Articles
`
`THE LANCET
`
`Bleeding complications of oral anticoagulant treatment: an
`inception-cohort, prospective collaborative study (ISCOAT)
`Gualtiero Palareti, Nicoletta Leali, Sergio Coccheri, Mario Poggi, Cesare Manotti, Armando D’Angelo, Vittorio Pengo,
`Nicoletta Erba, Marco Moia, Nicola Ciavarella, Gianluigi Devoto, Mauro Berrettini, Serena Musolesi, on behalf of the
`Italian Study on Complications of Oral Anticoagulant Therapy*
`
`Summary
`
`Background Bleeding is the most serious complication of
`the use of oral anticoagulation in the prevention and
`treatment of thromoboembolic complications. We studied
`the frequency of bleeding complications in outpatients
`treated routinely in anticoagulation clinics.
`
`Methods In a prospective cohort from thirty-four Italian
`anticoagulation clinics, 2745 consecutive patients were
`studied from the start of their oral anticoagulation (warfarin
`in 64%, acenocourmarol
`in the rest). The target
`anticoagulation-intensity was low (international normalised
`ratio [INR] 聿2·8) in 71% of the patients and high (> 2·8) in
`the remainder. We recorded demographic details and the
`main indication for treatment and, every 3-4 months, INR
`and outcome events. Such events
`included all
`complications (bleeding, thrombosis, other), although only
`bleeding events are reported here, and deaths. We divided
`bleeding into major and minor categories.
`
`Findings 43% of the patients were women. Nearly three-
`fifths of the patients were aged 60-79; 8% were over 80.
`The main indication for treatment was venous thrombolism
`(33%), followed by non-ischaemic heart disease (17%).
`Mean follow-up was 267 days. Over 2011 patient-years of
`follow-up, 153 bleeding complications occurred (7·6 per
`100 patient-years). 5 were fatal (all cerebral haemorrhages,
`0·25 per 100 patient-years), 23 were major (1·1), and 125
`were minor (6·2). The rate of events was similar between
`sexes, coumarin type, size of enrolling centre, and target
`INR. The rate was higher in older patients: 10·5 per 100
`patient-years in those aged 70 or over, 6·0 In those aged
`under 70 (relative risk 1·75, 95% CI 1·29–2·39, p<0·001).
`The rate was also higher when the indication was peripheral
`
`*Participants listed at end of article
`Cattedra e Divisione di Angiologia e Malattie della Coagulazione,
`Università Ospedale S Orsola, Bologna, Italia (G Palareti MD,
`N Leali MD, S Coccheri MD, M Poggi Bsc); Centro Emostasi, Ospedale
`Regionale, Parma(C Manotti MD); Ambulatorio Emostasi Trombosi,
`IRCCS Ospadale S Raffaele, Milano (A D’Angelo MD); Servizio
`Prevenzione Trombosi, Cettedra di Cardiologia, Università di
`Padova (V Pengo MD); Ambulatorio Emostasi Sezione Trasfusionale,
`Ospedale di Merate (N Erba MD); Centro Emofilia e Trombosi A
`Bianchi Bonomi, Università di Milano (M Moia MD); Servizio di
`Coagulazione, Policlinico, Bari(N Ciavarella MD); Laboratorio Analisi,
`Ospedale di Lavagna (G Devoto MD); Centro Emostasi e Trombosi,
`Istituto Medicina Interna e Vascolare, Università di Perugia
`(M Berrettini MD); e Istituto di Statistica, Università di Bologna
`(S Musolesi BStatSci)
`Correspondence to: Dr Gualtiero Palareti, Cattedra e Divisione
`di Angiologia e Malattie della Coagulazione, Università Ospedale
`S Orsola, 41038 Bologna, Italia
`
`venous
`than
`disease
`cerebrovascular
`and/or
`thromboembolism plus other indications (12·5 vs 6·0 per
`100 patient-years) (1·80, 1·2–2·7, p<0·01), and during the
`first 90 days of treatment compared with later (11·0 vs6·3,
`1·75, 1·27–2·44, p<0·001). A fifth of the bleeding events
`occurred at low anticoagulation intensity (INR<2, rate 7·7
`per 100 patient-years of follow-up). The rates were 4·8, 9·5,
`40·5, and 200 at INRs 2·0–2·9, 3–4·4, 4·5–6·9, and over 7,
`respectively (relative risks for INR>4·5, 7·91, 5·44–11·5,
`p<0·0001).
`
`Interpretation We saw fewer bleeding events than those
`recorded in other observational and experimental studies.
