2 months preceding the index event; (2) ischemic stroke event; (3) survival during hospital ization; (4) mRS grade 4 (moderately severe handicap: unable to walk and attend to owu bodily needs ¥;4thout assistance) or grade 5 (severe handicap: bedridden, incontinent, and requiring constant nmsing care and attention) at dischatge; (5) no echocatdiogtapt~c evi deuce o f valvular disease; and (6) no other possible sou*ces of catdiac emboli (eg, prosthetic valves, cardiac aneu*ysm, myocardial infarction in the preceding 3 months, atrial m}ccoma, catdiothotacic ratio >0,65), 7 Patients who were lost during fbllow up and who per manentty (> 1 week) discontinued anticoagulation ther apy were excluded fiom Farther evaluation.

Statistical Analysis Statistical analysis was performed to compute patients "treated with warfkrin ~ t h those teceiv[ng aspirin in terms of" demogaphic features, preexisting conditions, and functional outcome at hospital discharge. Dichotomous ot categorical variables were computed with the chi squat e test, and coutirmous variables were compared with the unpaired Student t test or Mann Whitney U test, as indicated. Continuous data are presented as mean (SD), and categorical data ate given as percentages. The Kaplan Meier product ffmit method was used m estimate "din pt obabllig ofs~vival and ofrecutrent vascular events at 1 and 5 },ears after stroke onset. 2U evaluate which fac tots contributed to 5 yea' mortaffty m~d recurrent vascu lat events, we used a univatiate Cox proportional hazatds model. Those factors that contributed m the outcome in the mffvariate analyses at P < 0.1 (because of the risk of Wpe II error ll'om low statistical power in such an arlaly

The A ~ e r i c ~ foe~r#e~I oF Ge~*i~rricI)lJ~rm~cor]Jer~

sis) were induded in the multivariate model as candidate variables. In the final multivariate Cox t egt ession analy:ses, statistical significance was defined as P < 0.05. Assoda lions are presented as hazard ratios (HRs) with thek cot responding 95% CIs. The Statistical Package for Sodal Sciences, version 10.0 for Wiudo*vs (SPSS Inc., Chicago, Illinois) was used for statistical ar~alyses.

RESULTS Four h u n & e d tlffr~y eight patients met the first 2 inclu sion criteria ofelectt ocardiogt aphically proven AF in the preceding 12 months and ischemic stroke event; 73 were excluded based on the t equit ement of st~vival dur ing hospitalization. One hundred thkty nine patients were excluded because they did not meet the criterion tbr mRS gsade 4 or 5; 9 patients were excluded due to the presence o f echocatdiogtaphic evidence of valvular disease. Thirteen patients were excluded due to other possible somces of cadiac emboli. Five patients were lost to fbllow up and 8 others permanently discontin ued anticoagulation therap?~: The present case series consisted o f 7 1 men and 120 women (mean [SD] ages, 75.3 [6.7] and 75.8 [6.8] yeats, respectively) with nonvalvular AF and moderate m severe handicap. Hypertension was the most c o m m o n risk factor (64.9% [ 1 2 4 / 1 9 1 ] ) , and intem~ittent AF was highly prevalent (35.1% [ 6 7 / 1 9 1 ] ) in this cohort. Nearly va,o thirds o f our study population (61.3% [ 1 1 7 / 1 9 1 ]) was severely handicapped, wittl mILS glade 5 at dischaxge. Only 22 of all 191 patients (11.5%) were raking warfarin before the index event, whereas 35 patients (18.39/0) weze receiving aspkin. No patient sug feting from paroxTsmal AF was raking oral warfh~in before stroke. The mean time of hospitalization was 17.6 days (range, 8 60 days). Auticoagulation was pro vided m 67 patients (35.194) at discharge and dur ing the fblIow up period, whereas 124 (64.9%) received aspkin therapy (dose range, 100 325 m y / d ) . N o stalls tically significant differ ences were documented between the warfarin and aspirin groups in terms o f baseline characteristics (Table I). Dining a mean follow np of 50.4 months (range, 112 60 months), 119 deaths (62.3% of rifle sample) and 39 (20.4%) tecuneut vascular events (32 strokes and 7 cases of s}stentic embolism) were documented. Death was recorded in 26 of the 67 patients taking w a r , i n (38.8%) and 93 of the 124 patients taking aspirin (75.0%). Inttacranial hemorrhage was identified by CT scan in 3 of the 32 recutrent cetebrovasculas events (all 3 patients were taking auticoagulation therapy; Table H). Vascular disease was the major underlying cause of death,

