ORIGINAL INVESTIGATION

Predictors of Recurrence After Deep Vein

Thrombosis and Pulmonary Embolism
A Population-Based Cohort Study
John A. Heit, MD; David N. Mohr, MD; Marc D. Silverstein, MD; Tanya M. Petterson, MS;
W. Michael O’Fallon, PhD; L. Joseph Melton III, MD

Background: The appropriate duration of oral antico-
agulation after a first episode of venous thromboembo-
lism (VTE) is uncertain and depends upon VTE recur-
rence rates.
Objective: To estimate VTE recurrence rates and de-
termine predictors of recurrence.
Methods: Patients in Olmsted County, Minnesota, with
a first lifetime deep vein thrombosis or pulmonary em-
bolism diagnosed during the 25-year period from 1966
through 1990 (N = 1719) were followed forward in time
through their complete medical records in the commu-
nity for first VTE recurrence.
Results: Four hundred four patients developed recur-
rent VTE during 10 198 person-years of follow-up. The
overall (probable/definite) cumulative percentages of VTE
recurrence at 7, 30, and 180 days and 1 and 10 years were
1.6% (0.2%), 5.2% (1.4%), 10.1% (4.1%), 12.9% (5.6%),
and 30.4% (17.6%), respectively. The risk of recurrence
was greatest in the first 6 to 12 months after the initial event
but never fell to zero. Independent predictors of first over-
all VTE recurrence included increasing age and body mass
index, neurologic disease with paresis, malignant neo-
plasm, and neurosurgery during the period from 1966
through 1980. Independent predictors of first probable/
definite recurrence included diagnostic certainty of the in-
cident event and neurologic disease in patients with hos-
pital-acquired VTE. Recurrence risk was increased by
malignant neoplasm but varied with concomitant chemo-
therapy, patient age and sex, and study year.
Conclusions: Venous thromboembolism recurs fre-
quently, especially within the first 6 to 12 months, and
continues to recur for at least 10 years after the initial
VTE. Patients with VTE with neurologic disease and pa-
resis or with malignant neoplasm are at increased risk
for recurrence, while VTE patients with transient or re-
versible risk factors are at less risk.
Arch Intern Med. 2000;160:761-768

From the Division of
Cardiovascular Diseases and
the Section of Hematology
Research (Dr Heit) and the
Division of Area General
Internal Medicine (Drs Mohr
and Silverstein), Department of
Internal Medicine, and the
Sections of Clinical
Epidemiology (Drs Silverstein
and Melton) and Biostatistics
(Ms Petterson and
Dr O’Fallon), Department of
Health Sciences Research, Mayo
Clinic and Mayo Foundation,
Rochester, Minn. Dr Silverstein
is now with the Center for
Health Care Research, Medical
University of South Carolina,
Charleston.
R
ECURRENT VENOUS throm-
boembolism (VTE) is an
important risk factor for
death after pulmonary em-
bolism (PE)1,2 and for ve-
nous stasis syndrome after deep vein
thrombosis (DVT),2 and is associated with
significantly increased long-term health
care costs.3 While anticoagulation therapy
is effective in preventing recurrence,4,5 the
optimal duration of anticoagulation after
an initial episode of DVT or PE is uncer-
tain6 and depends largely on the rate of re-
currence. However, the reported rates of
recurrent VTE vary widely, ranging from
0.6% to 5% at 90 days and from 13% to
25% at 5 years.2,7-14 Several factors may ac-
count for the variability in reported re-
currence rates. For example, studies that
identified cases from acute-care hospital
discharge data8,9 or Medicare claims diag-
noses11 are limited by diagnostic uncer-
tainty or misclassification, as well as fail-
ure to include autopsy-discovered
recurrence or to accurately separate ini-
tial from recurrent events.12 Moreover,
studies that identified cases solely from
acute-care hospital admissions may have
missed recurrent VTE among patients re-
siding in nursing homes or other long-
term care facilities9,11,12 as well as rapidly
fatal recurrences occurring out of hos-
pital.8 Follow-up studies of selected VTE
See also page 809
populations, such as elderly patients,11
patients referred to tertiary care centers
for diagnostic evaluation and treat-
ment,2,15 or patients enrolled in clinical tri-
als1,7,10,13,14,16 may not estimate the true re-
currence rate accurately since these studies
did not include the full spectrum of VTE
disease. Moreover, information regard-
ing long-term recurrence is limited since
follow-up in most studies was restricted

