OPTIMAL ORAL ANTICOAGULANT THERAPY IN PATIENTS WITH MECHANICAL HEART
VALVES
S.C. CANNEGIETER, M.D., F.R. ROSENDAAL, M.D., A.R. WINTZEN, M.D., F.J.M. VAN DER MEER, M.D.,
J.P. VANDENBROUCKE, M.D., AND E. BRIET, M.D.
Abstract Background. The optimal intensity of oral
an-ticoagulant therapy for patients with mechanical heart valves (i.e., the level at which thromboembohc complica-tions are effectively prevented without excessive bleed-ing) is not known. We attempted to determine the optimal intensity by calculating the mcidence of both complica-tions at different levels of anticoagulation.
Methods. Data were collected on all patients with me-chanical heart valves who have been seen at four region-al Dutch anticoagulation clmics smce 1985. The primary outcome events were episodes of thromboembolism or major bleeding. The intensity-specific mcidence of each type of event was calculated äs the number of events that occurred at a certam intensity of anticoagulation (ex-pressed m terms of the international normalized ratio [INR]) divided by the number of patient-years during
which the INR was at this level in the total patient popu-lation.
Results. A total of 1608 patients were followed during 6475 patient-years. Cerebral embolism occurred in 43 pa-tients (0.68 per 100 patient-years) and penpheral em-bolism in 2 (0.03 per 100 patient-years). Intracranial and spinal bleeding occurred in 36 patients (0.57 per 100 pa-tient-years) and major extracranial bleeding in 128 (2.1 per 100 patient-years). The optimal intensity of anticoagula-tion, at which the incidence of both complications was low-est, was achieved when the INR was between 2.5 and 4.9. Conclusions. The intensity of anticoagulant therapy for patients with prosthetic heart valves is optimal when the INR is between 2.5 and 4.9. To achieve this level of anticoagulation, a target INR of 3.0 to 4.0 is recommend-ed. (N Engl J Med 1995;333:11-7.)
P
ATIENTS with mechanical heart-valve prostheses
receive lifclong, high-inlensity oral anticoagulant
therapy to prevent thromboembolic complications, but
this trcatment is associated with an incrcased risk of
bleeding.
1The risks of thromboembolism and bleeding
depend on the intensity of anticoagulation. The optimal
intensity — that is, the level at which thromboembolic
complications are effectively prevented without
exces-sive bleeding — is not known. A number of trials have
becn conducted in which different levels of
anticoagu-lation were compared, some in combination with
anti-platelet therapy.
2' These studies, in which only two
lev-els were compared and which did not take mto account
discrepancies between achieved levels and target
inten-sities, do not offer comprehensive Information about
the optimal intensity of anticoagulation; therefore, a
di-rect relation between the intensity of anticoagulation
and the frequency of adverse events cannot be
deter-mmed from their results.
We developed a method to determine the optimal
intensity of anticoagulation^ which extends methods
developed by others for slightly different purposcs,
6 7and used it to study the dose—effect relation between
the intensity of anticoagulation and the benefits and
risks of the treatment. On the basis of the medical
rec-ords of 1608 patients with mechanical prostheses, wc
calculated the incidence of both thromboembolic and
bleeding complications at different achieved levels of
anticoagulation. The optimal intensity was defined äs
From the Hemostasis and Thiombosis Research Center, Depaitment of Hema-tology ( S C C , F R R , F J M M , E B ) , the Department of Clmical Epidemiol-ogy (F R R , J P V), and the Department ot NeurolEpidemiol-ogy (A R W), Umvei sity Hos> pital Leiden, and the Leiden Anticoagulation Chnic (FJ M M ) — both in Leiden, the Netheilands Address reprmt requests to Dt Cannegietei at the He-mostasis and Thiombosis Rcseaich Centei, Department of Hematology, Bldg l, C2-R, Univeisity Hospital Leiden, PO Box 9600, 2300 RC Leiden, the Nelhei-landsSuppoited in pait by a grant tiom the Nelhcrlands Thiombosis Foundation The institutions, mvcstigatois, and geneial practitioners paiticipatmg m the Leiden Artihcial Valve and Anticoagulation Study are hsted in the Appendix
the level at which the mcidence of both
thromboem-bolism and bleeding was lowest.
