Content uploaded by Josef G. Brecht
Author content
All content in this area was uploaded by Josef G. Brecht on Feb 15, 2016
Content may be subject to copyright.
Venous thromboembolism(VTE)inEurope
The number of VTE events andassociated morbidity andmortality
Alexander T. Cohen1 ,Giancarlo Agnelli2,Frederick A. Anderson 3,Juan I. Arcelus4,David Bergqvist5 ,Josef G. Brecht6,
Ian A. Greer7 ,John A. Heit8,Julia L. Hutchinson 9,AjayK.Kakkar10,Dominique Mottier11,Emmanuel Oger11,
Meyer-Michel Samama12,MichaelSpannagl13 fortheVTEImpact Assessment Group in Europe (VITAE)
1King’sCollege Hospital,London, UK;2Division of Internal and Cardiovascular Medicine,University of Perugia, Perugia, Italy; 3Center for
Outcomes Research, University of Massachusetts MedicalSchool,Worcester,MA, USA; 4University of GranadaMedicalSchool and Hospital
SanJuan de Dios, Granada, Spain; 5 Department of Surgical Sciences, University Hospital,Uppsala, Sweden;6InForMed GmbH–Outcomes
Research andHealth Economics,Ingolstadt, Germany;7Department of Obstetrics and Gynaecology, University of Glasgow,Glasgow Royal
Infirmary, Glasgow,UK; 8HematologyResearch, Mayo Clinic College of Medicine,Rochester,MN, USA; 9Fourth Hurdle Consulting,London,
UK;10Centrefor SurgicalSciences, Bartsand The London-QueenMarySchool of Medicine,London, UK and Thrombosis Research Institute,
London,UK; 11Groupe d’Etude de la Thrombose de BretagneOccidentale (GETBO), HôpitaldelaCavale Blanche,Brest,France;12Hôtel
Dieu, Départementd’HématologieBiologique,Paris,France;13Ludwig-MaximilliansUniversity Munich, Klinikum derUniversität, Abteilung
Haemostasiologe,Munich, Germany
Summary
Venous thromboembolism(VTE)isoften asymptomatic, mis-
diagnosed,andunrecognized at death,and thereisalack of rou-
tine postmortem examinations.Thesefactors arethought to re-
sult in marked underestimatesofVTE incidence.The objectiveof
our study was to estimate the total burden of VTE within the
European Union (EU)per annum.An epidemiological modelwas
constructedtoestimatethe number of community-and hospi-
tal-acquired incidents and recurrent cases(attack rate)ofnon-
fatalVTE and VTE-relateddeaths, as well as incidentand preva-
lent casesofpost-thromboticsyndrome (PTS)and chronic
thromboembolic pulmonaryhypertension (PH) occurring in the
EU perannum.Individual models were developed forsix EU
countries.Themodels were populated with datafrompublished
literatureand,where necessary, expert opinions. Thefindings
were tested using probabilistic sensitivity analyses. Theesti-
Keywords
Deep-veinthrombosis,post-thromboticsyndrome,pulmonary
embolism, pulmonary hypertension,venous thromboembolism,
VTE prophylaxis
matedtotal number of symptomaticVTE events (range basedon
probabilistic sensitivityanalysis)per annum within thesix EU
countrieswas 465,715(404,664–538,189) casesofdeep-vein
thrombosis, 295,982 (242,450–360,363) cases of pulmonary
embolism(PE), and 370,012(300,193–483,108) VTE-related
deaths.Ofthese deaths,anestimated 27,473 (7%) were diag-
nosed as beingantemortem;126,145 (34%) were sudden fatal
PE,and 217,394 (59%) followedundiagnosedPE.Almost three-
quarters of allVTE-relateddeaths were from hospital-acquired
VTE.VTE is amajor health problem in the EU,with over one mil-
lionVTE events or deathsper annum in thesix countriesexam-
ined. Giventhe availability of effectiveVTE prophylaxis,manyof
these events and deathscouldhavebeen prevented.Thesere-
sults have important implications forthe allocationofhealthcare
resources.
ThrombHaemost 2007; 98: 756–764
Blood Coagulation, Fibrinolysis and CellularHaemostasis
Correspondence to:
Alexander T. Cohen
VascularMedicine,Department of Surgery
King’sCollege Hospital
London, SE5 9RS, UK
Te l.: +44207346 3015, Fax: +44 20 7346 3927
E-mail: alexander.cohen@kcl.ac.uk
Received March 21, 2007
Accepted after resubmission July 25, 2007
Prepublished onlineSeptember 10,2007
doi:10.1160/TH07–03–0212
Introduction
Venous thromboembolism (VTE),comprising deep-vein throm-
bosis (DVT) and pulmonaryembolism(PE), is associatedwith
substantialmorbidity and mortality.Detailedestimatesofthe an-
nual number of VTE eventsare hard to obtain becauseVTE is
difficulttodiagnose.This is due to anumber of factors; VTEis
oftenclinically silent and,inmanycases,the first sign of the dis-
easeisasuddenfatal PE (1, 2). Despite modest increases in ante-
mortemdiagnosis of PE over the years,lessthan half of autopsy-
detectedPEcases are diagnosed antemortem(3). The lackof
routinely performed postmortem examinations means thatmany
©2007 Schattauer GmbH,Stuttgart
756
Cohen et al.Venous thromboembolisminEurope
757
fatalpulmonaryemboli remainunrecognized, leading to an
underestimation of their incidence. Previous analyses have
shown that, forevery caseofsymptomatic non-fatalPE, there are
2.5 cases of fatalautopsy-detectedPE(4). Only29% of patients
whosurvive an initialembolic event are diagnosed withVTE(5).
In 36% of patients diagnosedwith PE, an initialdiagnosis is
made on the date of deathorpostmortem, with PE being the
cause of death in 53% of these cases (2).Studieshavealso shown
thatasymptomatic DVTisstrongly associatedwith the develop-
ment of symptomatic VTE(6–8) and is also associated with an
increased risk of death (9).Due to these factors, modeling is the
onlyapproach to achieve annual incidencefigures and estimates
of complications.