(REPRINTED) ARCH INTERN MED/ VOL 160, MAR 27, 2000 WWW.ARCHINTERNMED.COM
761
Downloaded from www.archinternmed.com on April 12, 2010
©2000 American Medical Association. All rights reserved.
 METHODS
STUDY SETTING AND DESIGN
Using the data resources of the Rochester Epidemiology
Project,18 we identified the inception cohort of Olmsted
County residents with a first lifetime DVT or PE during the
25-year period from 1966 through 1990.17 All subjects were
followed forward in time through their linked medical rec-
ords19 (retrospective cohort study) until either death or emi-
gration from the community. For each subject, all inpa-
tient and outpatient medical records of any local health care
provider were searched for any evidence of recurrence as
of the last clinical contact. The study was approved by the
Mayo Clinic Institutional Review Board.
DEFINITION OF DVT AND PE
The initial episode of DVT or PE was categorized into the
highest possible of 3 levels of diagnostic certainty (pos-
sible, probable, or definite) according to previously de-
fined criteria.17 A DVT was categorized as definite when con-
firmed by venogram, computed tomographic scan, magnetic
resonance image, or pathologic examination of thrombus
removed at surgery or autopsy; as probable if either testing
for the definite level of diagnostic certainty was not per-
formed or the results were indeterminate and the results
of at least one of the following noninvasive tests were posi-
tive: impedance plethysmography, continuous-wave Dop-
pler ultrasound examination performed in the Mayo Clinic
Vascular Laboratory, compression duplex ultrasonogra-
phy, radionuclide venography, or radiolabeled fibrinogen
leg scan; and as possible if either confirmatory tests were
not done or the results were indeterminate and (1) the medi-
cal record indicated that a physician had made a diagnosis
of DVT (or possible DVT), (2) signs and symptoms con-
sistent with DVT were present, and (3) the patient re-
ceived a course of anticoagulation therapy with heparin,
warfarin, or a similar agent, or a surgical procedure for DVT.
A PE was categorized as definite when confirmed by pul-
monary angiogram, computed tomographic scan, mag-
netic resonance image, or pathologic examination of throm-
bus removed at surgery or autopsy; as probable if either
testing for the definite level of diagnostic certainty was not
performed or the results were indeterminate and the re-
sults of a perfusion or ventilation-perfusion lung scan were
interpreted as showing a high probability for PE; and as pos-
sible if either confirmatory tests were not done or the re-
sults were indeterminate and (1) the medical record indi-
cated that a physician had made a diagnosis of PE, (2) signs
and symptoms consistent with PE were present, and (3)
the patient received a course of anticoagulation therapy with
heparin, warfarin, or a similar agent, or a surgical proce-
dure for PE, such as an inferior vena cava filter. A short
to 1 to 3 years,1,9,11 with only 2 studies reporting fol-
low-up to 8 years.2,12
The duration of anticoagulation therapy also
depends upon accurate knowledge of predictors of VTE
recurrence. However, all of the 6 studies addressing
potential predictors of recurrence had limitations in
study design: 1 study had a relatively small sample
size,12 3 were clinical trials with extensive exclusion cri-
(REPRINTED) ARCH INTERN MED/ VOL 160, MAR 27, 2000
762
Downloaded from www.archinternmed.com on April 12, 2010
©2000 American Medical Association. All rights reserved.
period of anticoagulation therapy while awaiting comple-
tion of diagnostic evaluation for either suspected DVT or
suspected PE was insufficient grounds for inclusion. An epi-
sode of VTE consisting of both DVT and PE was catego-
rized into the highest level of diagnostic certainty present
for either manifestation.
To be categorized as a recurrent DVT or PE, an event
had to meet all of the criteria for an incident VTE event. Ad-
ditionally, recurrent events were categorized as probable/
definite recurrent DVT or PE if the results of confirmatory tests
or procedures for a probable or definite VTE were positive
and the site of recurrent VTE was previously uninvolved or
had documentation of resolution. They were categorized as
possible recurrent DVT or PE if confirmatory tests were not
done or the results were indeterminate but the other criteria
listed above were met and the site of recurrence was previ-
ously uninvolved or had documentation of resolution. Ex-
tension of a DVT or PE was classified as a recurrence if a sub-
stantial change from the incident event was documented by
history, physical examination findings, or a diagnostic test
result. Classification of autopsy-discovered VTE as a recur-
rence was adjudicated by a committee of the investigators
(J.A.H., D.N.M., and M.D.S.) and based on the likelihood that
the initial event had resolved. Pulmonary embolism diag-
nosed within 24 hours of beginning therapy for a DVT (or
vice versa) was considered a single incident event. If more
than 24 hours of therapy had elapsed, these events were con-
sidered separate and the second event was classified as a re-
currence. Because PE is a complication of DVT, results are
presented for DVT alone or for PE with or without DVT.