`Oral anticoagulation has become safer in recent years,
`especially if monitored in anticoagulation clinics. Caution is
`required in elderly patients and anticoagulation intensity
`should be closely monitored to reduce periods of
`overdosing.
`
`Lancet 1996; 348: 423–28
`
`Introduction
`Oral anticoagulant therapy is increasingly used for the
`prevention
`and
`treatment
`of
`thromboembolic
`complications of vascular disease.1 Bleeding is the most
`important complication. In a review of observational
`studies,2 average annual rates of fatal, major, and
`major/minor bleeding were 0·8, 4·9, and 15%,
`respectively. In another review,3 bleeding rates ranged
`from 0 to 4·8% for fatal bleeding and from 2·4 to 8·1% for
`major bleeding. Reliable data are lacking on the true
`frequency of complications
`in patients on oral
`anticoagulants because of methodological limitations.4
`Many studies were done before the introduction of the
`international normalised ratio system (INR)
`for
`or
`calculated
`INR
`prothrombin
`time
`(PT),5,6
`retrospectively. Most studies that used the INR system
`were in highly selected patients. The few observational
`studies were either retrospective or descriptive, and were
`not in a clearly defined inception cohort (ie, followed up in
`one clinic from start of treatment).4,7 Observational studies
`that included an inception cohort were retrospective and
`did not use INR8 or selected patients. 9
`We have prospectively assessed the rate of bleeding
`complications
`in outpatients monitored
`from the
`beginning of oral anticoagulation.
`
`Patients and methods
`Centres
`This study was done in thirty-four centres of the Italian
`Federation of Anticoagulation Clinics. Each centre is required to:
`give extensive instructions to all new patients enrolled; follow-up
`patients by INR; fix the date for next visit and meanwhile
`
`Vol 348 • August 17, 1996
`
`423
`
`MYLAN - EXHIBIT 1038
`
`

`
`Males
`
`Females
`
`<20 20–29 30–39 40–49 50–59 60–69 70–79 肁80
`Age (years)
`
`600
`
`500
`
`400
`
`300
`
`200
`
`100
`
`0
`
`Number of individuals
`
`High
`
`791
`532
`259
`
`. .
`. .
`
`630
`408
`222
`
`554
`237
`160
`50
`16
`9
`7
`
`THE LANCET
`
`Demography
`Males
`Females
`Age (years, mean and range)
`Males
`Females
`
`All patients
`
`Target INR
`
`2745
`1561
`1184
`
`Low
`
`1954
`1029
`925
`
`63·6 (8–91) . .
`61·5 (11–93) . .
`
`2011
`1137
`874
`
`Follow-up
`Patient-years
`Males
`Females
`Age (years)
`<70
`肁70
`Withdrawals
`Subjects who moved from centre
`Died
`Males
`Females
`Low INR=聿2·8, high=>2·8.
`Table 1: Demography and follow-up
`
`1779
`966
`829
`134
`102
`54
`48
`
`1381
`729
`652
`
`1225
`729
`669
`84
`86
`45
`41
`
`prescribe daily anticoagulant dose; monitor changes in patients’
`habits, diet, and co-medication, illnesses, bleeding complications
`and scheduled surgical or invasive procedures; and take part in
`external laboratory quality-control.
`
`Figure 1: Distribution of patients by age and sex
`
`Italiano per la Standardizzazione dei Metodi in Ematologia e
`Laboratorio. The thromboplastins used had ISIs under 1·2 in
`twenty centres, between 1·2 and 1·5 in nine, and over 1·5 in five.
`
`Design and patients
`This was an inception cohort study. In each centre, consecutive
`patients receiving for the first time and within 30 days of
`admission either warfarin or acenocoumarol (the only two
`anticoagulant commercially available in Italy) were included,
`independently of age, indication for anticoagulation, intended
`therapeutic range, or expected treatment duration. The two
`exclusion criteria were pregnancy and expected difficulty (usually
`geographic) in obtaining appropriate follow-up.
`Recruitment began in May, 1993, and stopped at the end of
`October, 1994. The observation period started the day of
`inclusion in the study and ended on March 31, 1995, or sooner if
`a major bleeding or thrombotic event occurred, if treatment was
`discontinued for any reason, or if the patient stopped attending.
`For each patient the main indication for oral anticoagulation
`was recorded. The therapeutic ranges recommended by the
`Italian Federation of Anticoagulation Clinics were: venous
`thromboembolism, INR 2–3; non-ischaemic heart disease
`(including atrial fibrillation and cardiomyopathy), 2–3; ischaemic
`heart disease (including coronary bypass surgery or coronary
`angioplasty), 2·5–4·5; cerebral/peripheral artery disease or after
`arterial surgery, 2·5–4; heart valve disease or biological valve
`replacement, 2–3; and prosthetic heart valves, 2·5–4·5.