The chronic nature of VTE and itsrecurrences and compli-
cations requires considerablehealthcare resources forits man-
agement.Additionally, morbidity and healthcare costs are in-
curred fromassociatedcomplications of VTE, such as post-
thromboticsyndrome (PTS), whichaffectsatleastone-third of
DVTpatients (10–17), and pulmonary hypertension (PH), which
occurs in 4%–5% of patients following PE (18, 19). Data from
the UK suggest that the total cost of VTE to the National Health
Service in 1993 was£235–£257 million ( e 349–e 382 million)
(20),and estimates of the combined direct and indirectcosts are
nowplaced at approximately £640 million ( e 950 million) (21).
The costs arefurther increased when long-termcomplications
such as PTS aretaken into consideration (22, 23).Therefore,itis
important for healthcare decision-makers to have reliable esti-
matesofthe total annualnumber of incident and recurrent VTE
eventsper year.
To date,the most comprehensive epidemiological datahave
beengenerated fromstudiesofspecific populations. US popu-
lation-based studies of defined geographical areas,which in-
cludedpatients with onlyincidentVTE at home or in hospital, re-
vealed aVTE incidenceof71–117 per 100,000 person-years
(24–26).One study also showedathirty-dayrateofVTE recur-
rence of 4.8% (26). In Europe, studiesinWestern France (27)
Event estimated Source of dataProbability(range) of event
Community-acquired events
Age and gender specific
Country-specific population data Oger 2000 (27)
US CensusBureau International
Database (28)
--
Hospital-acquired events
At-risk populationsHES (29) and PMSI (30)databases -
Surgical risk group
DVTand PE without prophylaxis
DVTand PE with prophylaxis Geerts et al. 2004 (31)
Derived from Geerts et al. 2004 (31) --
Medicalrisk group
DVTwith and without prophylaxis
PE with and without prophylaxis Geerts et al. 2001 (32)
Derived from Geerts et al. 2001 (32) --
PTS
Mild/moderate PTSinpatientwith DVT
Severe PTSinpatientwith DVT(new)
Derived from
Prandoni et al. 1996 (10)
Prandoni et al. 1996 (10)
0.147 (0.074–0.221) first year,
0.01 (0.005–0.015) in following years
0.026 (0.013–0.039) first year,
0.017 (0.009–0.026) in following years
PH Pengoetal. 2004(18) 0.31 (0.16–0.47) first year,
0.38 (0.19–0.57) in following years
DVT
Symptomatic and diagnosed
Death from DVT
Medical DVTthat has developed
into PE
Geerts et al. 2004 (31);
Piovellaetal. 2005(35)
Osteretal. 1987(33)
Osteretal. 1987(33)
0.10 (0.05–0.15)
0.006 (0.003–0.009)
0.115 (0.058–0.173)
PE
Symptomatic and diagnosed
SuddendeathfromPE
Death from undiagnosed PE
Death from diagnosedPE
Oster et al. 1987 (33)
Oster et al. 1987 (33)
Oster et al. 1987 (33)
Oster et al. 1987 (33)
0.29 (0.15–0.44)
0.11 (0.06–0.17)
0.30 (0.15–0.45)
0.08 (0.04–0.12)
DVTrecurrence
PE recurrence
Prandoni et al. 1996 (10);
Heit et al. 2000(34)
Prandoni et al. 1996 (10);
Heit et al. 2000(34)
0.104 (0.052– 0.156) first year,
0.02 (0.01–0.03) in following years
0.025 (0.013–0.038) first year,
0.005 (0.003–0.008) in following years
DVT, deep-veinthrombosis; HES, HospitalEpisodesStatistics; PE, pulmonaryembolism; PH,pulmonary hypertension; PMSI, Le Programme de Médicalisation
des Systèmesd’Information; PTS, post-thrombotic syndrome.
Table1:Sourcesofdata.
and Sweden (5), thatincludedboth incident and recurrent VTE
events, reported an overallVTE incidenceof160–180 per
100,000 person-years. Although these studies have contributed
considerablytoour understanding of the epidemiologyofVTE,
theyunderestimate the total burden of the diseaseonthe health-
care system because theydonot include undiagnosedormis-
diagnosedclinical VTEevents or unrecognizedVTE-related
deaths.
In view of the potential impact of VTE to public health and
the lackofcomprehensive estimatesonthe number of sympto-
matic VTE eventsand associated morbidityand mortality in Eu-
rope,theVTE Impact Assessment Group in Europe (VITAE) de-
signed amodified incidence-basedmodel to estimate the total
annualnumber of non-fatalincident and recurrent VTE events,
associatedcomplications (PTS and PH), and VTE-related deaths
in six countries within the European Union (EU).
Methods
Data searches
An extensive search of nationaldatabases wasperformed look-
ing at publishedpapersand grey literature. Thesearch wascon-
ductedusing MEDLINE and EMBASE databasesbetween 1994
to date.Seventy-five original papers were identifiedthat de-
scribed newepidemiologicalresearch.Onlythose giving infor-
mation about the prevalenceand incidenceinEurope in defined
geographical populations were used in the model. Keysearch
termswere “venous thromboembolism”,“deep-vein thrombo-
sis”, and “pulmonaryembolism”.These terms were crossed with
the terms“epidemiology”,“prevalence”, “incidence”, “risk fac-
tors”,and “naturalhistory”.
Modelstructure
Amodel wasdeveloped to estimate the number of country-spe-
cific,non-fatalsymptomatic VTE eventsand VTE-related
deaths in aone-year period. Eventsestimatedincludedincident
and recurrent cases of non-fatalVTE, and incident and existing
cases of PTS, PH,and VTE-related death. The base year fores-
timating eventswas 2004.