26'7

2%e A~eriec#~ ]or,~r~I o f Ger~e:~ ie ?br#rm~co~/Jer~py

~Z K Vem#geoae~ ~L

Table I. Baseline characteristics of the study sample

Chs acter Jstic Dem%raphics and risk %riots Age, mean (SD), 7 lime of ho,pitali ation, p-ean (SD), d Sex, no (%) Female Hale I !ypertenslo~/, !/o (%', Intel miEent afriai fibrilbdo!/, !/o £%; Onset of atrial flbrillaLior ~ l ;, no (96~ Diabetes rneiiitus, no (%) Smoking, no (9{,) Coronar 7 h e s t disease, !/u (%) Hyperc~oJesterolemia, r!o (%) I Jeart%iiure no (o8) Prevous transient ischernic ar/Jck, no (SO) Modi%d Ranl
r iodera~eto severe ,~dic@; gr ade 5

%adarin Group (n 67~

Aspirin Oroup (n J24)

746 (65) 176 ( / 4 )

76~' (69) 175 (95)

39 (5s% s's @is) 43 (6x~}} ss (3?8) sl (3i3) 16 (ass) I 5 (224) Ii(179) I0(i49) 6 (90) 2 (30)

si (693) 43 04,b 8J (653) 45 (363) 99 (934) st (177) i9 (i 53) 25 (s0>) i5 ( l l i ) ' (56) J4 (JJ3)

;,9(,,3)

<5 0 6 ) r9 (,%/)

s (569

0 13J 0996 0350

087x~ o75i

0 i73 0345 o 939 0848 0654 o394 0057

0355 0b5

severehat dica{~

Table II. Causes of death, recurrent vascular events, and bleeding complications in the study sample during the S-year follow-up. *

Out(ome Deafl-!s, no Causes of deaLh, no (%) h%ctien \/.~scuiar P\ecur~ent 4!o
Wartari!] ~roup (r 6 0

Aspirin ~r ou R (n i24)

76

93

IO 0 8 5 ) 8 (30@ 3 (i J5) 0 (00) 3 ( i i 5) P (7,,') I (38) I (38) I (38) I (38) 0 (0a) 0 (00) 6 (9~ i) / (i 04) 5 3 (45 .~' (33)

i6 ("8 @) 41 (44 !) !9 (104) 3 (3 i ) 9 (97) 5 (b4) 6 (6 b) 4 (43) S' (?7) I (i J) I (ii) i (i l) 2i (PP6) 25 ( " 0 / ) I 0 (00~ i (08)

The percent%es br c~usesof detk ,~ero co! puzedwRh the number of deSs as denor in~to~: Tbe pe~cent3gestot bleeding comp!ica%nsw~re corn p,Aed with the nurnoer of p~ents ss ~ deno~-in~tor

268

5£ N. Ve~¢ezos e~ ~L

The i ~ e H c ~

f(~nr~e~/ oF G~;d~ric l)b~rm~corher~

all major bleeding complications in aspirin and w a r ~ ' i n groups were 057% and 3 S°,% ~espectively. Determinants o f 5 y e a mortality are presented in Table U I . The initial mtivaiate analyses showed that age, smoking, pattern o f AF, mRS score, and anticoag ulation therapy were related to moKality. The multi va~iate Cox proportional hazards model ~evealed oaly age, grade of handicap, and anticoagulation therapy as significant independent outcome predictors. Increasing age and handicap were related to an increased ~isk o f 5 yea~ mortality, whereas anticoagulation was inde pendently associated with a decreased risk o f long term mo~talky (HR, 0.44; 95% CI, 0.27 0.70; P < 0.001). The pattern of AF, TLk, and anticoagulation were associated with vasculaz recuHence in the nulva~iate analysis, but only TIA and aspkin and warfarin therapy were retained as independent determinants of recur fence in the multivariate model (Table IV). The H R associated with warfarin versus aspirin therapy fbr 5 yem recmrence was 0.36 (95% CI, 0.1% 0.77; P < 0.01).