POTENTIAL RISK FACTORS
More than 25 baseline characteristics were tested as predic-
tors of recurrent VTE. Data on these characteristics were col-
lected by review of all medical records (inpatient and outpa-
tient) in the community for each subject.18 The characteristics
included the following: type of event (PE, DVT, or both); age
at incident event; sex; year of incident event; patient loca-
tion at onset (community, community but hospitalized within
the previous 90 days, hospital, or nursing home); body mass
index (BMI, calculated as the weight in kilograms divided
by the square of the height in meters: weight [kg]/[height
(m)]2), categorized as underweight (BMI ⬍20), normal
weight (BMI ⱖ20 and ⱕ24), overweight (BMI ⬎24 and ⬍30),
or obese (BMI ⱖ30)20; chronic heart disease (congestive heart
failure or other heart disease [congenital heart disease, car-
diomyopathy, ischemic heart disease, or valvular heart dis-
ease]); active malignant neoplasm (excluding nonmela-
noma skin cancer) with or without chemotherapy (cytotoxic
or immunosuppressive therapy for malignant neoplasm, ex-
cluding tamoxifen); serious neurologic disease (stroke or
other disease affecting the nervous system with associated
extremity paresis, or acute stroke with extremity paresis
teria,7,10,16 and 4 were limited to patients with DVT
only.2,8,12,16
We have identified the inception cohort of Olmsted
County, Minnesota, residents with DVT or PE first diag-
nosed during the 25-year period from 1966 through 1990.17
To address the limitations of previous studies, we performed
a population-based follow-up study to estimate the rate
of recurrent VTE and to determine predictors of recurrence.
WWW.ARCHINTERNMED.COM
 requiring hospitalization within the previous 3 months);
surgery requiring anesthesia (general, orthopedic, neuro-
logic, or gynecologic surgery); anesthesia (general, epidural/
spinal, other); trauma requiring hospital admission (ma-
jor fracture or severe soft-tissue injury); chronic lung disease
(chronic obstructive pulmonary disease, emphysema,
chronic bronchitis, bronchiectasis, interstitial lung dis-
ease, or pulmonary hypertension [asthma was included only
if there was documented evidence of fixed airflow obstruc-
tion]); chronic liver disease (including active hepatitis within
the previous 3 months); chronic renal disease (physi-
cian’s diagnosis and creatinine level ⬎175 µmol/L [2 mg/
dL] for at least 3 months, or nephrotic syndrome); inflam-
matory bowel disease; previous superficial vein thrombosis;
varicose veins (varicose veins or treated varicose veins [in-
jection sclerotherapy or stripping]); central venous cath-
eter or transvenous pacemaker; anticoagulation therapy or
prophylaxis immediately preceding or at the time of the in-
cident event; smoking status (none, former, or current [ciga-
rettes only]); hormone therapy (estrogen or progester-
one); and (for women only) pregnancy or postpartum
(within 3 months of delivery) at the time of the incident
event, oral contraceptive use, gynecologic surgery, and ta-
moxifen therapy. The characteristics related to active ma-
lignant neoplasm, chemotherapy, surgery, anesthesia,
trauma, hormone therapy, oral contraceptive use, and ta-
moxifen therapy were recorded as present only if docu-
mented within the 3 months prior to the VTE event. All
other characteristics were recorded as present if docu-
mented any time prior to the incident VTE event. Body mass
index calculations were based on the most recent height
and weight measurements prior to the incident event. The
BMI could not be calculated for 82 cases, mainly because
of missing height measurements. For these cases, BMI val-
ues were imputed by assigning to each case the average
height and/or weight of those incident cases of the same
sex and age group. For the one child younger than 15 years
with a missing measurement for height, we used the 50th
height percentile from the published 1976 National Cen-
ter for Health Statistics growth chart. Smoking status was
based on tobacco use in the 3 months prior to the incident
event. Smoking status was missing for 92 patients and was
evaluated only after determination of the otherwise final
model (including interactions). In the subset of cases with
complete smoking information, we verified that the final
multivariate model variables did not have hazard ratios dif-
fering from those computed using all cases. We then cat-
egorized those patients with missing smoking status as non-
smokers, reasoning that this would be the most conservative
approach. Because pregnancy and the postpartum period,
oral contraceptive use, gynecologic surgery, and tamoxi-
fen therapy could only be evaluated in women, these vari-
ables were assessed after determining the otherwise final
model, including interactions.
RESULTS
Altogether, 2218 Olmsted County residents had a con-
firmed first lifetime diagnosis of DVT or PE between Janu-
ary 1, 1966, and December 31, 1990. Their mean ± SD
age at onset was 61.7 ± 20.4 years, and 1244 patients
(56%) were female. Of the total, 938 patients (42%) had
DVT, while 969 (44%) had PE and 308 (14%) had evi-
(REPRINTED) ARCH INTERN MED/ VOL 160, MAR 27, 2000
763
Downloaded from www.archinternmed.com on April 12, 2010