METHODS Oral Anticoagulation
In the Netherlands, treatment with coumarm denvatives (phen-procoumon or acenocoumarol) is momtored by regional anticoagu-lation chnics " All patients with mechanical prosthtüc heart vakes roceive anticoagulant tieatment through these clmics, which they visit regularly eveiy onc to six wceks At each visit, a short htstory is taken and patients are asked about thiomboembohc or bleeding complications, other dtseascs, and hospital admissions Samples of antecubttal vcnous blood are obtamcd, prothrombm times are meas-uied, and the results aic expressed m tcrms of the intei national nor-malized ratio (INR) The daily dose of the anticoagulant agcnt for the upcommg peiiod and the length of this penod are determmed by physicians, the taiget lange for the prothrombm time is an INR
of 3 6 to 4 8
Patients
Four anticoagulation clmics that use the samc computcr-assisted System to deteimme dosage'1 paiticipated m this study those in Leiden, The Haguc, Cnschede, and Oost-Gelderland Begmnmg on January l, 1985, all patients with mechanical prosthetic valves who were tieated m the clmics in Leiden and The Hague, which wcrc us-ing the System at that Urne, were mcluded in the study The antico-agulation clmics at Enschcde and Oost-Gelderland later mtroduced computeri7ed registiation Their patients weie theiefoie mcluded from the date their lecords were computenzcd, which vancd between January l, 1985, and May l, 1987 Patients who later underwent heai t-valve t cplacement wci e mcluded äs of the date of discharge (at least 30 days aftei surgery) Follow-up cnded when a patient dicd or moved to a different legion, when the anticoagulation was stopped permancntly foi any i eason, 01 at the end of the study after six yeai s of follow-np at each site Palients who wcte seen at the anticoagula-tion clmics foi less than onc month and who were not residents of the rcgion were cxcluded
Data Collection and Definitions of Events
The following infoimation was cxtiacted dorn the computenzed filcs of the anticoagulation clmics date öl bnth, sex, date oi valvc rc-placement, position and type of valves, all piothrombm-lime nieas-urements (expressed m tetms of the INR) and then dates, data on all hospital admissions, and date of death for pattents who died Additional mioi mation about hospital admissions (such äs discharge
12 THE NEW ENGLAND JOURNAL OF MEDTCINE July6, 1995
letters and autopsy rcports) wcre collcctcd from thc hospitals or from gcneral practitioners' archivcs.
The chief outcomc variable was all thromboembolic and bleeding events — that is, cercbral infarction, peripheral embolism, valve thrombosis, intracranial and spinal bleeding, and major extracranial bleeding. Cerebral infarction was defined äs neurologic dcficil of sud-den onset documented by a computed tomographic (GT) scan indi-cating the presence of infarction or the absencc of hemorrhage or at autopsy. Peripheral cmbolic events werc included in the category of acute peripheral ischcmia, demonstrated by angiography, at Opera-tion, or at autopsy. Valve thrombosis was defined äs impairmcnt of the valve by the deposition of thrombus, demonstrated at Operation or at autopsy. Intracranial and spinal bleeding was defined äs a neu-rologic deficit of sudden or subacute onset, confirmcd by CT scan-ning, surgery, or autopsy. Major extracranial bleeding was defined äs an acute bleeding event that led to dcath or to admission to the hos-pital for treatmcnt of thc bleeding; bleeding that led to hoshos-pital ad-mission for diagnostic procedures only was not considered major. Bleeding caused by trauma was also included. All strokes that could not bc catcgorized äs hemorrhagic or ischemic were designatcd "un-classified" strokes.
Since the aim of this study was to assess the optimal level of anti-coagulation in otherwisc hcalthy outpatients, we excludcd all events that occurred during a period of endocarditis. We also excluded all complications that occurred during hospitahzation for a differenl rea-son, since other diseases, interventions, and diagnoslic procedures may affect the risk of complications far more than thc intcnsity of the anticoagulant therapy. For the samc rcason, bleeding that occurred aftcr surgery or after any othcr invasive procedure was also excluded from our analysis.
All complications were reviewed by an expert panel, whosc mcm-bers classified events according to thcse dcfinitions. They also classi-fiecl the outcome of the intracranial and spinal events äs death, if the patient died äs a dircct conscqucncc of the event, or äs survival with or without neurologic deficit (äs judged three weeks after the event); neurologic deficits were categorized according lo whcthcr paticnts were dependent or indcpendent in pcrforming activitics of daily liv-ing. Thc pancl was blindcd to the intensity of anticoagulation at the time of thc event.
When the outcome of interest was "all adverse events," follow-up ended at the first event. In the analysis of a particular event, follow-up ended at the time that thc event of interest occurred; for example, if joint bleeding occurred, follow-up for cerebral embolism continued if this was the event of interest. As a result, thc number of patient-ycars, uscd äs thc dcnominator in dctermining the diffcrent incidence ratcs, varied dependmg on the outcome of interest.
Assessment of the Intensity of Anticoagulant Therapy The incidence rates of adverse events for specific levels of intensity of anticoagulation were calculated äs the ratio of the number of events that took place when thc prothrombin time was in a particular INR ränge to thc number of patient-years during which the INR was at this level in the patient population. This mcthod has been de-scribed in detail previously.' Ninety-five percent confidcnce intervals for the incidence rates were calculated with the assumption of a Pois-son distribution for the number of cases."1
Assessment of the Numerator
After an outcomc event had been reviewed and classified by the ex-pert panel, the INR at the time of the hospital admission was ob-tained from the hospital records. When thc prothrombin time had not been measured or could not bc rctricvcd, wc used thc most recent measurcment at the anticoagulation clinic before the event, but only when the test had been pcrformed less than eight days before the event. If this Information was not available, the event was not includ-ed in this analysis.
Assessment of the Denominator
The total number of paticnt-ycars for all paticnts combined was subdivided according to INR intervals of 0.5. The number of patient-years in each INR category was estimated äs lollows. For each pa-tient, thc actual INR (luctuated between and around the targct values and was measured at regulär intervals. By assuming that the INR
changed in a linear fashion belwcen mcasurcments, it was possible to allocatc a spccific INR to each day. All days werc thcn groupccl ac-cording to INR rangcs (in intcrvals of 0.5) and summcd for all pa-ticnts."' With this approach, individual patients contributed pcrson-timc to pcriods of anticoagulation at different intcnsitics, and the risks werc estimated for the INR Icvcls achicved, rcgardlcss of the target levels. When thc Icngth of limc bclwecn two mcasurcments ex-ceeded eight weeks, no INR was allocatcd, since wc feit we could not assume a linear change bclwecn two visits so far apart.