accounting %r 44.1% of deaths (20.4% from recmrent stroke) in the aspirin group (41/124) and 30.7% of deaths (11.5% fiom recmrent stroke) in the warfarin group (8/67; Table II). One yea~ after the index event, 27.7% of the stroke smvivors (31/17_2) had improved m a Ngher level of fancdonal status (mRS ~3). The cumu lative 1 and 5 year mortalky rates were 47.0% (95% CI, 39.5 54.5) and 76.7% (95% CI, 69.0 84.3), respectively. '['he risks o f recurrent vascula~ eveuts within 1 and 5 yea~s o f stroke onset wea'e 19.0% (95% CI, 12.6 25.4) and 33.7% (95% CI, 23.3 44.1[), respectively. The annual event rate of vascular recurrence was estimated at 12.9%, and the observed rate ofintracramal hemor rhage du~ing anticoagulation in tNs cohort was 2.0%. Additionally, 2 noncerebral major bleeding events re q@ring hospital admission and blood transfusion (1 case of gastrointestinal bleeding and 1 case of geul t o u r i n a y bleeding) were recorded in the warf?rin group. Only 1 noncerebral bleeding event was docn mented in the aspirin group. The annual event rates for

Table III, Univariate and multivariate Cox regression analyses determining the effect of various factors on B-year mortality among 191 patients with nonvalvular atrial fibrillation and severe ischemic stroke,

~actor

Univaria% An iysJs, i iR (95% Ci)

ABe*

171 Cl3b i S i ) t

r !uiti,~ari;£e An iysis HR (9S% C:i) 145 Ci i i

i8/) ~

Sex

Male bemSe blyper%snsiotl Diabetes meiiitus Hyperchoiestc oiernia Srr okinF; Cor',;rary heart disease Paten- oi st!ial fibrills~iorl Cons%nt Jrterrnitterff Flea!at *aiiure Plevieus transient ischemic attxk Modified Ran~.n Scale score Grade 4 Grade 5 kmtithrombotic th~qspy AspHn AnUcoa~ul;£ion I-R hazard ratio; £rade ff ~%r I0 year i~crease ¢P < O 0 0 l $P
10 I 3? CO9O 194) 10/ (0 ,3 Iss) 094 (0 S9 ~S9 090 (0as JSS) I /4 @ 0S~ as)5 I £/~ (0 66 J67)

i ;,2 COn 2o7)

I0

I0

0 i I CO48 i OG)ii I 16 133 o/-~ 24 l)

o69(o47 i o )

I0 2 15 (i43

i0 Zi2(i4i

,'3)+

I0 034 @?2 052; r

moderate to severe handicap; £rsde 5

3i2)r

i0 044 (07/ ©70)t

severe hsr dicap

269

77Jr A'~ericg~ ]~,r,~r~I of Geri;:r~ ic Pg~'m~,co@er~py

5£ K Vem~;ezose~ ~L

Table 1~4 Univariate and multivariate Cox regression analyses determining the effect of various factors on B-year recurrence of vascular events among 191 patients with nonvalvular atrial fibrillation and severe ischemic stroke. UI Ikfir late £,n iysis, kiR ('~5% CI)

Factor Age * }3ex ~¢ale Fer~ale Llypertenslon Diabetes rnellitus Hyper !!oJesteroJernJa Smokir8 Corona W heart diseuse Patem of atrial fibiilbtJor~ Constant lrte!~rnittent Heart%flute Previeu transient ischen/ic a£ack Hodified Rankin 8oAe score Grs:e 4 G~ace 5 Antithrombotic the! a!6,, A pirin AnLicoaF{uiatien J JR " J a d ratio; graJe 4 ~Per I0 yea" increase tp<0i SP< 005 '~P< 0OJ

moderate to seve~chandic@; grdae 5

P! uitk~ir isle ~' rlsbsis, LIR (95?6 Ci)

125 (084 i86) I0 I 24 lOb I 60 098 !6i ! 27

(0 64 (054 (080 {038 (06 (0 60

£42) ~05) 327) £ 45 885) 168)

!0 048 ( 0 i b i Jb)t !5 (0544~P} P81 (i24 63/)$ !0 0%' (05i

i0 054(025

JiO',

?93 (i2/ 6 / / ) $

i8;')