©2000 American Medical Association. All rights reserved.
ANALYSIS
Survival free of VTE recurrence was estimated using the
Kaplan-Meier product-limit method. Hazard rates for re-
currence were computed using the life table method for ana-
lyzing survival. Follow-up began on the date of the inci-
dent VTE event. Data were censored at the earliest of death,
last clinical contact, or December 31, 1990. Because our
criteria for a possible VTE recurrence did not require ob-
jective diagnostic testing, some of these events may not have
been true recurrences. On the other hand, some patients
with a true recurrence may not have undergone objective
testing and therefore could not meet our criteria for a defi-
nite or probable event. For example, many noninvasive di-
agnostic tests (eg, venous duplex ultrasonography, lung scan,
computed tomography, magnetic resonance imaging) were
unavailable early in our study. Moreover, many patients were
too ill to tolerate invasive procedures (eg, pulmonary an-
giography), or exposure to radiation was contraindicated
(eg, pregnancy). Since the true rate of clinically recog-
nized recurrence is bounded by these 2 rates, we per-
formed 2 analyses for first VTE recurrence: (1) first over-
all VTE recurrence, defined as the first possible, probable,
or definite recurrent event, and (2) first probable/definite
VTE recurrence.
The relationship of the independent variables to time
to recurrent VTE was assessed using the Cox proportional
hazards model. We used stepwise and backwards condi-
tional logistic regression to identify a final model. Ini-
tially, Pⱕ.10 was required to enter the model. The result-
ing model was validated using a bootstrap method,21 in
which random samples of the same size as the incidence
cohort were drawn with replacement from these cases; a
stepwise proportional hazards model was run on the
sample, with Pⱕ.05 required for a variable to enter the
model. This was repeated 500 times, and the percentage
of times a variable came into the 500 models was used to
validate its presence in the final model. Variables were
considered validated and were retained in the final model
if they entered more than 70% of the models.21 When the
list of main effect variables was finalized, all 2-way inter-
actions of the main variables in the model were explored
using the bootstrap technique. Discrete variables with
multiple components could not be assessed using the
usual stepwise technique. Such variables were checked
separately, as were all their interactions. We used a modi-
fication of the bootstrap technique described above to
validate interactions with Pⱕ.05. We also checked all
2-way interactions of those variables that were no longer
in the model with those that were in the final list of main
effect variables. The proportional hazards assumption was
checked for all variables in the final model by testing
whether the interaction of time since the incident VTE
event and the variable was significant.
dence of both DVT and PE; 3 patients (0.1%) had chronic
thromboembolic pulmonary hypertension.
Of the 2218 incident cases, 494 either died on the
event date or were first discovered to have VTE at au-
topsy. Three additional patients had no follow-up avail-
able (all 3 died within a few days). One patient’s last medi-
cal contact occurred 201 days before her autopsy-
confirmed PE following an out-of-state motor vehicle
WWW.ARCHINTERNMED.COM
accident. One additional patient was excluded from most years for patients with DVT and 6.1 years for patients with
analyses because of missing values, leaving 1719 pa- PE. During the follow-up period, 404 patients had a to-
tients available for multivariate analysis. Of the patients tal of 588 recurrences. The first such recurrence was a
diagnosed prior to death, 96% received standard treat- possible recurrence in 205 of the 404 patients, probable
ment; 86% received anticoagulation therapy, with recurrence in 50, and definite recurrence in 149. The es-
mean ± SD durations of intravenous heparin and oral an- timated cumulative incidence of first overall VTE recur-
ticoagulation therapy of 5.9 ± 4.2 days (median, 5) and rence was 1.6% at 7 days, 5.2% at 30 days, 8.3% at 90
197.7 ± 559 days (median, 88), respectively. Other stan- days, 10.1% at 180 days, 12.9% at 1 year, 16.6% at 2 years,
dard treatments included subcutaneous heparin (n = 48) 22.8% at 5 years, and 30.4% at 10 years (Table 1 and
in the early years of the study; inferior vena cava liga- Figure 1). The hazard rate per 1000 person-days (±SD)
tion, interruption, or filter (n = 33) or other treatment for the first overall recurrence was highest in the first 6
(eg, vein ligation or thrombectomy; n = 13) for patients to 12 months after the initial event, ranging from 170 ± 30
in whom anticoagulation therapy was contraindicated; recurrent VTE events at 7 days to 130 ± 20 events at 30
and thrombolytic therapy (n = 14) or pulmonary embo- days, 30 ± 5 events at 90 days, 20 ± 4 events at 180 days,
lectomy (n = 3) for patients with phlegmasia or hypo- and 20 ± 2 events at 1 year. However, the recurrence haz-
tension. Patients not treated were often terminally ill with ard rate never fell to zero, continuing at 10 ± 1 events at 2
cancer. years, 6 ± 1 events at 5 years, and 5 ± 1 events at 10 years.