The core model structure, methodologies, and assumptions
were validatedbyanadvisoryboardconsisting of VTE experts
from across Europe and the USA. Individual models were devel-
opedfor six EU countries: France, Germany,Italy,Spain,
Sweden, and the UK. Whereverpossible, the parameter esti-
mates for models were derivedfrom datainthe published litera-
ture (Table 1) (10, 18, 27–35). Where data on model parameters
were unavailable, the advisoryboardprovided an expertopinion
on the plausible ranges for sensitivity analyses. Published data
were used as asourcefor all event rates,for the at-risk hospital
population, and for the probabilities of events. Estimatesofpro-
phylaxisrates fromexperts in the individual countries were only
used when publisheddatawere notavailable. Wheretherewere
anydifferences, or disagreements arose,experts were asked to
reach aconsensus.
The model estimated the number of community-acquired
VTE eventsand hospital-acquiredVTE events. Hospital-ac-
quired eventswere definedasevents that occurredfollowing ex-
posure to hospital-related risk factors, surgical procedures,or
admission to amedical ward,and whichoccurred within ahospi-
tal setting or in the community within a90-dayperiod of admis-
sion to hospital. Eventsnot meeting these criteria were classified
as community-acquired. Thegeneral modelstructureapplies to
both community- and hospital-acquiredevents (Fig. 1).
Estimation of the numberofcommunity-acquired
events
Community-acquired eventsfor each of the six EU countries
were estimatedonthe basis of alarge French epidemiological
study (the Groupe d'Etude de la Thrombose de Bretagne Occi-
dentale [EPI-GETBO] study) (27) with age-and gender-specific
incidencerates appliedtocountry-specific population data.The
authorsofthe EPI-GETBO study (27) re-analyzedtheir study
and provided additional data.Theyseparated the eventsthat took
placeinthe community occurring within threemonths of hospi-
Figure1:General model
structurefor estimating
the number of country-
specific,non-fatal sympto-
maticvenousthromboem-
bolism (VTE) events and
VTE-related deaths in a
one-year period. *Treated
deep-veinthrombosis(DVT)
and pulmonaryembolism (PE)
(for thepurpose of themodel,
we assume all diagnosedVTE is
treated). PH, pulmonaryhy-
pertension;PTS, post-throm-
boticsyndrome.
Cohen et al.Venous thromboembolisminEurope
758
talization (16%ofthese events). As in the VITAEstudy these
would not constituteacommunity-acquired event.The re-analy-
sis of the EPI-GETBO study (27) also allowedthe calculation of
age-adjusted numbersofevents for community-acquiredVTE.
Estimation of the numberofhospital-acquiredevents
Patients were includedinthe hospital-acquiredevents model if
theywere deemed to be at risk of developingVTE (the at-risk
population), and met the aforementioned criteria. At-risk popu-
lations were estimatedonthe basis of patient numbersderived
from the HospitalEpisodes Statistics (HES) database (29) in the
UK and Le Programme de Médicalisation desSystèmes d’Infor-
mation (PMSI) database(30) in France (Table 2).All HESand
PMSI diagnostic and procedurecodes relating to disease groups
were carefully reviewed and thosedeemed to put apatient at risk
of VTEaccording to currentAmerican CollegeofChest Phys-
icians (ACCP) consensus guideline definitions (31) were in-
cluded. At-risk patients from both public and private sectors
were included. Thehospital at-risk population wasthen divided
into categoriesaccording to eithertype of surgery(moderate,
high,and highest risk of VTE),ormedical diagnosis on admis-
sion (myocardialinfarction, stroke, and other medicaldiagnoses
associatedwith VTErisk, e.g.heartfailure, respiratoryfailure,
infectious and inflammatorydiseases)(36). Cancer patients
were notseparatelyincludedinorder to avoid double counting
with cancerleading to surgery.This is because coding does not
specify whether an admission for cancer leads to operation or
whether an operation, suchascolectomy, is for abenign or ma-
lignant condition. We also did not include hematological cancers
suchasleukemia,and other cancers not known to be associated
with VTE (e.g. skin cancers).Hence we took aconservative ap-
proach.Inaddition, patients whowere in hospitalfor daypro-
cedures and minor surgery were excluded. Diagnoses requiring
therapywith anticoagulants were alsoexcluded.
The at-risk populations for the UK and France were based on
the HES(29) and PMSI (30) data,respectively. These databases
are robust and readilyaccessible sources of patient-leveldata.
Equivalent databaseswere eithernot available, or the method of
data collection wasinconsistent with HES and PMSI for the
other countries. Forthese countries (Germany, Italy, Spain,and
Sweden), aweightedaverage of the UK and French datawas
used to derive patient numbers. Foreach at-risk group, the age-
adjustednumber of VTE eventswas calculated on the basis of the
weighteddataand this wasapplied to age-specific breakdowns
of the relevant countrypopulations.
Hospital-acquired eventswere estimatedbyapplying spe-
cific risk frequenciesfor developingVTE to the surgical and
medical at-risk populations, according to whether or notthey
were receiving VTE prophylaxis. Theprobability of developing
VTE withand without prophylaxis (Table 3) (31–33, 37) was
combined with estimatesofcurrent prophylaxis use (Table 4) in
ordertocalculatethe number of incident clinicalVTE eventsin
each at-risk group. Prophylaxis use wasestimatedbased on pub-
lisheddatawherever possible and,ifotherwise, on expertcon-
sensus.
Estimation of the numberofrecurrent VTE events
In order to fully capturethe number of VTE eventsexpected in a
givenyear,weestimatedthe number of recurrent clinicallyevi-
dent eventsalongside incident cases of clinical VTE. Themajor-
ity of cases of recurrence occurwithin the first year of the index
France Germany ItalySpain Sweden UK
Generalpopulation 60,424,000 82,425,000 58,058,000 40,281,000 8,986,000 60,271,000
At-risk population
Surgical
Moderate risk
High risk
Highest risk
577,362
1,155,368
365,212
751,436
1,487,275
474,948
532,200
1,035,322
339,129
385,328
660,542
216,299
84,178
153,689
50,576
599,355
742,324
250,162
Medical
Myocardial infarction*
Stroke
Other medical
†
15,250
67,499
1,130,796
25,883
106,691
1,416,345
18,685
79,025
1,019,501
11,930
50,464
651,004
2,795
11,876
152,406
18,651
74,678
720,880
†
Cardiac, respiratory, and inflammatorydiseases and severe infections. *Themajorityofpatientswith myocardial infarction receiveanticoagulant therapyand
areexcluded.