I0 0 3 / (0 i8 0/Q'~

i0 03o (0 i7 O7/IQ

sever haridic~

DISCUSSION

Stroke associated with AF is severe and has a poor prog nosis in terms of early survival, residual disability, and the destination after discharge, s,4 AF increases the risk o f functional impairment at 3 and 12 monflls after stroke by -~50% and --90%, respectively, independent o f other baseline risk factors, s y AF pauents ~ith severe stroke were excluded from prospective rando~Nzed r~ials such as the E~kFT and file Stroke Prevention in Atrial Fibrillation (SPAF) III study. 7,i8 Ther@ore, oux data provide the best mMlable infUrmation about the value o f anticoagulatiou in these frequently encountered patients. "2he principal findings of the present analysis were that after adiusting for known risk fhctors, anticoagnalation @erapy independently decreased the risk of recurrent vas cula~ events by almost ~ o thirds (HR, 0.36) and halved "dle risk o f long term mortality (}]JR, 0.44) in severely handicapped stroke suzvivors with AF. The RAFT estab fished the efficacy and safety profiles o f oral anticoagu lants fbr secondary prevention in AF patients ,Mth a

270

receut (<3 months) TIA or minor ischen'dc strokeJ In the SPAF III study, adiusted dose warfarin, wheu corn pared with combination therapy (low dose warfarin w~th aspirin), sigulficanfly reduced recurrent vascular eventos (P < 0.001) in patients ~@b a more remora ischemic stroke or TIA (median time from previous ~ o m b o e m bolism to study erzollment, 2.4 yea~s). '~ However, extrap olation o f these findings m AF patients who have s~zvived a severely disabling stroke may not be justified. N o ran domized climcal trial has addressed tile value ofanficoag ulation ~br secondary prevention in such patients, whose residual handicap may cause additional difficulties in compliance and m o N m r i n g of anticoagulation. Several hospital based studies have established ca~ diac disease and recurrent stroke as the major causes of death in stroke sarvivors with AF. 8,%±~: An autopsy study verified that AF is an important cause o f first ever or recmrent cerebral infarction in older patients (mean age, 79.3 yea's; range, 6 [ 101 years).*" ~ he efEcacF of ,

70

~

.

anticoagulants in preventing recurrent stroke and sys temic embolism may therefore rsanslate into lower long term mortality rates in this specific subgroup o f sttoke patients with significant functional impairment. Apart from anticoagulation therap); our data identi fled age and tile degree ofpoststroke handicap as inde pendent predictors o f mortalit3; whereas prex4o us TIAs correlated w~th recurrent vascular events. These results corroborate previous reports supporting age as the most robust predictor of death w-ithin it to 5 years after cere bral infarction. 2° ~2 Functional status (assessed by tile mRS) and urinary incontinence, a marker ofpostsr~oke handicap, have been acknowledged as predictors of long term mortality. ;°~s Elistory of TIA has been associated with art increased risk of recurrent stroke among partic ipants receix~ng aspirin therapy in the SPAF I to III tri als. ;4 Furthermore, a recent pooled analysis of pa~tici pants from 2 large randomized trials (FD~FT and SPAF III) has suggested that AF plus prior TIA causes a high stroke risk and that seconda~ y prevention ~ t h adjusted dose warfarin instead of aspirin results in substantial absolute reductions in ischenic stroke. 2~ In addition, commudity based studies have identified TL& as a base line predictor o f stroke recurrence among ischemic st~ oke patientsY~6,;7 The pattern o l A F did not emerge as an independent outcome predictor in multivariate analyses of our study population, despite its association with decreased srsoke risk and mortality in the initial nniva~iate analyses. This finding is in line w~th pre~ions reports suggesting that intermittency o f rhythm does not independently influence stroke risk when other risk factors ate consideredT;~4~ Limited data exist about the long term outcome o f AF patients who have survived a first ischemic stroke. Gnsta£sson and Britton ? prospectivdy monitored for 5 years 88 AF patients tteated in a stroke unit. The reported mortality rate at 5 years (70.3%) is similar to our' finding (76.7%), but we documented a higher' 1 year mortality rate (47.0%) when compared x~ith the previous smdy (29.7%) 9 and the F I N M O N I C A stroke register (30.5%)] ± This difference could be attributed to the more severe handicap of our study population at discharge. The 1 yea~ mortality rate a m o n g 30 day su~ vivors o f brain infarction ~ith AF in the Framingham Study (47.6%) is very close m ou~ results. ±2 The annual event rate of stroke recurrence estimated by our data (12.9%) is in accord with the emlier reported art nnal risks o f recurrent stroke in AF patients w-ith brain ink:arction that had varied between 11% and 23%. 7,9,±°,±~,±s,~ It should be noted tttat in most o f these studies, the observation period was limited at 1 to