CUMULATIVE INCIDENCE OF FIRST OVERALL CUMULATIVE INCIDENCE OF FIRST

VTE RECURRENCE PROBABLE/DEFINITE VTE RECURRENCE

The 1719 patients were followed up for a total of 10 198 Of the 404 patients with a VTE recurrence, 240 had at
person-years. The median duration of follow-up was 7.4 least one probable/definite recurrence (probable in 59 pa-
tients and definite in 181 patients). The numbers of first
probable and first definite recurrences described herein
Table 1. Kaplan-Meier Estimates of the Cumulative are higher than the numbers described above, since if a
Incidence of First Overall or First Probable/Definite Venous
Thromboembolism Recurrence Among Olmsted County,
possible recurrence preceded a probable/definite recur-
Minnesota, Residents With a First Lifetime Venous rence in the analysis of time to first overall recurrence,
Thromboembolism Diagnosed From 1966 Through 1990 the subsequent probable/definite recurrence was not
counted. The estimated cumulative incidence of first defi-
Probable/Definite nite or probable recurrence was 0.2% at 7 days, 1.4% at
Overall Recurrence Recurrence 30 days, 3.1% at 90 days, 4.1% at 180 days, 5.6% at 1
Time to No. at Cumulative No. at Cumulative
year, 7.6% at 2 years, 12.4% at 5 years, and 17.6% at 10
Recurrence Risk Incidence, % Risk Incidence, % years (Table 1 and Figure 2). The hazard rate per 1000
0d 1720 0.0 1720 0.0
person-days (±SD) for first probable/definite recurrence
7d 1638 1.6 1656 0.2 also was highest in the first 6 to 12 months after the ini-
30 d 1502 5.2 1563 1.4 tial event, ranging from 30 ± 10 recurrent VTE events at
90 d 1389 8.3 1462 3.1 7 days to 50 ± 10 events at 30 days, 20 ± 4 events at 90
180 d 1292 10.1 1366 4.1 days, 10 ± 3 events at 180 days, and 10 ± 1 events at 1
1y 1171 12.9 1256 5.6 year. Similar to first overall recurrence, the hazard rate
2y 1002 16.6 1102 7.6
for the first probable/definite recurrence also never fell
5y 723 22.8 821 12.4
10 y 418 30.4 501 17.6 to zero, continuing at 6 ± 1 events at 2 years and 4 ± 1
events at 5 and 10 years.