The influcncc of sex and age and position and type öl valvcs was studicd by further stratification of the INR-days. For each INR-day, thc corrcsponding age was calculated. Thc paticnts werc groupcd in three age categories (younger than 50 ycars, 50 to 69 ycars, and 70 ycars or older). Intcnsity-specific incidcncc ratcs werc calculated for these subgroups. For this analysis, thc INR values for the days werc grouped according to intcrvals of 1.0. To control for the cffects of age, we pcrformed a multivariatc analysis with Poisson rcgrcssion tcchniqucs.
RESULTS
Patients
At the start of the study, 795 paticnts who had
al-ready undergone valve replacement wcre enrollcd. An
additional 813 patients later entered the study: 767 who
had a valve replacement aftcr the study began and 46
who moved into the region. These 1608 patients were
followed for a total of 6475 person-years (mean per
pa-tient, 4 years). Follow-up ended before the termination
of thc study for 254 palients who died and for 91
pa-tients who moved to a different geographic region.
An-ticoagulant therapy was discontinucd in 11 patients for
various reasons, such äs noncompliancc. In three
pa-tients the mechanical prosthesis was replaccd by a
bio-prosthesis, so follow-up ended at the time of thc valve
replacement.
Of the 1608 patients, 879 were admitted to the
hos-pital at least once during the follow-up period (a total
of 2394 hospital admissions). Sufficient clinical
Infor-mation was available for 98 percent of these
admis-sions.
The general charactcristics of the 1608 patients
in-cluded in the study are shown in Table 1.
Control of Anticoagulation
For all 1608 patients, 123,254 measurements of the
prothrombin time were obtained during the follow-up
period. The average time between two measurements
was 2.7 weeks. With the method dcscribed above, wc
were able to estimate the number of patient-days per
INR category. No INR value could be allocated to 139
patient-years of follow-up (2.1 percent of thc total
ob-servation time), since the Icngth of time between two
consecutive measurements excccded eight weeks. Thc
distribution of the paticnt-time ovcr thc INR categories
is summarizcd in Figure 1. The INR was within the
tar-get ränge of 3.6 to 4.8 for 61 percent of the total
pa-tient-time; for 31 percent of thc time the INR was
be-low 3.6; and for 8 percent thc INR was abovc 4.8.
Outcome Events
Vol 333 No. l OPTIMAL ORAL ANTICOAGULATION FOR PATIENTS WITH MECHANICAL HEART VALVES 13
Table 1 Charactenstics of the Patents at the Begmnmg of Follow-up. CHARACTLRISIIC No (%) No o) patients 1608 No of patient-years of follow-up 6475 Age (yi) <50 343 (21 3) 50-69 995 (61 9) s=70 270(168) Sex Male 879 (54 7) Female 729 (45 3) Valve position Aoitic 961(598) Mitral 485 (30 2) Both 158 (9 8) Unknown 4 (0 2) Valve model1
Caged ball or diskl 53 (3 0) Tiltingdisk] 1354(767) Bileafletfc 347(196) Unknown 12(07) *A total of l ^8 patients had two valve prüstheses FStarr-Edwards in 46 and Cooley-Cutter in 7
IBjoik-Shiley m 10·>2, Sonn ra 164 Medtromi,-Hati m 117, and Lillchei-Kastci m l
i)St Jude m 242 DmoMediLs in "S6, and CarboMedic·, m 49
except for two cases of peripheral cmbolism (onc in the
kidncy and onc retinal; incidence, 0.03 per 100
patient-years). Thcrc were no cases of valve thrombosis that
mct our criteria. In 36 patients, central nervous System
hemorrhage occurred (34 had intracranial hemorrhage,
and 2 spinal; incidence, 0.57 per 100 patient-years).
Major extracranial bleeding occurred in 128 patients
(2.11 per 100 patient-years). These events were
gastro-intestinal (in 47 cases), genitourinary (12), or
retroperi-toneal (1); involved bleeding in the skin or muscles (33)
or joints (5); were characterized by epistaxis (17); or
were classified äs "other" (13).
Seventy-two perccnt of all patients with a stroke died
or had persistent neurologic impairmcnt. The outcome
of the hemorrhagic strokes was more scvere than that of
the ischemic strokes: 44 percent of the patients with an
episode of intracranial bleeding died, whereas only
5 percent of the ischemic strokes were fatal. Also, 79
percent of the patients with ischemic strokes recovered
without deficits or with a deficit allowing them to
func-tion indcpcndcntly, äs compared with only 33 percent
of the patients with hemorrhagic strokes.
The extracranial events were less severe than the
other events; 124 of the 128 patients with extracranial
bleeding rccovcrcd, and 4 died (incidence, 0.07 per 100
patient-years). Three of thesc four patients died of
pro-fuse gastrointestinal bleeding (one had colon canccr; in
the other two there was no known underlying discase),
and one died of bleeding from an ovarian tumor.