2 years, and only a flew reports conducted ~iollow up fbr S yeats ~cl3 or longer, i° The annual rate o f major bleeding complications in the group of patients raking warfarin in a m report (3.3%) is slighdy higher than that documented in the ECFT (2.8%) 7 and the SPAF trial (2.1%). Is The higher mean age o f our patients and the lower efficacy of anti coagulation control in clinical practice as compared with the setting of a randomized uial may account for this ~figher incidence o f bleeding events. Certain limitations of the present report should be addressed. We evaluated the possible association o f anti coagulation therapy with outcome in a sample o f selected patients after adjusting tlor known prognostic fffctors. ThereFore, OL~ assessment o f the effect o f anti coagulant therapy on long term recurrence and mortal ity is obsezvational, and ouly a prospective, randomized clinical study can verify our results. [t should also be noted that the patients t~eated with warfarin were younger, had a lower rate of'pre~ions TDk, and had bet rer functional outcome at hospital discharge than those taking aspirin, although these differences did not reach statistical significance. Because age, history of TIA, and functional status at hospital disch~ge were independent predictors of mortality and recurrence, the differences in baseline characteristics between the ~ o n p s should be acknowledged in rite interpcetation of our results. A further limitation o f am' report is potential observer bias during follow up, which could affect the docnmen ration o f recurrent vascula~ events. Most of these events, howevex, were devastating or fatal, and most cases were confirmed by means of" hospital records and autopsy findings. Furthermore, we cannot provide detailed data concerning ttte patients' compliance ~Mth antithrom boric ttterapy, the intensity o f anticoagulation, or possi ble complications o f anticoagulation. Moreover, be cause the 2 ueatment groups were selected according to the perceived likelihood o f compliance, it is plansiNe that the incidence o f hemorrhagic complications in sim ilat patients subjected to chronic anticoagulation would be higher outside the confines of a clinical research tdal snch as ours. Also of some concern is die ett~cal issue involved in pursuing a relatively ag~essive r~eatment in a g~onp o f patients ~ t h moderate or severe stroke induced nenro logic deficit. It is difficult to determine whedler we a~e really doing dmse patients any good by preventing re current stroke and prolonging survival. However, the fiAlowing considerations may provide some answers regarding this issue. First, a substantial proportion o f patients (27.5%) improved m a higher functional level

271

during f~llow up. Second, the recent development o f user friendly a n t i c o a ~ a n t drugs such as the ural direct "d~ombin inNbitors, which do not require anticoagula don momtoriug and aze associated with fewer bleeding events,~9 simplifies the management of seconday stroke prevention in such patieuts. Third, only a prospective, rando~Nzed tri~ could da'ify die issue of whether AF stroke patients ~ t h severe residual disability should be excluded ~om ant[coagulation therapy solely on the basis of their functional status. On the other hand, our data set has a number o f streugths. The observation period was longer than in the EAFT and SPAF tzials. Although randomization was not performed, no significant differences in base line factors were documented bewceen the r~eatnient gtoups. Finally, the influence o f ant[coagulation on long term prognosis was assessed atker conuolling for age, sex, major vascula risk factors, functional depen dency, and comorbid conditions. The magnitude o f benefit in terms o f absolute risk reduction conferred by oral anticoagulants varies vdth the inherent dsk of srzoke. ~ AF patients with previous minor stroke a~e at particularly high risk of recuzrent vascula~ events and potentially have the most m gain ~ o m rzeatment ~ t h warfarin instead o f aspirin. 7 The findings of tNs study corroborate the eddence of pre vious trials and cleazly indicate that ant[coagulation may be most beneficial in AF patients, who are at high inuinsic risk o f tt~'omboembolism and surf?ring froni disabling cardioembolic strokes* ,~" CONCLUSIONS Our results suggest that ci~onic ant[coagulation ther apy may be eff?ctive in lengthening su~;G'al and p~e venting recurrent thtomboembolism in AF patients who have sufthted a severely disabling ischemic suoke. In the absence o f randomized trials, our data suggest that AF patients with mode[ate to severe stzoke benefit ~oni ant[coagulation to a similar degree as do patients with mild srzoke. ACKNOWLEDGMENT The authors thank P~ofessor K.G. Hart fbr his guid ance, support, and his vNuable comments and help in the pt eparation o f tNs manuscript. REFERENCES

1. Hart RG~ 1Ialperin JL, Pearce L/\~ etal, for dLe Stroke Prevention in Atrial Fibrillation Investigators. Lessons from the Stroke Prevention in Atrial Fibrillation trials A~zr~bzref~&ied 2003;138:831 838.