`
`Data collection and monitoring
`All centres sent records every 3-4 months of all enrolled patients.
`Each centre was required to fill in an admission form, including
`demographic data, indication for anticoagulation (from a
`standard list), drug used, the start day, the targeted therapeutic
`range, important co-diseases, and other drugs. Also collected
`were date of visit, INR, dose (mg per week), date of next visit, log
`of events. Centres were asked to contact the patient, his or her
`family, or doctor if an appointment was missed by 20 days.
`Twenty-five centres used compatible computerised systems for
`results and prescriptions. Sixteen used the same system.10
`The results of all visits were checked by the coordinating centre
`for inconsistencies. When necessary the centres were asked to
`provide further information, mainly to avoid loss of follow-up.
`
`Laboratory monitoring
`Oral anticoagulation was monitored by PT expressed as INR.
`Centres were asked about their thromboplastin reagent used and
`its international sensitivity index (ISI). In Italy, many of the most
`widely used thromboplastin reagents are calibrated by Comitato
`
`Assessment of anticoagulation
`A program provided by F R Rosendaal (University Hospital
`Leiden, Leiden, Netherlands), was used to calculate the observed
`percentage time spent at different INRs. The program11 calculates
`the total number of days accumulated at different INR in intervals
`of 0·5 regardless of target values. Patient-years accumulated in
`INR categories can be calculated for all patients or by
`stratification. The method does not allocate INR when the
`interval between two consecutive measurements exceeds 8 weeks,
`and censors patients after the first outcome event.
`The same program was used to calculate the frequency of
`events at different achieved intensities of anticoagulation (INR<2,
`2–3·4, 3·5–4·4, 4·5–6·9, 肁 7) by dividing the number of events in
`patients with “temporally related” INR in each category by the
`total number of patient-years accumulated in that range. INR was
`defined as temporally related to an outcome event when it was
`obtained at the time of the event or during the preceding 8 days.
`The few outcome events without a temporally related INR were
`excluded from this evaluation.
`
`Venous thromboembolism
`
`Non-ischaemic heart disease
`Dilated cardiomyopathy
`Atrial fibrillation
`Endocavitary thrombosis
`Other
`
`Ischaemic heart disease
`Post-myocardial infarction
`After ACBP or PTCA
`Other
`
`Atrial vascular disease
`Peripheral
`Cerebral
`After vascular surgery
`After peripheral emboli
`Other
`
`Heart-valve prosthesis
`Biological
`Mechanical
`
`Heart-valve disease
`
`Other diagnoses
`
`Total
`
`892 (32·5%)
`
`661 (24·1%)
`136
`462
`24
`39
`
`403 (14·7%)
`144
`135
`124
`
`281 (10·2%)
`48
`93
`80
`44
`16
`
`296 (10·8%)
`34
`262
`
`183 (6·7%)
`
`29 (1·1%)
`
`2745
`
`ACBP=aorto-coronary bypass, PTCA=percutaneous transluminal coronary angioplasty.
`Table 2: Indication for oral anticoagulation
`
`424
`
`Vol 348 • August 17, 1996
`
`

`
`THE LANCET
`
`into two groups: INR targets of 2·8 or lower and over 2·8
`(low and high intensity, respectively). This cutoff was
`chosen on preliminary analysis of the patients’ inclusion
`forms which showed good discrimination. The low-
`intensity group had an intended range between 1·8 and
`3·5, and included 1954 patients (71·2%). The high-
`intensity group (therapeutic range 2·5–4·5) included 791
`patients (28·8%), mainly with coronary, cerebral, or
`peripheral arterial disease and surgery, or mechanical
`heart-valve prosthesis.
`During the study 829 patients (30%) withdrew from
`anticoagulation. In 756 of these cases, the intended
`treatment period was completed.
`In 35 cases,
`anticoagulation was interrupted because of other diseases
`that required different medical and/or surgical treatments.
`Although we cannot exclude that some of the patients who
`discontinued were given other antithrombotic drugs, 38
`patients withdrew from anticoagulants because they were
`recommended by their family doctor or specialist to shift
`to antiplatelet drugs (aspirin, 31; ticlopidine, 7). 134
`patients (4·8%) changed monitoring centre, thus
`terminating follow-up. 30 patients, mostly because they
`were on vacation, spent from 30 to 90 days without being
`monitored. When they resumed follow-up they were
`carefully questioned about possible complications. 102
`patients (3·7%) died during study.