Table2:At-risk hospital
population estimates
(baseline)extractedfrom
or based on the Hospital
EpisodesStatistics (29)
and Le Programme de
Médicalisation desSystèm-
es d’Information (30) data-
bases2004.
ProbabilityofVTE event
Without prophylaxisWithprophylaxis
At-risk population
DVT* PE
†
DVT* PE
†
Surgical
Moderate risk
High risk
Highest risk
0.150
0.300
0.600
0.052
0.103
0.241
0.041
0.081
0.162
0.027
0.054
0.126
Medical
Myocardialinfarction
Stroke
Other medical
‡
0.240
0.547
0.160
0.025 (0.27)
§
0.057 (0.067)
§
0.017 (0.047)
§
0.075
0.235
0.040
0.008
0.024
0.004
*Objectivelyverifiedbyvenography(i.e.all deep-veinthrombosis [DVT] events –both symptomatic
and asymptomatic).
†
In order to calculate mortality,the frequency of pulmonary embolism (PE) for
surgical patients was inflated from the probability of diagnosedclinical PE to the probability of all PE
(i.e.toinclude undiagnosedclinical PE) (37). Formedical patients, the PE frequency was calculated
by applying the Oster algorithm (33) to DVTfrequencies.
‡
Cardiac, respiratory, and inflammatory
diseases andsevereinfections.
§
Figures based on clinical trials and inflatedfromthe probability of
diagnosedclinical PE to the probability of all PE.
Table3:Probability of venous thromboembolism (VTE)events
extracted or derived from Geerts et al.2001 (32) and 2004
(31).
Cohen et al.Venous thromboembolisminEurope
759
event;however,the cumulative frequencycontinues to increase
as timefrom the indexevent elapses (34).Total recurrent events
expected in agiven year therefore comprised recurrences of inci-
dent cases occurring in thatyear plus recurrences fromincident
eventsoccurring in previous years.Inorder to incorporatethis
into the model, we assumedaconstant incidenceofindexVTE
over the previous five years and applied1-to5-year recurrence
rates to these data (Table 5) (10,18, 34).
Estimation of the numbers of PTS and PH events
Asimilarapproach wastaken to estimate the expected number of
PTS cases. Totalcases expected in agiven year therefore com-
prised PTS related to incident cases of VTE occurring in that
year plus PTSrelated to incident cases of VTEoccurring in the
previous five years (10).However,given thatPTS is achronic
disease, thereisalso an underlying levelofdiseaseexisting in
anygiven year thatisequivalent to the sum of cases developedin
previous years.Inthe model, PTS wasfurther separated into
mild/moderate disease and severe disease. This distinction was
made becausediseaseseverity drives resource use. We assumed
thatonce apatient haseithersevere or mild/moderatePTS,they
willremaininthis state (23).Inreality,itislikelythat patients
will pass from one state to anotheroverthe course of the disease.
The expected number of PH cases wasestimatedbased on
publishedincidence rates (18).SimilartoPTS,expected cases of
PH in anygiven year comprisedPHrelated to incident cases of
VTEoccurring in thatyear plus PH related to incident cases of
PE occurring in the previous twoyears.Rates of chronic PH ex-
pected to result in surgicalintervention were appliedtothe esti-
mate of incident cases of PE in ordertoestimate the total number
of PH cases in agiven year.
Estimation of the numberofVTE-relateddeaths
Estimates of the number of VTE-related deaths included esti-
matesofsudden death, death following diagnosedand treated
disease, and death following undiagnosed, untreateddisease.
Thenumber of deaths from recurrent cases wasalso estimated.
Thenumber of VTE-related deaths wasestimatedfrom
numbers of treated VTEevents on the basis of an algorithm de-
velopedbyOster et al.(33). Asimplifiedversion of the algo-
rithm is shown in Figure 2(31, 33, 35). The keyparameters de-
fined by the algorithm are rates of diagnosed DVT, likelihoodof
progression from undiagnosedDVT (symptomaticand asympto-
matic) to PE, and rates of death from diagnosed and undiagnosed
VTE (symptomatic). Forthe purpose of the analysis, we as-
sumedthat diagnosed VTE wassynonymous with treatedVTE.
We thenapplied the algorithm to the numbersoftreated DVTand
PE eventsinorder to estimate the number of deaths related to
diagnosedVTE in each of the sixEUcountries. This allowedus
to estimate the number of expected deaths from cases of VTE
that arenot diagnosedaswellasexpected deaths from cases that
arediagnosed. In order to assess whether the algorithm wasap-
plicable and reflectedcurrent understanding of VTE, it wasre-
viewed by the advisoryboard. On the basis of this review, one
parameter wasadjusted. The likelihood of all DVT(initially
asymptomatic and symptomatic) being detected and treated was
adjustedfrom 16%, based on areport by Oster et al. (33), to 10%
on thebasis of morerecent data(31, 35), whichsuggestedthat
the original estimate overstatesthe current likelihood of clinical
presentation and diagnosis.
Probabilistic sensitivityanalyses
Probabilistic sensitivity analyses (PSAs) were conducted to as-
sess the impactofuncertaintyonthe model results. One-way
At-riskpopulation France Germany Spain ItalySwedenUK
Surgical
Moderate risk
High risk
Highest risk
0.80
0.90
0.95
0.80
0.85
0.95
0.60
0.80
0.78
0.50
0.70
0.90
0.80
0.80
0.95
0.40
0.90
0.70
Medical
Myocardial infarction
Stroke
Other medical*
0.85
0.85
0.60
0.55
0.85
0.62
0.60
0.60
0.62
0.30
0.20
0.25
0.20
0.20
0.20
0.48
0.48
0.30
*Cardiac, respiratory, and inflammatorydiseases and severe infections.
Table4:Probability (base-
line estimate)ofvenous
thromboembolism pro-
phylaxis use in theat-risk
population estimated from
publisheddata and clini-
cian estimates, 2004.