272

2. Hart RG, Halperin J L Atrial fibrillation mid throm boembolism: A decade of progress in stroke prevention. A ~ I~re±~ MeK 1999;131:688 695. 3. Lamassa M, Di Carlo A, Pracucci G, etal. Characteristics, outconl G arid care of stroke associated with atrial fibrilla [ion in Europe: Data fi"om a rrmldcentex multinational hospital based registry (The European Commurfity Strok~ Project). 8r~vke 2001;32:392 398. 4. Saxena R, Lewis S, Bcrge E, etal. Risk of early death arid recurrent stroke and effect of heparin in 3169 patients with acute ischemic stroke and atrial fibrillation hi the International Stroke Trial. Stroke. 2001;32:2333 233Z 5 Risk factors for strok~ and efficacy of antitt~'ombotic therapy ki attial fibrillation. Analysis of pooled data fl'om five randomized cozm'olled trials. Arc~ Irz~rrz Med. 1994; 154:1449 145Z 6. Hart RG, Benavente O, McBride R, Pear'ce I~\. Anti rlva'ombodc therapy to prevenE sa'ok¢ in patients with atrial fibrillation: A meta arlalysis.Arz'~ b~u-fr~ Med. 1999; 131:492 501. 7. The EAbW (Europearl Atrial Fibrillation Trial) Stu@ Group. Secondary prevention in non rheumatic xtria[ fibrillation i fher transient ischaemic attack or minor stroke. L~rzcea 1993;342:1255 1262. 8. Cmlddise L, Pinardi G~ Morabito A~for the Italiari Acute Stroke Stu@ Group. MortaliW in acute stroke w-lth atrial fibrillation. Sr~vke. 1991;22:169 174. 9. Gust~sson C, Britton M. Pathogenedc mecbmdsm of stroke in non ,Mvula"atrial fibFfllation:Follow~upof s~oke padents with and without atria[ fibFdladon. J I~z~efrz i~[ed. 1991;230:11 16. 10. Sage JI, Van Utter[ RL Risk of recurrent stroke in patients widl atrial fibrillation arid non vahmlal" heart disease. Sz~vhe. 1983;14:537 540. 11. Kaarisalo MM, Immonen Raiha P, Mar[ilia RJ, et N. Atrial fibrillation aid m'oke. MortaliV/ arid causes of deadl after the fu'st acute ischemic stroke. S~fofe& 1997; 28:311 315. 12. Lin HJ, Wolf PA~ Kelly Hayes M, etal. Stroke severityin atrial fibrillation. Fhe Framingham Study. 8z;,0ke. 1996; 27:1760 1764 13, Sandercock P, Barnf;ord J, Dear[is M, et N. Anial fibrilla don and stroke: Prevalence in different Wpes of arokc and influence on early and long term prognosis (Oxford shh'e Commurfity Sa'oke Project). B M S 1992;305:1460 1465. 14 Vemmos KN, Takis CE, Georgilis K, et N. The Athens S~'oke Registry: Results of a five yea/"hospital based studv Ce~'eN'op~sc lNs. 2000;10:133 141. 15. Stroke Prevention in Atrial Fibrillation Sin@. Final resuks. C~fcu[l~NofA 1991;84:527 539.