INR-Specific Incidence
Of the 79 patients with ischemic or hemorrhagic
stroke, the prothrombin time was measured at the time
of the event in 62. In scven additional cases, it had
been measured at the anticoagulation clinic within eight
days before the event. Thus, INR-specific incidence rates
could bc calculated on the basis of data on 87 percent
of the patients with stroke. In 111 of the 128 cases of
extracranial bleeding, the INR was measured at the
time of the event, and in 12 additional cases it had been
measured less than eight days before at the
anticoagu-lation clinic. Therefore, 96 percent of these events
could be used for further analysis. The INR-specific
in-cidence rates that were calculated from these data are
prcsented in Figures 2 and 3.
The optimal intensity of oral anticoagulant therapy
is that at which the incidence of both thromboembolic
and bleeding complications is lowest. In Figures 2 and
3, this level can be found at INR values betwecn 2.5
and 4.9. At these INR levels, the incidence of all
ad-verse events (all episodes of bleeding and embolism
and unclassified strokes) was only about 2 per 100
pa-tient-years (95 percent confidence interval, 1.0 to 3.8).
The incidence rose sharply when the INR feil below
2.5, ranging from 7.5 per 100 patient-years (95
per-cent confidence interval, 3.6 to 12.6) for an INR of 2.0
to 2.4 to 27 per 100 patient-years (95 percent
confi-dence interval, 3.3 to 99) when the INR was in the
very low ränge of 1.0 to 1.4 (a 13.5-fold increase from
the lowest incidence). Once the INR rose to 5.0 or
above, the incidence of adverse events again increased
steeply: 4.8 per 100 patient-years (95 percent
confi-dence interval, 2.6 to 7.7) for an INR of 5.0 to 5.5 and
75 per 100 patient-years (95 percent confidence
inter-val, 54 to 101) when the INR was 6.5 or above (a
37.5-fold increase).
In 77 of the 164 bleeding events (47 percent), eithcr
trauma or some underlying disease was a contributing
cause of the bleeding. This perccntage did not differ
substantially for the different INR categories. Thus, at
higher INRs, the same percentage of bleeding episodes
occurred without an apparent contributory cause äs at
lower INRs.
When the analysis was repeated for the events with
the most unfavorable outcomes only (death or handicap
ϊχ °> t* ^^tx Q , t »K0 ) t x , o , t x <o
>· >· <V- Ο,· "}· ">· >· >· <3· <Τ <0· Λ·
INR
Figure 1. Distribution of Patient-Years of Follow-up among the INR Categories.
Follow-up totaled 6475 patient-years. A target INR of 3.6 to 4.8 was used. For 139 patient-years of follow-up, no INR value could be allocated, smce the time between two consecutive
14
THE NEW ENGLAND JOURNAL OF MEDICINE
JulyG, 199^resulting in dependency), the optimal intensity of
anti-coagulation remained the same — an INR of 2.5 to 4.9.
Differences among Subgroups Overall Incidence
Women had a slightly higher risk of both
throm-boembolism and bleeding than men. The incidence of
thromboembolism was 0.8 per 100 patient-years for
women and 0.6 per 100 patient-years for men; the risk
of bleeding was 3.1 per 100 patient-years for women
and 2.4 per 100 patient-years for men.
The risk of thromboembolism was very low among
patients younger than 50: only one patient in this age
group had a thromboembolic event (incidence, 0. l per
100 patient-years; 95 percent confidence interval, 0.0 to
0.5). Patients who were 50 or older had a much higher
risk of thromboembolism than younger patients; the
in-cidence of this complication was 0.8 per 100
patient-years (95 percent confidence interval, 0.5 to 1.1) among
patients between 50 and 69 years of age and 1.1 per
100 patient-years (95 percent confidence interval, 0.6
to 1.8) among those 70 or older. The risk of bleeding
did not vary much with age for those younger than 70
(under 50 years, 2.5 per 100 patient-years; 50 to 69
years, 2.8 per 100 patient-years). For patients 70 or
old-er, the risk of bleeding was twice äs high (5.6 per 100
patient-years).
The overall risk of thromboembolism was also
influ-enced by the position of the valve. The incidence was
0.5 per 100 patient-years among patients with an aortic
valve, 0.9 per 100 patient-years among patients with
a mitral valve, and 1.2 per 100 patient-years among
those with both an aortic and a mitral valve. As
expect-ed, the risk of bleeding did not depend on the position
of the valve.
The overall incidence of thromboembolism was 0.5
per 100 patient-years for the bileaflet valves, 0.7 per
100 patient-years for the tilting-disk valves, and 2.5
per 100 patient-years for the caged-ball and caged-disk
valves. The type of valve did not affect the risk of
bleeding substantially.
To assess whether the differences among the
sub-groups were independent findings that could not be
ex-Table 2. Adverse Events.
iNciDENcr (PER 100 PATIENT-YEARS)* 12.1 Thromboembolism Cerebral infarction Penpheral embohsm Valve thrombosis Any thromboembolism Fatal thromboembolism Bleeding episode
Intracramal and spinal bleeding Extracramal bleeding Any bleeding Fatal bleeding Unclassified stroke First events 43 2 0 45 2 36 128 164 20 14 210 068 003 0 071 003 057 2 1 1 268 030 023 350 INR
Figure 2. Incidence of Ischemic and Hemorrhagic Stroke Accord-ing to INR Category.
cd ω j, c ω Ο o ω o. ω o c ω ;σ Ό 100· 80· 60- 40- 20-Ο
*Since foHow-ιιρ ended when the event of mterest occurred, the denom-mators differ for the various end points
Ix 0> tx „°> t* 0> b, Ο, fr 0> b. <r>
\· \· O,. O,' o,. O,. b.· K,. «,. «,. (Q. n,.