Recurrent event/outcome Year 1Year 2Year 3Year 4Year 5Cumulative
Deep-veinthrombosis 0.104 0.030 0.017 0.017 0.017 0.185
Pulmonaryembolism 0.025 0.007 0.004 0.004 0.004 0.044
Mild/moderate PTS 0.147 0.038 0.001 0.001 0.001 0.188
Severe PTS0.026 0.0170.017 0.0170.017 0.094
Pulmonaryhypertension0.014 0.003- --0.017
PTS, post-thrombotic syndrome.
Table5:Probability of re-
currence of venous throm-
boembolism (10, 34) and
associated outcomes (10,
18) in theat-risk popu-
lation.
Cohen et al.Venous thromboembolisminEurope
760
Figure2:Simplifiedver-
sion of the Osteretal. (33)
algorithm showing poten-
tial outcomes of hospital-
acquired venous throm-
boembolism.*The likelihood
of adiagnosisofdeep-vein
thrombosis (DVT) wasad-
justed from 16% as reported
by Oster et al. (33) to 10% on
thebasisofmorerecentdata
(31, 35) estimatingthe current
likelihood of diagnosis. PE, pul-
monaryembolism;P=Probabil-
ity.
sensitivity analyses were conducted according to the methodolo-
gy of Sullivanetal. (38)with allinputs varied by ±50%. All pa-
rametersthat had an impact >1%onthe overalloutput of the
model were includedasinputs in PSAs. The distribution of the
at-risk population and rates of VTE prophylaxis use were in-
cluded apriori in the PSAs. Distribution and ranges of modelin-
puts were definedaccording to availableinformation and recom-
mendedbestpractice as follows:clinicianestimateswere used
forrates of VTEprophylaxis use;ACCP-specified ranges (31,
32)were used forthe rates of VTE with and without prophylax-
is; published data were used for recurrence rates (34)and PTS
rates (10);and methodology as outlined by Sullivanetal. (38)
wasusedfor the at-risk population, rates of death, and the prob-
ability of symptomatic/diagnosedDVT and PE.
Extrapolationofdata
To estimate the number of VTEevents occurring per annum
across the whole of the EU,the country-specific outputs from the
six EU countries modeled were summedand then inflatedbythe
ratio of population size in the six included countries (310.4 mil-
lion) to the population sizeinall 25 member-state countries
(454.4 million).
Results
To tal numberofnon-fatal VTE events and associated
complications
In total, the modelestimatedthat 761,697 non-fatalVTE events
(465,715cases of DVTand 295,982 cases of PE) and 399,808 as-
sociated complications (395,673 patients with PTS and 4,135
with PH) occurredacross the six EU countries in 2004(Table6).
Event,n(95% CI) Community-acquired Hospital-acquired To tal
Non-fatalVTE event
Deep-veinthrombosis
Pulmonaryembolism
200,482
(172,548–226,239)
86,511
(73,967–99,626)
265,233
(209,844–332,407)
209,471
(153,817–273,371)
465,715
(404,664–538,189)
295,982
(242,450–360,363)
VTE-related deaths
†
TreatedVTE
†
Untreated VTE
Suddendeath
108,535
(77,243–178,968)
8,124
(6,151–10,470)
63,541
(41,574–114,074)
36,870
(25,467–60,724)
261,477
(211,782–325,823)
18,349
(12,422–25,695)
153,853
(110,943–211,670)
89,275
(64,718–117,822)
370,012
(300,193–483,108)
26,473
(19,158–35,271)
217,394
(154,910–317,068)
126,145
(92,352–170,949)
Associated outcome
Post-thrombotic syndrome
‡
Pulmonaryhypertension
177,236
(147,893–207,420)
1,173
(991–1,371)
218,437
(162,065–285,321)
2,961
(2,162–3,860)
395,673
(328,154–477,185)
4,135
(3,311–5,089)
*Modeledcountries were France,Germany, Italy, Spain,Sweden, and theUK.
†
Including deep-veinthrombosis- and pulmonary embolism-related deaths.
‡
Including all post-thrombotic syndrome (newcases plus those underlying from previous years). CI,confidence interval.
Table6:Total non-fatal ve-
nous thromboembolism
(VTE)events, VTE-related
deaths, and associated
outcomes across allsix
European Union countries
modeled* in 2004.
Cohen et al.Venous thromboembolisminEurope
761
Hospital-acquired eventsrepresent 63%ofall cases of VTEplus
associatedcomplications.
Basedonatotal population estimate of 310.4 million inhab-
itants and the number of non-fatalVTE eventsacross the six EU
countries includedinour model, the estimated DVTattack rate
(incidence of first lifetime and recurrent DVT) was148 per
100,000 person-years (65 per 100,000 for community-acquired
DVT, and 83 per 100,000for hospital-acquiredDVT). Similarly,
the estimatedPEattack ratewas 95 per 100,000person-years
(28per 100,000for community-acquiredPE, and 67 per 100,000
for hospital-acquired PE).
To tal numberofVTE-relateddeaths
The total estimated number of VTE-related deaths for 2004
across the sixEUcountries was370,012 (Table 6). Of these,7%
(26,473)were expected to have resultedfrom diagnosed (and
presumably treated) VTE, 34%(126,145) from sudden fatalPE,
and 59% (217,394)from PE-related deaths followingundiag-
nosed (untreated) VTE. Deaths that occurredasaconsequence
of hospital-acquiredVTE comprised 71% of the total number of
VTE-related deaths in the sixEUcountries.
Sensitivity analyses
The results of the sensitivityanalyses arereported as 95%con-
fidenceintervals in Table6.
Discussion
This study describes the extent of the health burden attributedto
VTE in theEUinterms of the totalnumber of incident and recur-
rent non-fatal DVTand PE clinicalevents, and VTE-related
deaths per year in six EU countries.Our estimates indicate that
VTE is amajor public-health probleminthese six countries, with
apredicted totalnumber of just under half amillion DVTevents
and almost athird of amillion PE eventsper annum.Fur-
thermore,athird of amillion deaths occurper year duetosudden
PE or following undiagnosedand untreatedVTE. Giventhe
availability of effectiveVTE prophylaxis(31), manyofthese
eventsand deaths could be prevented –particularlythose that are
hospital-acquired(39).