5~ }.e'. Vem#geoser M

16. Bmnford J, Smadercock P, Dennis M, et aL A prospective study of acute cerebrovascniar disease in the commurtity: The Oxfordshke Community Stroke Project 1981 86.2. Incidence, case fatality rates and overall outcome at one year of cerebral hdarction, primary intracerebral mad sub~u'acbaloid haemorrhage J Neurol Ne~rosu~2~ Psycb~z;,3,. 1990;53:16 22. 17. Vemmos KN, Bots ML, TsibomC~s PK, et al. Prognosis of stroke in the south of Greece: 1 Year mortalitT, fanctiorml outcome and its determinants: The Arcadia Stroke Registrg f Ne~rd Neuro.~*~ P~ycfd~z"O, 2000;69:595 600. 18. Adjusted dose warfmem versus low intensity, fixed dose warfaaem plus aspirin for high risk patients with a ~ a l fibrillation: Strokc Prevention in Atrial Fibrillation III randomised clinical trial. L~r~cet 1996;348:633 638. 19. Yaxnanouctal H, Tomonaga M, Shimada I-I, et al. Non valvalm" atrial fibrillation as a cause of fhtal massive cerebral infarction ha the elderly Stroke. 1989;20:1653 1656. 20. Petb, GW, Brown RD Jr, ~Vaisnant JK et al. Ischemic s~roke subtypes: A popalation based sin@ of fi~nctional outcome, survival, and recurrence. Sz-v0ke.2000;31:1062 1068. 21. Petty, GW, Brow.n KD Jr, Whisnant ~P, et aL Ischemic stroke: Outcomes, patient mix, mad practice variation for neurologists and generalists in a communi w, Ne~edoe~y. 1998;50:1669 1678. 22. Hankey GJ. Long term outcome at'ter ischaemic stroke/ ~urmasient ischaemic attack. Cere&oe~sc Dis. 2003;16 (Suppl 1):14 19. 23. Hankey GJ, Jamrozik I<, Broadhurst RJ, et al. Five yea" survival after first ever stroke and related prognostic Sac tots in the Perth Community Stroke Study Stroke. 2000; 31:2080 2086.

The A~eric~ ffoe~r#e~Iof Ge;*i~rric Pl~rm~coff~er~

24. Hart RG~ Pearce L/k~ McBride R, ct al, for the Stroke Prevention ha AErial Fibrifiation (SPAF) Investigators. Factors associated w-lth ischemic stroke dutemg a.sp~m ther apy in atrial fibrillation: Analysis of 2012 pm'ticipaa~ts kl the SPAF I III cikfical trials. Stroke 1999;30:1223 1 2 2 9 2 5 Hart RG, Peea'ce LA, Koudstaal PJ. Transient ischemic attacks in pat[trots with a~ml fibrillation: Implications for secondary prevcx~tion: The European AtYxal Fibrillation Frial and Stroke Prevention in Atrial Fibrillation III Tte~al. Stroke. 2004;35:948 951. 2 6 Sobel E, Alter M, Davanipour Z, et aL Stroke in the Lehigh VNley: Combined risk factors for recurrer~t ischemic stroke Ne~rdo.gy. 1989;39:669 672. 2 7 Ehaeihoum AM~ Goransson M~ Falke I), ~ranzon L Three year surglval a i d recurrence afier stroke in Malmo, Sweden: An aa~alysis of stroke registry data. Sz-~0ke. 1998; 29:2114 2117. 28. Hart RG, Peaxce LA, Rothbart BaM, et al, for the Stroke Prevention in Atrial Fibrillation Investigators. Stroke wida intermittent atrial fibrillation: Incidence and predic Eors duFmg aspirhl therapy. J Af~ Co]l Ca~*~d~o[. 2000;35: 183 1 8 7 29. Olsson SB, [;or the ExecutNe Steetemg Committee on behalf of" the SPORTIF III Investigators. Stroke: preven tion ~*~th the oral dh'ect tbromb[n intalbitor ximdagatran compared with warfaaem hi patients ~4th non valvalar atria[ fibrillation (SPOKI'IF III): Randomised controlled trial. L~ezcet. 2003;362:1691 1698. 30. Hart P,G, Pearce LA, Miller VF, et al. Cardioembolic v s noncardioembolic strokes hi atrial fibrillation: Frequency and effect of antitt='ombotic agents in the stroke preven lion ha atrial fibrillardon studies. CereF1o~sc Dis. 2000; 10:39 43.

A d d r e s s c o r r e s p o n d e n c e to: Konstant{nos N . V e m m o s , M D , A c u t e Stroke U e d t o f the DepartmenLs o f Clirfical T h e r a p e u t i c s and N e m o l o g T , University o f A t h e n s Medical School, ~,Mexandra H o s p i t a l , Vasilissis Sofias 80, 1 1 5 2 8 A t h e n s , G*eece. E mail: v e m m o s k @ a t h . f o r t h e t . g r

273