X X X X X X X X X X X ^
v N· 1·· <V 'b· 'b' V V Ό' <r <b· INR
Figure 3 INR-Specific Incidence of All Adverse Events (All Epi-sodes of Thromboembolism, All Major Bleeding EpiEpi-sodes, and
Unclassified Stroke).
The dotted lines indicate the 95 percent confidence interval.
plained by a different age distribution for the various
types and positions of valves, we performed a
multi-variate analysis with Poisson rcgression techniques.
Since the rate ratios we found were equal in a
univari-ate and a multivariunivari-ate model, we concluded that the
ef-fects of age, the position of the valve, and the type of
valve were independent.
Optimal Intensity of Anticoagulation
Vol. 333 No. l OPTIMAL ORAL ANTICOAGULATION FOR PATIENTS WITH MECHANICAL HEART VALVES 15 7-1 co ω O o CO D. O o φ Q. ω o c 0) T3 Όc: co ω .i O O O) Q. 4-S2 6 co ω 4- 3- 2- 1-11 10 9 6 5 4-5?7Q yr _, '
ncidenc
e
2- 1-¥ 50-69 yr <50yr 2.0-2.9 3.0-3.9 4.0-4.9 INR Both Π. _ _ _ • - -D Mitral Aortic 2.0-2.9 3.0-3.9 INR4.0-4.9
s N ^ Caged ball or disk
ω o c CD -σ Ό c 3- 2- 1- n-+ Tilting disk \ .^m m J — - — - Vi Bileaflet 2.0-2.9 3.0-3.9 4.0-4.9 INR
ränge for the INR (Fig. 4). For patients with a higher
risk of thromboembolism (mitral valve position and
older valve types), the optimal level was narrowed on
the left side of the curve; in other words, the patients
with these characteristics had a higher incidence of
thromboembolism at lower intensities of
anticoagula-tion. For patients with both an increased risk of
throm-boembolism and a higher risk of bleeding (those 70 or
older), the optimal level was narrowed on both sides of
the curve. However, confidence intervals for these
in-tensity-specific incidence rates for particular subgroups
overlapped considerably.
DlSCUSSION
Although the need for oral anticoagulant therapy in
patients with mechanical heart-valve prostheses is not
in dispute, the optimal intensity of anticoagulation has
been a matter of debate. The intensity of
anticoagula-tion at which thromboembolism is effectively prevented
without excessive bleeding is not known, even though
anticoagulant agents are given äs a lifelong therapy to
a very large number of patients. It was not even
possi-ble to express the intensity in a standardized way until
1985, when the INR System was introduced."
The only studies in which an attempt has been made
to determine the optimal intensity are randomized
clin-ical trials in which two target levels have been
com-pared.
2"
4Recently, we developed a method that permits
the direct assessmcnt of the relation between the
meas-ured intensities of the anticoagulant therapy and its
benefits and risks.
5With this method, the incidence of
both thromboembolism and bleeding can be calculated
at cvery level of anticoagulation. Application of this
method to the records of more than 1600 patients with
a total follow-up of more than 6000 patient-years
showed that the optimal intensity is that which
produc-es an INR between 2.5 and 4.9. In this ränge, the
inci-dence of all adverse events was about 2 per 100
patient-years. This rate rose sharply both below and above this
ränge — up to an incidence of 75 per 100 patient-years
(95 percent confidence interval, 53 to 101) when the
INR was 6.5 or above.
In order to achieve this INR in actual measurement
of the prothrombin time, a target INR needs to be
de-fined, and to do this the severity of both types of events
must be taken into account. Since the outcome of the
bleeding events in these patients was more severe than
the outcome of the thromboembolic events, it might be
argued that the optimal target level should be on the
lower side of the optimal INR ränge, in order to avoid
Figure 4. Incidence of All Adverse Events at INRs within the INR Range of 2.0 to 4.9, According to Age, Valve Position,
and Valve Type.