The estimated total number of VTE events(DVTand PE: 148
per 100,000 and 95 per 100,000, respectively)inthe EU is higher
than that reported for US communities (24–26),but is close to
that reported in the EPI-GETBO study (27) in WesternFrance
(124per 100,000and 60 per 100,000, respectively). Onlythe
community data from the EPI-GETBO study (27) were used in
our analysis; the EPI-GETBO hospitaldatawas not. Many of the
other differencesbetween estimatescan be accounted forbythe
fact thatthe Wo rcester,MA(24) and Olmsted County, MN (25)
studiesexcludedrecurrentevents, whereas the EPI-GETBO
study (27)and our study included both incident and recurrent
events; recurrentevents accountedfor 25% of eventsinthe EPI-
GETBO study (27) and 19% of eventsinthis study.Our model
conservativelyassessedthe contribution of asymptomatic hospi-
tal eventstosymptomatic events(10% insteadof16% conver-
sion rate).Inaddition, acomparison of the number of VTE
eventspredicted by our model with data from the above epidemi-
ological studies(24–27) is complicated by the fact thatthe latter
onlycollectedsymptomatic, diagnosedVTE events, while our
modelalso accounted forundiagnosed eventsleading to VTE
mortality.The number of deaths in the 25 member states of the
EU has beenreported to be 4,432,177 perannum (40).Based on
apopulation of 454.4 million, this givesanannual death rate of
approximately 1%. The total number of deaths occurring in
France, Germany,Italy,Spain, Sweden, and the UK has beenre-
ported to be 2,982,816 perannum (40).Our model estimates that
370,012 (300,193–483,108) (12%)deaths per annum in these six
EU countries are duetoVTE, withthe proportion varying from
10%inthe UK to 14% in Italy.While this figure mayatfirst
seem high,itisconsistent with data from community- and hospi-
tal-based postmortem examination studies, whichhaveshown
that approximately 10% of deaths in hospitals aredue to PE
(41–43).Furthermore,ofthe estimated 370,012 (300, 193–483,
108) VTE-related deaths predicted per annum, 93% were aresult
of either sudden fatalPEorundiagnosed, untreatedVTE. This
means that only7%ofdeaths followeddiagnosedand treated
VTE, and would have beenrecognized as VTE-related by the
treating physician. Treating physicians maynot be aware that
VTE is responsiblefor more than 14 times as manydeaths as
thoserecognized as being VTE-related,reflecting the poor
awarenessthat manyphysicians have of the potentiallyfatal out-
comes of VTEand the importanceofproviding at-risk patients
with VTEprophylaxis. In addition, at least25% of VTE-related
deaths were estimatedtobe“sudden” and,therefore onlypre-
ventable with appropriate prophylaxis as currentlyrecom-
mended(4, 31).
The cost of treating VTE and related morbidityissubstantial.
Our data have important implications regarding VTE prophylax-
is in both the surgical and medical settings. While some formof
prophylaxis is giventomost major-surgerypatients, studieshave
shown thatthe type, duration, and intensityofprophylaxis are
frequentlyinappropriate, usuallyleading to inadequate prophy-
laxis (44). Furthermore, the use of recommendedprophylaxis in
medical patients, asizeable proportion of whom will develop
VTEifleft unprotected,has been reportedtobesuboptimal (45,
46).
Estimatesofthe number of VTE eventsacross the EU,based
on extrapolation from the six modeledcountries, were 684,019
DVTevents, 434,723 PE events, and 610,138PTS eventsper
annum.The number of VTE-related deaths wasestimatedat
543,454 across the EU per annum, whichismore than double the
number of combined deaths due to AIDS (5,860),breast cancer
(86,831), prostate cancer (63,636), and transport accidents
(53,599)(40). This figure of 543,454 deaths is similar to that cal-
culated from the most complete postmortem examination data
setknown, whichestimatesaround 514,000deaths fromPEin
the EU (42). These datagiveastrong indication of the burden of
VTE in the EU,but further researchiswarrantedtoestimate this
numbermore accurately.However,these figuresleave no doubt
that thereisanurgent need forhealthcare policy-makers across
the EU to review their nationalguidelines and hospital protocols
with respecttothe accepted international guidelines for VTE
prophylaxis and to ensurethat theyare consistently implemented
(21).
Cohen et al.Venous thromboembolisminEurope
762
Limitations
Our approach for estimating thenumber of VTE eventsand re-
latedmorbidity and mortality has some limitations.These types
of modeling exercisesare widelyusedinmanyclinical areas to
estimate the diseaseburdenwhereotherepidemiologicalmeth-
odshavefailedtocapturethe complete picture or arenot feas-
ible. External validity hasbeen testedfor the mortality data as
previouslypresented. Some country-specific data were unavail-
able.Wherethis wasthe case, data were extrapolatedfrom the
UK and France. Forsurgicalpatients, the frequencies of PE were
availablefrom observationalstudies(31) and these were used;
however, for non-surgical patients, onlyclinical trialdata(see
Table3)were available, hence we used the Oster algorithm (33)
to calculate the frequencyofPE. Usingthis modified algorithm
rather than clinicaltrial dataresultedinmore conservativeout-
comes(16% conversion rate would have resultedinmore
events). Also, as there are no current dataavailableregarding
howmanyVTE eventsare symptomatic butundiagnosed, we
used the Oster algorithm (33) to calculate this as it is still the best
availablemethod. Although the Osteralgorithm wasoriginally
used for the assessment of surgicalpatients, its inputs comefrom
studiesexamining all VTE, henceitwas suitable forour model.