`
`Control of anticoagulation
`The total number of INR results was 51 566, with an
`average time between two measurements of 15 days. The
`number of patient-years at different INR categories was
`calculated for 1980 out of the 2011 patient-years of total
`follow-up (98·4%). We could not allocate INRs to 31
`patient-years.
`In the whole study population, patients were within,
`below, and above therapeutic ranges 68·0, 26·1, and 5·9%
`of the time, respectively. The proportions of time spent
`within and below the therapeutic ranges were significantly
`higher and lower, respectively, in patients with low
`intended anticoagulation intensity (p<0·001). The
`proportion of time below the therapeutic range was
`significantly lower (p<0·05) when thromboplastins with
`low ISI values (<1·2) were used (data not shown). No
`differences were found for sex, age, and anticoagulant
`drug used.
`The quality of anticoagulation treatment control was
`examined in the 141 patients who had bleeding events. In
`
`Minor
`
`Fatal + major
`
`100
`(2239)
`
`300
`200
`400
`(1584)
`(1059)
`(640)
`Days of therapy
`(number of patients at risk in parenthesis)
`
`500
`(292)
`
`600
`(107)
`
`0·10
`
`0·08
`
`0·06
`
`0·04
`
`0·02
`
`0
`
`0
`(2745)
`
`Cumulative frequency
`
`Figure 2: Cumulative frequency (Kaplan-Meier curves) of fatal
`plus major and of minor bleeding events during outpatient
`anticoagulant treatment
`
`All
`
`153 (7·6)*
`
`Fatal (all cerebral, 4 women) 235 (0·25)
`
`323 (1·1)
`
`Major
`7 digestive
`5 ocular (2 with diabetic retinopathy)
`4 cerebral
`3 haemarthrosis
`2 haemoptysis
`1 retroperitoneal
`1 haematuria
`
`125 (6·2)
`
`Minor
`32 haematuria
`25 proctorrhagia
`16 uterine bleeding
`14 gastrointestinal bleeding
`14 haematoma
`13 large bruising
`2 epistaxis
`9 other or multiple sites
`6 with two minor bleeding episodes
`4 with three minor bleeding episodes
`
`*Per 100 patient-years.
`Table 3: Bleeding events
`
`Outcomes
`The occurrence of all types of complications (bleeding,
`thrombosis, other) was recorded, although only bleeding events
`are considered in this report. Deaths for all causes were recorded
`and coded as: bleeding, cardiovascular (acute myocardial
`infarction, stroke), underlying or other diseases, or sudden death.
`We classified major bleeding as:
`fatal (death due to
`haemorrhage); intracranial (documented by imaging), ocular
`(with blindness), articular, or retroperitoneal; if surgery or
`angiographic intervention was required to stop bleeding; and if
`bleeding led to haemoglobin reduction of 2 g/dL or more and/or
`need for transfusion of two or more blood units. Minor bleeding
`was all cases of bleeding not classified as major. Non-relevant
`(small) bleeding was bruising, small ecchymoses or epistaxis,
`occasional haemorrhagic bleeding, or microscopic haematuria.
`In the few cases where this classification was not used, the
`coordinating centre asked for a more detailed description of the
`event and adjudicated the event. If classification of the
`haemorrhagic complication was still uncertain, the event was
`coded as major bleeding.
`
`Statistics
`The SOLO package (version 4.0) was used. Differences between
`groups were assessed by ␹2 or two-sample proportion tests as
`appropriate. Cumulative frequency of fatal, major, and minor
`bleeding events was analysed separately and altogether with the
`Kaplan-Meier method. Data were censored after the first
`bleeding complication, after the cessation of oral anticoagulation,
`or when a patient stopped being monitored. The independent
`effect of possible risk factors was investigated with Poisson
`regression.
`
`Results
`2745 patients (43% female) were included (table 1).
`Twenty-five centres enrolled under 100 cases, and nine
`enrolled over 100. In the centres that enrolled over 100
`patients, the number of anticoagulated patients not
`included in the study (second anticoagulation course,
`anticoagulation started after 30 days, geographic
`inacccessibility) ranged from 6 to 11% of the patients
`included. The anticoagulant drug used was warfarin in
`1752 (63·8%) and acenocoumarol in the remaining 993
`(36·2%) patients. During the study, 22 patients changed
`from warfarin
`to acenocoumarol and 28
`from
`acenocoumarol to warfarin. Most patients were aged
`between 60 and 79 (57·8%) but 8% were over 80 (figure
`1). The mean length of follow-up was 267 days (range
`5-660).