Overall Probable/Definite
40 200 40 200
Hazard Rate, per 1000 Person-Days (

Hazard Rate, per 1000 Person-Days (

)

)
Cumulative Recurrence, % (

Cumulative Recurrence, % (

30 150 30 150

20 100 20 100

10 50 10 50

0 0 0 0
)

0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time Since Incident VTE Event, y Time Since Incident VTE Event, y

Figure 1. Kaplan-Meier estimates of cumulative incidence of first overall Figure 2. Kaplan-Meier estimates of cumulative incidence of first
venous thromboembolism (VTE) recurrence and hazard of first overall probable/definite venous thromboembolism (VTE) recurrence and hazard of
recurrence per 1000 person-days among Olmsted County, Minnesota, first probable/definite recurrence per 1000 person-days among Olmsted
residents with a first life-time VTE diagnosed from 1966 through 1990. County, Minnesota, residents with a first life-time VTE diagnosed from 1966
through 1990.

(REPRINTED) ARCH INTERN MED/ VOL 160, MAR 27, 2000 WWW.ARCHINTERNMED.COM
764
Downloaded from www.archinternmed.com on April 12, 2010
©2000 American Medical Association. All rights reserved.
 PREDICTORS OF VTE RECURRENCE
Table 2. Univariate Cox Proportional Hazards Analyses of
Hazard ratios and 95% confidence intervals for more than Potential Predictors of First Overall or First Probable/Definite
Venous Thromboembolism Recurrence Among Olmsted
25 baseline characteristics tested as predictors of VTE re- County, Minnesota, Residents With a First Lifetime Venous
currence are presented in Table 2 (univariate analyses). Thromboembolism Diagnosed From 1966 Through 1990*
In general, “persistent or irreversible” characteristics pres-
ent at the initial VTE were associated with an increased Hazard Ratio (95% CI)†
risk of recurrence, both overall and for probable/ Baseline
Characteristic Overall Definite/Probable
definite recurrence. These characteristics included in-
creasing age, being male, institutionalization, inflamma- Event year‡ 0.99 (0.98-1.01) 1.06 (1.04-1.08)
Age§ 1.18 (1.12-1.24) 1.36 (1.27-1.47)
tory bowel disease, neurologic disease associated with Male 1.29 (1.06-1.57) 2.07 (1.60-2.67)
extremity paresis, previous central venous catheter or Weight (BMI, kg/m2)
transvenous pacemaker placement, and malignant neo- Underweight (⬍20) 0.72 (0.45-1.14) 0.67 (0.35-1.27)
plasm (with and without chemotherapy). Several addi- Normal (20 to 24) 1.00 1.00
Overweight (⬎24 to ⬍30) 1.12 (0.88-1.42) 1.24 (0.91-1.69)
tional persistent characteristics were variably associated Obese (ⱖ30) 1.21 (0.92-1.60) 1.15 (0.80-1.66)
with an increased risk of either overall or probable/ Smoking history
definite recurrence, including the incident event year, None 1.00 1.00
smoking history, congestive heart failure, other heart dis- Current 0.96 (0.75-1.22) 0.84 (0.60-1.17)
Former 1.06 (0.82-1.37) 1.40 (1.02-1.91)
ease, chronic lung disease, and chronic renal disease. Body Incident PE with or without DVT 1.07 (0.88-1.30) 1.15 (0.89-1.48)
mass index, previous superficial vein thrombosis, vari- vs incident DVT only
cose veins, serious liver disease, and nephrotic syn- Probable/definite vs possible 1.09 (0.89-1.34) 2.11 (1.62-2.74)
drome were not predictors of recurrence. incident event
Previous superficial vein 1.08 (0.79-1.48) 1.00 (0.66-1.51)
Of equal importance, most “transient or revers- thrombosis
ible” baseline characteristics either were associated with Varicose veins 0.88 (0.70-1.11) 0.85 (0.63-1.14)
a reduced risk of recurrence or were not predictive of re- Anticoagulation 1.19 (0.69-2.07) 1.36 (0.67-2.76)
currence. For women, pregnancy or the postpartum state Location at onset
Community 1.00 1.00
and oral contraceptive use were associated with a re- Community, recent 1.01 (0.79-1.30) 0.98 (0.70-1.37)
duced risk of recurrence, both overall and for probable/ hospitalization
definite recurrence. Similarly, gynecologic surgery was Hospital 1.29 (1.02-1.64) 1.46 (1.08-1.98)