The top panel shows the incidence rates for three age groups: younger than 50 years, 50 to 69 years, and 70 years or older. The middle panel shows incidence according to valve position: aortic, mitral, or both aortic and mitral. The bottom panel shows the incidence according to valve type: bileaflet, tiltmg disk, or
16 THE NEW ENGLAND JOURNAL ΟΓ MLDICINE July6, 1995
bleedmg On the other band, the risk of cmbohsm rises
sharply below an INR of 2 5 (Fig 2) Furthcrmore, the
actual INR achieved tends to be somewhat lowei than
the target level (Fig 1), this imphes that the lower limit
of the target INR ränge should be well above 2 5
Therefore, we favor a target INR ränge of 3 0 to 4 0 for
the total patient group
We identified three charactenstics m the patients
that appeared to be associated with a lower risk of
thromboembohsm an age younger than 50, a
prosthet-ic valve m the aortprosthet-ic posiüon, and a bileaflet valve
Pa-tients who were 70 or older had an mcreased risk of
bleedmg To study the possibihty of modifymg the
op-timal mtensity accordmg to these charactenstics, we
calculated the INR-specific mcidence of events for each
subgroup As Figure 4 shows, a higher risk of
throm-boembohsm narrows the optimal level on the left side
of the curve, whereas a higher risk of bleedmg narrows
the optimal level on the right side of the curve Smce
these findmgs are statistically not very stable because
of the hmited numbers of observations, they should be
mterpreted with caution
In 1990, the Bntish Society of Haematology
recom-mended a target ränge of 3 0 to 4 5 for the INR, and m
1992 the American College of Chest Physicians
recom-mended that the target ränge be 2 5 to 3 5
1 2 1 3There
seems to be a tendency to lower these recommended
values, however, on the basis of the rcsults of recent
studies in which lower target ranges were used
2'
4 l7Most of these studies were carned out in patients with
valves m the aortic position and with bileaflet valves
l4 l bAlthough such patients may havc a lower risk of
throm-boembohsm, our data do not support lowermg the
tar-get INR ränge below 2 5 for these groups On the
con-trary, they mdicate that a target of 3 0 for the INR
should be the mimmum, especially for patients with a
mitral valve or an older-model valve
With regard to bleedmg, three rccently pubhshed
studies also showed a clear rclation between the
mten-sity of anticoagulation and the frequency of
hemor-rhage
li! 20In two of these studies, only
prothrombin-ümc ratlos were reported, a fact that prevents the
assessment of the INR at which the risk of bleedmg
be-comes unacceptable
In other recent studies combmation therapy with
coumann derivatives and antiplatelet agents has been
studied
3 4Turpie et al reported a beneficial effect of
addmg aspirm to warfarm, with a target INR of 3 0 to
4 5 for the prothrombm time
3However, the patients
who reccived both drugs still had an mcidence of major
embohsm that was more than twice äs high äs that m
our patients (l 6 vs 0 7 per 100 patient-years) The
m-cidence of bleedmg was also higher 8 5 per 100
pa-ücnt-years, äs comparcd with 2 7 per 100 patient-years
m our study In our opmion, well-momtored
anücoagu-lant treatment is preferable to combmed therapy When
well-momtored anticoagulant treatment is not feasible,
the addition of aspirm might be considered
In conclusion, we found a clear relation between the
mtensity of oral anticoagulation with coumann
deriva-tives m patients wilh mechamcal hcart-valvc
prosthc-scs and the mcidence of thromboembohsm, on the one
hand, and oi bleedmg, on the other The optimal
mten-sity of anticoagulation, producmg the fcwest events of
both types, lies between measurcd INR values of 2 5
and 4 9 As a target ränge, we recommend an INR of
3 0 to 4 0
Wc arc mdcbted to the medical dircctors of the participatms; clm ics (FJM van der Meer of the St Tromboscdienst van de afdeling Leiden c o van hct Nc derlandse Rode Kruis, E van Mctgcn of the St Rode Kruis Tiomboscdienst 's Gravenhage c o , FJW Han nivooit of the St Tromboscdienst Enschede e o in samcnweikme; met htt Ncderlandse Rode Kruis and WG M Brcukmk Engbcrs öl the Trombosedicnst Oost Gclderland) and all the other employees of these centcrs, to Mrs I C Rccuwi]k-Hornung foi data management and assistancc m the preparation öl the manuscript, to N de Gioot of St Infotrom foi ictricvmg data from the computni/cd files of the anticoagulation clinies, and to the staffs of all thc participatmg cen-tcrs and the general practitioncrs