`The most frequent indication for oral anticoagulation
`was venous thromboembolism, followed by non-ischaemic
`heart disease (mainly atrial fibrillation (table 2). To check
`the variety of the therapeutic ranges and target values in
`the participating centres (therapeutic ranges that
`sometimes differed from those recommended by our
`federation), we evaluated the effects of the intended
`anticoagulation intensity by dividing the study population
`
`Vol 348 • August 17, 1996
`
`425
`
`

`
`Patient-years of
`follow-up
`
`Bleeding
`
`Fatal
`
`Major
`
`Minor
`
`⫻100 patient-years
`
`Relative risk (95% CI)
`
`THE LANCET
`
`Sex
`Female
`Male
`
`Age (years)
`<50
`50–69
`肁70
`
`Relative risk 肁70 vs70
`
`Indication
`Venous thromboembolism
`Arterial disease
`All others
`
`Relative risk arterial disease vsothers
`
`Centres
`<100 recruited
`>100 recruited
`
`Coumarin
`Acenocoumarol
`Warfarin
`
`Target INR
`聿2·8
`>2·8
`
`Temporally related INR (not available)
`<2
`2–2·9
`3–4·4
`4·5–6·9
`肁7
`
`Relative risk values 肁4·5 vs<4·5
`
`Timing of events (days)
`聿90
`>90
`
`874
`1137
`
`288
`997
`726
`
`558
`223
`1230
`
`820
`1191
`
`753
`1258
`
`1381
`630
`
`377
`1116
`442
`42
`3
`
`566
`1445
`
`4
`1
`
`0
`0
`5
`
`1
`1
`3
`
`2
`3
`
`2
`3
`
`3
`2
`
`2
`0
`1
`2
`0
`0
`
`1
`4
`
`10
`13
`
`0
`7
`16
`
`7
`5
`11
`
`6
`17
`
`7
`16
`
`17
`6
`
`1
`6
`8
`3
`2
`3
`
`9
`14
`
`61
`64
`
`20
`50
`55
`
`30
`22
`73
`
`46
`79
`
`58
`67
`
`94
`31
`
`2
`23
`45
`37
`15
`3
`
`52
`73
`
`8·6
`6·8
`
`6·9
`5·7
`10·5
`
`6·8
`12·5
`7·1
`
`6·6
`8·3
`
`8·9
`6·8
`
`8·2
`6·2
`
`7·7
`4·8
`9·5
`40·5
`200
`
`11·0
`6·3
`
`1·25 (0·91–1·71)
`
`1·75 (1·28–2·39, p<0·001)
`
`1·80 (1·20–2·70, p<0·01)
`
`0·79 (1·10–0·57)
`
`1·30 (0·95–1·79)
`
`0·75 (1·08–0·52)
`
`7·91 (5·44–11·5, p<0·0001)
`
`1·75 (1·27–2·44, p<0·001)
`
`Relative risk 聿90 vs>90
`Relative risks are univariate.
`Table 4: Bleeding events stratified by risk factors
`
`these patients the proportion of time within, below and
`above therapeutic ranges was 66·3, 24·1, and 9·6%,
`respectively; the difference in this distribution compared
`with the distribution in the whole study population was
`not statistically significant.
`
`Bleeding complications
`Bleeding events are detailed in table 3 and figure 2. The
`rates of bleeding events were not different according to
`sex, coumarin type, size of enrolling centre, and target
`zone (table 4). However, the rate was higher in older
`patients and when the indication for anticoagulant
`treatment was arterial disease. Among these patients,
`bleeding was frequent in those with cerebrovascular
`disease (n=107: 2 major events [1 fatal] both intra-cranial;
`10 minor events, 14·5 per 100 patient-years of follow-up)
`or peripheral emboli (n=44: 2 major and 6 minor events,
`21·6 per 100 patient-years). The risk of haemorrhagic
`events during therapy was higher during the first 90 days
`of treatment (table 4).
`The frequency of bleeding events at different achieved
`intensities of anticoagulation was investigated by dividing
`the number of events in patients with temporally related
`INR in five increasing INR categories by the total number
`of patient-years accumulated in these categories (table 4).
`Many bleeding events (29 out of the 147 [20%] with
`available related INR) occurred at low anticoagulation
`intensity. However, in 4 of these 29, the low INR on the
`day of the event had been preceded (within 3-10 days) by
`value over 4·5, indicating that erratic anticoagulation may
`have been a cause of bleeding in these cases. The rate of
`
`426
`
`bleeding was significantly lower (p<0·05) in the 2·0–2·9
`INR category which had the lowest frequency of events.
`With further increase in INR, there was an increase in
`bleeding. Multivariate analysis confirmed that the risk of
`bleeding was higher when INR exceeded 4·5, when
`arterial disease was the indication for anticoagulation, and
`during the first 90 days of treatment (table 5).