Rates of VTE prophylaxis use arenot currentlyavailableinthe
literaturefor all settings. We therefore used the most appropriate
prophylaxis rates based on publishedfigures and the opinions of
experts in each country. Our calculations followedcurrent con-
vention and classified eventsoccurring up to 90 days post-hospi-
taldischargeasbeing hospital acquired (34);this led to most
eventsbeing related to hospitalization. Theseevents were ex-
cluded fromcommunity-acquired total event rates in orderto
avoid double-counting. The classification method did not there-
foreaffect overallnumbers. Finally, even with the uncertainty re-
vealed in the PSA, the estimatesofthe number of eventsare high.
Additional research is needed to refinethe inputs to the model
reported here, and to assess the impact of increasing prophylax-
is use in medicaland surgicalat-risk populations.
Conclusions
VTE is apreventable disease and effective prophylaxis is widely
available(4, 31), butimplementation remains suboptimal
(44–46). Giventhe current mortality rate associated with VTE
and its enormous burden in the EU in termsoflong-term morbid-
ity and costofpatient management, there is aclear opportunity
for physicians and policy-makers to have alarge, positive impact
on public health by making implementation of appropriate pro-
phylaxis use ahigherpriority (47). Strategiesfor improving the
uptakeofthe evidence-based guidelinesfor the use of thrombo-
prophylaxis should be reviewedand implemented.
Acknowledgements
TheVITAE study wasfunded by sanofi-aventis. Themodel structure and
underlying data were approvedbyanindependent expertboard. Editorial
assistance wasprovided by sanofi-aventis.
References
1. SteinPD, HenryJW. Prevalence of acute pulmon-
aryembolismamong patients in ageneralhospital and
at autopsy. Chest 1995;108: 978–981.
2. Heit JA,Silverstein MD,Mohr DN et al. Predictors
of survivalafter deepvein thrombosisand pulmonary
embolism: apopulation-based,cohort study.Arch In-
tern Med 1999;159: 445–453.
3. Pineda LA,Hathwar VS,Grant BJ.Clinical suspi-
cion of fatalpulmonary embolism. Chest2001;120:
791–795.
4. NicolaidesAN,BreddinHK, Fareed J, et al. Preven-
tionofvenousthromboembolism. International Con-
sensus Statement. Guidelines compiled in accordance
withthe scientif ic evidence. IntAngiol 2001; 20: 1–37.
5. Nordstrom M, Lindblad B, Bergqvist D, et al. A
prospective study of theincidenceofdeep-vein throm-
bosis within adefined urban population. JInternMed
1992; 232:155–160.
6. Hull RD, Pineo GF,Stein PD,etalExtended out-of-
hospital low-molecular-weight heparin prophylaxis
against deepvenous thrombosisinpatients afterelec-
tive hip arthroplasty: asystematic review.Ann Intern
Med 2001;135: 858–869.
7. Mismetti P, LaporteS,DarmonJY, et al. Meta-
analysisoflow molecular weight heparin in thepreven-
tionofvenousthromboembolismingeneralsurgery. Br
JSurg2001;88: 913–930.
8. DouketisJD, EikelboomJW, QuinlanDJ, et al.
Short-durationprophylaxisagainst venousthrom-
boembolismaftertotal hip or kneereplacement: a
meta-analysisofprospectivestudies investigating
symptomatic outcomes. Arch Intern Med 2002;162:
1465–1471.
9. Vaitkus PT,LeizoroviczA, CohenAT, et al. Mortal-
ity rates and risk factors forasymptomatic deepvein
thrombosisinmedical patients. Thromb Haemost
2005; 93: 76–79.
10. Prandoni P, Lensing AW,CogoA,etal. The long-
term clinical courseofacute deepvenous thrombosis.
AnnInternMed 1996; 125:1–7.
11. McColl MD, EllisonJ,Greer IA,etal. Prevalence
of thepost-thrombotic syndromeinyoungwomen with
previous venousthromboembolism. BrJHaematol
2000; 108:272–274.
12. Mohr DN,Silverstein MD,Heit JA,etal. The ve-
nousstasissyndrome afterdeep venousthrombosis or
pulmonary embolism: apopulation-based study.Mayo
Clin Proc 2000;75: 1249–1256.
13. AsbeutahAM, RihaAZ, CameronJD, et al. Five-
year outcome study of deepvein thrombosisinthe
lowerlimbs. JVascSurg2004;40: 1184–1189.
14. Bova C, RossiV, Ricchio R, etal. Incidenceofpost-
thrombotic syndromeinpatients withprevious pul-
monaryembolism. Aretrospectivecohort study.
ThrombHaemost2004;92: 993–996.
15. GabrielF,Labios M,PortolesO,etal. Incidenceof
post-thrombotic syndromeand its associationwith
various risk factors in acohort of Spanish patients after
one year of follow-up followingacute deepvenous
thrombosis. ThrombHaemost2004;92: 328–336.
16. Roumen-Klappe EM,den HeijerM,Janssen MC,
et al. The post-thrombotic syndrome: incidence and
prognostic valueofnon-invasive venousexaminations
in asix-year follow-up study.Thromb Haemost2005;
94: 825–830.
17. Wille-Jorgensen P, Jorgensen LN,CrawfordM.
Asymptomatic postoperative deepvein thrombosisand
thedevelopment of postthrombotic syndrome. Asys-
tematic review and meta-analysis. ThrombHaemost
2005; 93: 236–241.
18. Pengo V, Lensing AW,Prins MH, et al.Incidenceof
chronic thromboembolic pulmonary hypertensionafter
pulmonaryembolism. NEnglJMed 2004;350:
2257–2264.
19. Ribeiro A, Lindmarker P, Johnsson H, et al. Pul-
monaryembolism: one-year follow-up withechoc-
ardiographydopplerand five-year survivalanalysis.
Circulation 1999;99: 1325–1330.
20. GriffinJ.Deepvein thrombosisandpulmonary em-
bolism. London: Office of HealthEconomics 1996;
48–59.
21. HouseofCommonsHealthCommittee. The pre-
vention of venousthromboembolisminhospitalised
patients. Second reportofSession2004–05.Available
at: http://www.parliament.the-stationery-office.co.uk/
pa/cm200405/cmselect/cmhealth/99/99.pdf.Accessed
February27, 2007.