APPENDIX
Vol 3« No l OPTIMAL ORAL ANTICOAGUI ΑΠΟΝ ΓΟ11 PA1IENTS WITH MECHANICA1 HEART VALV1 S 17 G B II W Lamcrs, JWH Lrei, A E Mundus, PM Rozmg, C G
Stcik, R T WM ThomcOi,J Thompson, B J Vcrmeei, R Willemzc, A R Wmt/en, J Zwaitcndi]k, and A Zwavelmg, Diaioneswnhms Leiden Λ A de Rotte, J D Holtzcr, R Vink, IIJ Volku-Dieben, and S A G ] Wiltcvecn St Lhsabtth 7iekenhms, Leidirdorp M Boekhout, E P de Vnes, AR Jansscns, JM II Mooirecs, P C M Rosekians, T L See, WJ van Amstr], G van Rec s, PDJ Vcgt, E M Vioom, and R J W Wittevccn, Sint Anlomushove Leiduhendam JA Bckci, MG Herben, II Holländer, J WM Loeff, A M M Pennmgs, R E van dci Hier, and P Wigt, Antomus Ziekenhms Nieuwegem G API Ascoop, J M P G Einst, R PH M Hamerli|nck, HA van Swictcn, and R G J R A Vandcischueren, Sl Radboud Ziekenhms Nijmegen JS Bös, P J L Davclaai, I TM Huysmans, L K Lacjuct, W A J van Daal, and J J Manni, Acadimisch 7ukenhms Dijkzigt Rotterdam G E Bös, E Bös, JW Deekers, and M J B M \an den Biand, Oogiiekm hms Rotterdam DA E Mcrtcns, Zuiderziekenhms Rotterdam X H Kiaussand R Slmgc>land,AcadermsihZieiinhm<, Utrecht JJ Bredee, H Lubsen, E O Robles de Medma, PC Schölten, and W E Ticffcis, Acadcmisch Zukenhms Velp HA Rebel-de Vnes and J P Staal, Dia concnerihms, Voorburg Π Bijlmer-Gorter, C M G J A Marcoen, I H
Oei, G Slagboom, J R van der Mey, J L van Doorn, E S van Em menk, H Walinga, and JG de Wall, and Stieekziekenhuis Konwgin Bcalnx, WmUnmjk JP de Jong, JP Klaasscn, P Sijbung, JVC Stevens, RAH Stukart, TPM van Thiel, and A A Van
The following gcneral praetitioners also participated m the study Aarlandirveen P K van der Ent Braat, Abbenes PJ Otto, Alphen a/d Rijn JG C M van Engelcn, RJ Hoogendoorn, RJ Oveibeek, M E van de Steen, and R W Vissn,/l»ufcrfl'flm B R van dei Mculen, Bor culo H W Andnessen, De Lier ] Blom and R van Styn, The Hague A Apperloo, R B M R Bakker, E A Baldcr, L K W Beukman, WII BhatosJW BnzaJA Blök, P Boot, A C A Borst, C A E Biandts-BoerefijnJE Burger, WII Chen, WH L China Paw van Wyk,J W Groese, GAT de Beus, JJ Dekktr, RJD de Gioot, APM den Biggclaar, N den Hcycr, X de, Roest, H A Dirkst, J Ehngs, K P Go, NTG Gioot, HG Haitstra, J E P M Hcitkamp, R Hulscbosch, R H J Jamin, A G G van Kaam, R Kanters, A G J van Keulen, A E G Knüttel, H I I M Kianendonk, J A Kuypei, RPM Lemders, R A Lent/e, AJ Lullen, J J H Luykx, PE Mcyei,JJB Nouwens, AJ Ongkiehong, WJ Pameyer, P Pool, M C L Piad]arahard]a, R Rambalclo, P Sanders, R Schuchncr, AJ Sjouke, A M Slagcr, C M Snoeek, A E Sonnenberg, A I L Ste\cns, PJM Stryers, R M Sukul, H J T ten Beige, T H Tjiook, A loet, G S M Trorap, K E van de Beut, PMA van der Biugh, E M M van Extci, H M van Manen, PJ \ an Metnen, A M van Meui s, D E van Wanrooy, D A Vcrhagcn, P R H Veimeulen, l P Weimai,A Woerdman,JD Wuis-tei, D N Wijma, and S Zcilmaker, Den Hildtr G N Ilaasjes, Duren H Polce, Dmxpcrh W Nawyn and GJM Wcitjcns, Doesburg JC Bakx, WJA Besselmk, II Ilecnng, PJM Jaspeis, and JC Tientcl man, Doelinchem J PJ Bei tmg, J J de Bocr, M Draaisma, H F Pas ma, O M Scchteiberger, and A PJ van Haailem, Dnehuis P P L M Molkcnboer, Enschedt H J Biockman, R Pnnscn, and A van di lugt, Gendrmgen A G Mculenkamp and WH M Schattevoct, s Gravenzande J M Baks, R H Dupuis, G S Huizenga, and PD Vis ser, Grocnlo J Cannegicter and J H M van de Holst, Groningen J Koch, Hazerswoudi Dorp AJ Bosch, fiazirswoude Ryndijk J C A Sprangcrs, 's Heeienberg J H A Massop, Hengelo J M Körnen, Hüle gom S C W Keldcrman, Honselersdyk JJ Vadei, T M van der Hock, a n d j van der Stclt, Hummelo J I I G Nicuwcnhuiscn, Katimjk JCJ Loclder, I J W Schoncnbeig, and J H tcn Hove, Laag Kippet B T E M Assmk, Luden A Boels, W de Bruyne, E M M Huibers, I I P JuigensJW Lely,JP Rcmdcis, andj Rus, I nderdorp O Du R) van Beest Holle, J van dci I cdcn, and P van 't Woud, Leidtchendam A V M Pcrqum, JH RunlsJH Schonten, T WJ van de Duft, and M WA Bootsma-van dci Voort, Luhlinvoordt GJM Doesbuig and HJ de Nooyci.Lme l G WM Haasc, JPM Seelelaai, and PJM Steldei, Maarn C MJ Meycis, Made H AM Veimeulen, Monster L E M Klupool, P van der Endt, and E H van Heymngcn-de Lange, Naaldtmjk Z van de Weg, C van Guhk, and H W Vissei, Nuuwkoop J H Boonstra and P Sluis, Nieuw Vtnmp G Smaal, M van Doorn,