`During the whole follow-up, 70 thrombotic events (20
`fatal, 39 major and 11 minor) occurred in 67 patients, 5 of
`whom also had bleeding. The rate of thrombotic
`complications was 3·5 per 100 patient-years of treatment.
`About one-third of the patients who had bleeding
`complications (46/141) had more than one indication for
`oral anticoagulation. Besides the main one, there was:
`peripheral and/or cerebral arterial disease (n=22),
`ischaemic heart disease (8), and atrial fibrillation and
`venous thromboemobolic disease (6 each). At least one
`other disease or risk factor was present at the start of
`treatment in 78 of 141 who had bleeding (table 6). In a
`few cases we could correlate the occurrence of a bleeding
`episode with onset of specific pro-haemorrhagic
`conditions, such as trauma (1 major and 5 minor
`
`Sex (women vsmen)
`Age (肁70 vs<70 years)
`Target INR (聿2·8 vs>2·8)
`Indication (arterial disease vsothers)
`Actual INR (肁4·5 vs<4·5)
`Coumarin type (acenocoumarol vswarfarin)
`Timing of events (聿90 vs>90 days)
`
`Table 5: Multivariate risk ratios
`
`Relative risk (95% CI)
`
`1·21 (0·86–1·70)
`1·69 (1·21–2·37, p<0·001)
`0·83 (0·56–1·22)
`1·72 (1·17–2·54, p<0·001)
`5·96 (3·68–9·67, p<0·0001)
`1·20 (0·85–1·69)
`2·5 (1·4–3·3, p<0·001)
`
`Vol 348 • August 17, 1996
`
`

`
`Hypertension
`Cancer
`Diabetes
`Renal and bladder disease
`History of gastrointestinal bleeding
`
`Gastrointestinal disease
`Renal insufficiency
`Lung disease
`Lupus-anticoagulant/anticardiolipin-antibody
`Chronic liver disease
`
`Cardiac insufficiency
`Uterine fibromatosis
`Thrombocytopenia
`von Willebrand’s disease
`Psychiatric disease
`
`Number
`
`25
`20
`10
`7
`6
`
`5
`4
`4
`3
`2
`
`2
`2
`1
`1
`1
`
`Table 6: Other diseases and risk factors at start of oral
`anticoagulation in 78 of 141 patients with bleeding
`complications
`
`bleedings), urinary infections and/or nephrolithiasis (4
`minor), heparin co-administration (2 minor), throm-
`bocytopenia and lung disease (1 minor each). Finally,
`cancer was diagnosed in 2 patients after minor events.
`Except in venous thromboembolism, many of the other
`patients received more than one treatment. In 77 of 153
`cases drugs other than coumarins were administered near
`to the bleeding event, mostly antihypertensive drugs (2
`fatal, 6 major and 34 minor), vasodilators and nitrates (11
`minor), aspirin or other antiplatelet drugs (2 major, 9
`minor), amiodarone (6 minor), allopurinol (4 minor),
`cyproterone acetate (2 major, 1 minor), and antidiabetics
`(3 minor). Minor bleeding occurred a few days after
`withdrawal of rifampicin or barbiturates (1 case each).
`During the study, 102 patients died, 5 due to bleeding
`events. The causes of death other than bleeding were:
`cancer (34), acute myocardial infarction (6), heart failure
`(22), sudden death (12), post-surgery complications (4),
`non-specified cardiovascular events (4), ischaemic stroke
`(3), pulmonary embolism (7, 5 of whom highly probable
`and 2 confirmed by necropsy), sepsis (2), acute hepatitis
`(1), and respiratory insufficiency (1).
`
`Discussion
`We were able to estimate haemorrhagic risk complications
`during oral anticoagulation in outpatients monitored by
`INR in specialist anticoagulation clinics. The results
`reflect the normal practice of Italian centres. The rate of
`fatal, major, and minor bleeding events was 0·25, 1·1, and
`6·2 per 100 patient-years of follow-up, respectively. These
`figures are lower than the average annual frequencies of
`bleeding (0·8, 4·9, and 15 per 100 patient-years for fatal,
`major, and major/minor bleeding, respectively) in a
`review2 of studies of similar design, and lower than those
`in experimental studies (0·4, 2·4, and 8·5 per 100 patient-
`years).2 Levine et al,3 reviewing randomised controlled
`trials, reported rates ranging from 0 to 4·8 per 100 patient-
`years for fatal and 2·4 to 8·1 per 100 patient-years for
`major bleeding. The rates of bleeding we found are also
`lower than those in studies in which patients were
`routinely treated with oral anticoagulation for various
`indications.12-15 Others have reported bleeding rates higher
`than7,16,17 or similar9 to ours.