22. BergqvistD,JendtegS,Johansen L, et al. Costof
long-term complications of deepvenous thrombosisof
thelower extremities: an analysisofadefined patient
populationinSweden.Ann Intern Med 1997;126:
454–457.
23. CapriniJA, Botteman MF, Stephens JM,etal.
Economicburden of long-term complications of deep
vein thrombosisafter totalhip replacement surgeryin
theUnited States.Value Health2003;6:59–74.
24. Anderson JrFA, Wheeler HB,GoldbergRJ, et al. A
population-based perspective of thehospital incidence
and case-fatality rates of deepvein thrombosisand pul-
monaryembolism. The Worcester DVTStudy.Arch In-
tern Med 1991;151: 933–938.
25. SilversteinMD, Heit JA,Mohr DN,etal. Trendsin
theincidenceofdeepvein thrombosisand pulmonary
embolism: a25-year population-based study.Arch In-
tern Med 1998;158: 585–593.
Cohen et al.VenousthromboembolisminEurope
763
26. Spencer FA,Emery C, Lessard D, et al.The Wor-
cester Venous Thromboembolismstudy: apopulation-
based study of theclinical epidemiology of venous
thromboembolism. JGen InternMed 2006; 21:
722–727.
27. Oger E. Incidence of venous thromboembolism: a
community-based studyinWesternFrance.EPI-
GETBO Study Group. Groupe d'Etude de la Throm-
bose de Bretagne Occidentale. ThrombHaemost 2000;
83: 657–660.
28. US Census Bureau. International Data Base (IDB).
Availableat: http://www.census.gov/ipc/www/idbnew.
html. Accessed February27, 2007.
29. Department of Health.Hospital Episode Statistics
(HES) data 2004–05. Av ailableat: http://www.dh.gov.
uk/PublicationsAndStatistics/Statistics/HospitalEpi-
sodeStatistics/fs/en.Accessed February27, 2007.
30. AgenceTechnique de l’Informationsur l’Hospitali-
sation. Programme de médicalisationdes systèmes
d'information(PMSI)2004. Availableat: http://www.
atih.sante.fr/. Accessed February27, 2007.
31. Geerts WH,Pineo GF,HeitJA, et al.Preventionof
venous thromboembolism: theSeventhACCP Confer-
ence on Antithrombotic and Thrombolytic Therapy.
Chest 2004;126 (3 Suppl): 338S–400S.
32. Geerts WH,HeitJA, ClagettGP, et al.Preventionof
venous thromboembolism. Chest 2001; 119(1 Suppl):
132S–175S.
33. OsterG,Tuden RL,Colditz GA. Acost-effective-
ness analysis of prophylaxisagainst deep-veinthrom-
bosis in majororthopedicsurgery. JAmMed Assoc
1987;257: 203–208.
34. Heit JA,Mohr DN,Silverstein MD,etal. Predictors
of recurrence after deep vein thrombosisand pulmon-
aryembolism: apopulation-based cohortstudy.Arch
InternMed 2000; 160: 761–768.
35. Piovella F, Anderson Jr FA ,etal; for theIMPROVE
Investigators.Rates of clinically apparent venous
thromboembolismand deathresultingfrom pu lmonary
embolismamong unselected hospitalized acutelyill
medical patientsversus rates predicted from clinical
studies. Blood (ASH Annual Meeting Abstracts)2005;
106: Abstract1620.
36. Cohen AT,Alikhan R. Prophylaxisofvenous
thromboembolisminmedical patients. Cur rOpinPulm
Med 2001;7:332–337.
37. DalenJE, Alpert JS. Natural historyofpulmonary
embolism. ProgCardiovasc Dis 1975; 17: 259–270.
38. SullivanSD, Davidson BL,Kahn SR, et al.A cost-
effectiveness analysis of fondaparinux sodium com-
pared with enoxaparin sodium as prophylaxisagainst
venous thromboembolism: use in patientsundergoing
majororthopaedicsurgery. Pharmacoeconomics2004;
22: 605–620.
39. Kucher N, KooS,Quiroz R, et al.Electronic alerts
to prevent venous thromboembolismamong hospital-
ized patients. NEngl JMed 2005; 352: 969–977.
40. Eurostat.Eurostatstatisticsonhealth and safety
2001. Availableat: http://epp.eurostat.cec.eu.int/port
al/page?_pageid=1996,45323734&_dad=port-
al&_schema=PORTAL&screen=welcome-
ref&close=/health/public&language=en&prod-
uct=EU_MASTER_health&root=EU_MASTER_he
alth&scrollto=0. Accessed February27, 2007.
41. Cohen AT,Edmondson RA,PhillipsMJ, et al.The
changing pattern of venous thromboembolic disease.
Haemostasis1996; 26: 65–71.
42. LindbladB,Sternby NH, Bergqvist D. Incidence of
venous thromboembolismverifiedbynecropsyover30
years.BMJ 1991;302: 709–711.
43. Sandler DA,Martin JF.Autopsy provenpulmonary
embolisminhospital patients: are we detectingenough
deep vein thrombosis? JRSoc Med 1989; 82: 203–205.
44. ArnoldDM, Kahn SR, Shrier I. Missed opportun-
itiesfor preventionofvenous thromboembolism: an
evaluationofthe use of thromboprophylaxis guide-
lines. Chest 2001; 120: 1964–1971.
45. Ageno W, SquizzatoA,Ambrosini F, et al.Throm-
bosis prophylaxisinmedical patients: aretrospective
review of clinical practice patterns. Haematologica
2002; 87: 746–750.
46. RashidST, Thursz MR,Razvi NA,etal. Ve nous
thromboprophylaxis in UK medical inpatients. JRSoc
Med 2005; 98: 507–512.
47. ShojaniaKG, Duncan BW,McDonaldKM, et al.
Makinghealth care safer: acritical analysis of patient
safetypractices. Evidence report/technologyassess-
ment:number 43. Av ailableat: http://www.ahrq.gov/
clinic/ptsafety/.Accessed February27, 2007.
Cohen et al.VenousthromboembolisminEurope
764