a n d j van Nuencn, Nootdwyk I I J L M Immeizeel, NooidwyLet/wut JJ \an de Laak, C JM \an Gueken, and H C M \an Hecsch, Oegsl geut C KL}U,Poddyk PPM Sehyen and l C M Touw, Rotterdam TO B o o n s t i a a n d T C Γ CdiliLi, Rijnsburg ECM Priest« and A J Veihage, Rijswijk NJ Ammeilaan, B Bakker, M W Beuger, H L Holschei, G E W Hoogslag, F H Kersch, C M de Putter, JJ Steen kist, a n d L W N Stern, Sassenluvn H Mouman, fei Am TF Zand süa, Ulß E B J Biemci, WG van Loenen, and AH van de Wal, Utrecht FAM Baien, Vanseveld A E J de Block and JE I ckkerker kei, Voorburg H W Bouman, E Kootte, H A Langendoen, D R Sander, PA Soetci, P van Aikel, and WJM \an Gcnnip, Voorhout J Ilooinweg and J Schmkelshock, Vooisckoten F Koopmans and C H M M Me)cr, Wassuiaar B Dinger\an Krummgcr and M E van der leen, Winlersioyk H Brouwci,JG Klein Hancveld, and F PM Riekhoff, Zeddam H G M Dekkcrs, 7elhem J van Noort, and Zoeler meer F X C Krapels
REFERENCES
1 Cannegieter SC Rosendaal FR Briet E Thromboembolic and bleeding comphcations in patients with mechanical heart valve piostheses Circula tion 199489635 41
2 SaourJN Sieck JO Mamo LAR GallusAS Trial of different mtensities of anticoagulation in patients with prosthetic heart \alves N Engl I Med 1990 322 428 32
3 Turpie AGG Gent M Laupacis A et al A companson of aspinn with pla cebo in patients treated with warfann aftei heart valve replacement N Engl JMed 1993 3295249
4 Altman R Rouvier J Gurfinkel E et al Companson of two levels of anti coaguhnt theiapy in patients with substitute heart valves J Thorac Cardio vascSuig 1991 101 427 31
5 Rosendaal FR Cannegieter SC van der Meei FJM Briet E A method to determme the optimal mtensity of oial anticoagulant therapy Thromb Hae most 1993 69 236 9
6 Duxbuiy BM Theiapeutic quality control leading to further chmcal assess ment of oral anticoagulation Acta Haematol 1986 76 65 7
7 Kent DL Vermes D McDonell M Henikoff J Fihn SD A raodel for plan nmg optimal follow up tor outpatients on warfaun anticoagulation Med De cisMakmg 1992 12 132 41
8 van den Besselaai AMHP van der Meei FJM Gemts Drabbe CW Thera peutic contiol of oral anticoagulant treatment in the Netherlands Am J Clm Pathol 1988 90 685 90
9 Wiegman H Vossepoel AM A computei progiam for long term anticoagu lation conlrol Comput Methods Programs Biomed 1977 7 71 84 10 Vandenbroucke JP A shortcut method for calculatmg the 95 pei cent confi
dencc interval of the standardized moitality ratio Am J Epidemiol 1982 115 3034
11 International Committee for Standardization m Haematology International Committee on Thiombosis and Haemostasis 1CSH/1CTH recommendations for reportmg prothrombm time m oral anticoagulation control Thromb Hae most 1985 53 155 6
12 Hirsh J Dalen JE Deykm D Poller L Oral anticoagulants mechamsm of action chmcal effectiveness and optimal therapeutic ränge Chest 1992 102Suppl312S 326S
13 Stein PD Alpeit JS Copeland J Dalen JE Goldman S Turpie AGG Anti thiombotic therapy m patients with mechanical and biological prosthetic heart valves Chest 1992 102 Suppl 445S 455S
14 DiSesa VJ Colhns JJ Jr Cohn LH Hematological comphcations with the St Jude valve and reduced dose Coumadm Ann Thorac Surg 1989 48280 3
15 Nair CK Mohiuddm SM Hilleman DE et al Ten year results with the St Jude mcdical piosthesis Am J Caidiol 1990 65 217 25
16 Krat? JM Ciawford FA Jr Sade RM Crumbley AJ Stroud MR St Jude pioslhesis for aortic and mitral valve replacement a ten year expenence Ann Thoiac Surg 1993 56 462 8
17 Wilson DB Dünn MI Hassanem K Low mtensity anticoagulation m mc chamcal cardiac prosthetic valves Chest 1991 100 1553 7
18 van der Meer FJM Rosendaal FR Vandenbroucke JP Briet E Bleeding comphcations m oral anticoagulant therapy an analysis of nsk factors Arch Intein Med 1993 153 1557 62
19 Fihn SD McDonell M Martin D et al Risk factors for comphcations of chromc anticoagulation a multicenter study Ann Intein Med 1993 118511 20
20 Hylek EM Singer DE Risk factors for intiacramal hemoirhage in outpa tients takmg warfann Ann Intein Med 1994 120 897 902
THE NEW ENGLAND JOURNAL OF MEDICINE is pubhshed
weekly m the Enghsh language by the Massachusetts Medical Society (Waltham, MA, USA) Material prmted m the Journal is covered by Copyright No part of this reprmt may be reproduced or transmitted in any form without wntten permission All nghts re-served Direct permission requests to the Permissions Depart-ment at the USA subscription Office Editonal Office 10 Shattuck Street, Boston, MA 02115, USA SUBSCRIPTIONS: Subscrip-tion Offices 1440 Main Street, Waltham, MA 02154-1649, USA, 1800 Ironstone Manor, Pickering, Ontano, L1W 3J9, Canada, and c/o E M D GmbH, Zeitschriftenvertrieb, Attn Doreen Haak,