`van der Meer et al4 found higher rates of major bleeding
`events than we did, probably because of higher
`anticoagulation intensity in the patients. The low bleeding
`rate we recorded, most likely due to the moderate
`
`THE LANCET
`
`treatment intensity, did not seem to be counterbalanced
`by more frequent thrombotic complications, since the rate
`of such complications was actually lower (3·5 per 100
`patient-years) than that in other studies (5·9 to 9·5).7,17
`The use of INR increases the reliability of anticoagulant
`control and makes possible inter-study comparisons.18
`Unnecessarily high doses, associated with higher bleeding
`rates, can be avoided19 and optimum therapeutic ranges
`can be more easily achieved.20 In line with the results of
`some4 but not all3 studies, we found no relation between
`risk of bleeding and target zone. However, intensity of
`anticoagulation achieved was related to bleeding. About
`one-fifth of all bleeding events occurred at very low INR
`(<2). This confirms other reports21, 8 that many bleeds
`during oral anticoagulation are not related to the intensity
`of anticoagulation but to a local bleeding source that may
`be unmasked by anticoagulant therapy. A slight but
`significant increase of risk was recorded for INRs of
`3·0–4·4, the risk of bleeding becoming much higher for
`values over 4·5. Similar results are suggested by other
`studies.7,22 Based on these findings it would be prudent to
`avoid INR of 4·5 or more.
`It is debatable whether the risk of bleeding during oral
`anticoagulation is higher in older patients.3,23 In our study
`patients over 70 had a relative risk of 1·75 compared with
`all the others. Similar results have been reported,4,8,18,22
`although not by Fihn et al.7 Our results also indicate a
`relation between older age and intracranial bleeding, as
`was reported by Landefeld and Goldman.8 In our study we
`found no relation between intracranial bleeding and
`achieved anticoagulation.
`Although most physicians are aware of the higher risk of
`oral anticoagulation in the elderly,24,25 an increasing
`number of elderly patients are treated with anticoagulants.
`Older patients on anticoagulants should be: treated at a
`low target zone; monitored closely to keep their INRs
`within the therapeutic zone; and carefully followed up so
`that conditions which may
`interfere with oral
`anticoagulation can be monitored.
`In our study more than one-third of all bleeding
`episodes occurred within the first 90 days of each
`anticoagulant course; the frequency of bleeding then
`stabilised. A higher frequency of bleeding early in the
`course has been reported in many7,8,14 but not all26,27
`studies. Several factors may contribute to the increased
`risk of early bleeding. First, anticoagulant therapy can
`unmask a cryptic lesion. Second, dose adjustment may be
`less well-controlled at the start of treatment. As clearly
`pointed out by Landefeld and Goldman,8 studies that
`examine non-inception cohort and/or include patients who
`have resumed anticoagulation are likely to underestimate
`the true risk of bleeding by either missing early events or
`excluding from any second course patients who had bled
`in the first course.
`Our patients on oral anticoagulation for arterial vascular
`disease had a higher frequency of bleeding (12·5 per 100
`patient-years of follow-up) than the others; the rate of
`bleeding was even higher if cerebrovascular patients were
`considered alone (14·5). Since the arterial vascular disease
`was also the most frequent secondary indication in the 22
`patients who bled, this indication whether main or
`secondary was most frequently associated with bleeding
`(50 patients out of 141). These results confirm the
`particularly high risk of oral anticoagulants in patients with
`arterial disease, especially cerebrovascular disease,
`recorded in experimental trials3 or observational studies,8
`
`Vol 348 • August 17, 1996
`
`427
`
`

`
`THE LANCET
`
`and raise the question of whether the risk of bleeding
`during anticoagulation outweighs the benefit in such
`patients28, 29
`Finally, the quality of anticoagulation obtained over the
`whole study was high (68% of all the period was within the
`therapeutic ranges), especially given that thirty-four
`centres did the monitoring with a wide range of
`thromboplastins. The quality of treatment was higher for
`patients in the low target zone and when thromboplastins
`with low ISI were used. This result, consistent with our
`previous findings,18 supports the switch from low-
`sensitivity to high-sensitivity reagents.
`
`The following investigators and centres participated in ISCOAT.
`S Coccheri, G Palareti (chairman), M Poggi, N Leali, Catteda e Divisione
`di Angiologia e Malattie della Coagulazione, Università di Bologna
`(coordinating centre, enrolled 240 patients). C Manotti, R Quintavalla,
`Centro Emostasi, Ospedale Regionale, Parma (291); A D’Angelo,
`L Crippa, Ambulatori Emostasi Trombosi, IRCCS Ospedale S, Raffaele,
`Milano (228); V Pengo, Servizio Prevenzione Trombosi, Cattedra di
`Cardiologia, Università di Padova (184); N Erba, D Restifo, Ambulatorio
`Emostasi Sezione Trasfusionale, Osepdale di Merate (173); M Moia,
`P Bucciarelli, Centro Emofilia e Trombosi A Bianchi Bonomi, Università
`di Milano (142); N Ciavarella, C Ettorre. Servizio di Coagulazione,
`Policlinico, Bari (137); G Devoto, Laboratorio Analisi, Ospedale di
`Lavagna (129); M Berrettini, F Poeta, Centro Emostasi e Trombosi,
`Istituto Medicina Interna e Vascolare, Università di Perugia (111);
`G Ballerini, Servizio di Fisiopatologia della Coagulazione, Arcispedale
`S Anna, Ferrara (99); F Baudo, Divisione di Ematologia, Ospedale
`Niguarda, Milano (95); F Veschi, Laboratorio Analisi Ospedale
`Pediatrico Apuano, Massa (95); G Piseddu, Centro Emostasi e Trombosi,
`Servizio Trasfusionale, Sassari (85); S Testa, Laboratorio Analisi Istituto
`di Patologia Clinica, Odpedale di Cremona (68); M Molinatti, Servizio di
`Immunoematologia, Ospedale Maria Vittoria, Torino (67); L Frigerio,
`Laboratorio Analisi e Centro Trombosi, Ospedale Valduce, Como (61);
`M Pagliarino, Servizio di Immunoematologia, Ospedale di Ivrea (57);
`L Ria, Centro Emostasi e Angiologia Medica, Medicina Interna, Ospedale
`di Gallipoli (52); L Gatti, Centro Trasfusionale, Istituti Clinici Milano
`(36); G Malcangi, Servizio Trasfusionale, Ospedale di Molfetta (36);
`E Rossi, Centro Emostasi e Trombosi, Ospedale di Cosenza (35);
`C Agazzi, Laboratorio Analisi, Ospedale di Seriate (33); F Fusco,
`Laboratorio Emocoagulazione, Ospedale di S Vito Al Tagliamento (32);
`G Bazzicalupo, Laboratorio Analisi, Ospedale di San Secondo Parmense
`(28); F Corbara, Servizio di Cardiologia, Ospedale di Este (27); D Prisco,
`Clinica Medica Generale e Cardiologia, Università di Firenze (27);
`S Guarino, Emostasi e Trombosi, Divisione di Ematologia, Ospedale di
`Latina (27); E Tiraferri, Centro Angicoagulati, Ospedale di Rimini (23);
`V Brancaccio, Divisione Ematologia, Ospedale Cardarelli di Napoli (22);
`V Rocco, Laboratorio Patologia Clinica, Ospedale di Benevento (22);
`G Labò, Laboratorio Analisi, Ospedale di Castelnuovo Monti (21);
`F Marongiu, Istituto Medicina Interna, Università di Cagliari (21);
`L Steidl, Clinica Medica, Università di Pavia, Ospedale Multizonale di
`Varese (21); R Del Bono, II Laboratorio Analisi, Ospedale di Brescia (20)
`and S Musolesi, Department of Statistical Science, University of Bologna.
`Acknowledgements
`We thank Jack Hirsh for helpful suggestions about study design and
`revision of the manuscript, Frits R Rosendaal for supplying software, and
`Stephen Jewkes for revising language. This work was partly supported by a
`grant from MURST.
`
`References
`1 Hirsh J. Drug Therapy: oral anticoagulant drugs. N Engl J Med 1991;
`324: 1865–75.
`2 Landefeld CS, Beyth RJ. Anticoagulant-related bleeding-clinical
`epidemiology, prediction, and prevention. Am J Med 1993; 95:
`315–28.
`3 Levine MN, Hirsh J, Landefeld CS, Raskob G. Haemorrhagic
`complications of anticoagulant treatment. Chest 1992; 102: 352s–63s.
`4 van der Meer FJM, Rosendaal FR, Vandenbroucke JP, Briet E.
`Bleeding complications in oral anticoagulant therapy-an analysis of risk
`factors. Arch Intern Med 1993; 153: 1557–62.
`
`5 WHO Expert Committee On Biological Standardization, 33rd Report.
`Geneva: WHO. Tech Rep Ser 1983; 687: 81–105.
`6 Loeliger EA. ICSH/ICTH recommenda