associated with a reduced risk of overall recurrence. Ad- Nursing home 1.38 (0.88-2.15) 1.68 (0.94-3.00)
Heart disease
ditional transient baseline characteristics that had no sig- No CHF or other heart disease 1.00 1.00
nificant effect on recurrence risk included the incident CHF 1.43 (1.04-1.97) 1.54 (0.99-2.38)
event type (DVT vs PE), anticoagulation prophylaxis im- Other heart disease 1.08 (0.79-1.47) 1.53 (1.05-2.24)
mediately preceding or at the time of the incident event, Chronic lung disease 1.25 (0.90-1.74) 1.88 (1.27-2.78)
Serious liver disease 0.99 (0.41-2.40) 1.09 (0.35-3.41)
immobilization, general surgery, orthopedic surgery, Inflammatory bowel disease 2.37 (1.12-5.01) 2.57 (1.06-6.25)
trauma, fracture, hormone therapy, and tamoxifen Chronic renal disease 1.61 (0.67-3.90) 2.83 (1.05-7.62)
therapy. Neurosurgery within the 3 months preceding Nephrotic syndrome 0.54 (0.08-3.87) 0.95 (0.13-6.80)
the incident VTE event was the only “transient” charac- Neurologic disease with 1.92 (1.33-2.77) 2.02 (1.24-3.27)
extremity paresis
teristic that was associated with an increased risk, both
Prior central venous catheter or 1.62 (1.06-2.50) 2.14 (1.26-3.62)
for overall and probable/definite recurrence. pacemaker
In the multivariate analyses of first overall VTE re- Malignant neoplasm
currence (Table 3) or first probable/definite VTE re- None 1.00 1.00
currence (Table 4), many persistent baseline charac- Malignant neoplasm with 3.57 (2.20-5.78) 7.53 (4.29-13.22)
chemotherapy
teristics remained significant independent predictors of Malignant neoplasm without 2.56 (1.86-3.51) 3.16 (2.10-4.75)
increased recurrence risk. For example, the risk of over- chemotherapy
all VTE recurrence was increased by 17% per decade in- General surgery 1.14 (0.86-1.52) 1.22 (0.86-1.75)
crease in age at incident VTE and by 24% per 10-point2 Orthopedic surgery 0.85 (0.61-1.18) 0.66 (0.41-1.07)
Neurologic surgery 2.25 (1.37-3.72) 2.59 (1.41-4.75)
increase in BMI. The level of diagnostic certainty was an Trauma 1.03 (0.76-1.38) 0.92 (0.61-1.38)
independent predictor of probable/definite recurrence; Fracture 0.97 (0.69-1.37) 0.96 (0.61-1.50)
patients with a probable or definite initial DVT or PE had Women only
a 56% increased risk of recurrence compared with pa- Pregnancy or postpartum 0.44 (0.24-0.78) 0.20 (0.06-0.63)
Postpartum 0.44 (0.22-0.90) 0.20 (0.05-0.82)
tients with a possible incident event. Neurologic disease Oral contraceptive use 0.58 (0.35-0.94) 0.30 (0.12-0.73)
with extremity paresis was an independent predictor of Estrogen or progesterone 0.90 (0.53-1.52) 0.63 (0.28-1.44)
overall recurrence; however, the risk of probable/ replacement therapy
definite recurrence among patients with neurologic dis- Estrogen replacement therapy 1.02 (0.60-1.72) 0.70 (0.31-1.59)
Gynecologic surgery 0.48 (0.24-0.97) 0.49 (0.18-1.32)
ease varied by patient location at incident event onset; Tamoxifen therapy 1.05 (0.15-7.47) 3.08 (0.43-22.25)
patients with neurologic disease requiring hospital con-
finement at the time of the initial event had a more than *CI indicates confidence interval; BMI, body mass index; PE, pulmonary
3.5-fold increased risk of recurrence. Compared with fe- embolism; DVT, deep vein thrombosis; and CHF, congestive heart failure.
male patients with VTE without malignant neoplasm, male †Unless otherwise specified, the comparison group consists of incident
cases without the indicated disease or condition.
patients with VTE without malignant neoplasm had a ‡Per 1 calendar year increase.
higher risk of recurrence for all ages and incident event §Per decade increase in age.

(REPRINTED) ARCH INTERN MED/ VOL 160, MAR 27, 2000 WWW.ARCHINTERNMED.COM
765
Downloaded from www.archinternmed.com on April 12, 2010
©2000 American Medical Association. All rights reserved.
 COMMENT
Table 3. Multivariate Cox Proportional Hazards Analysis
of Predictors of First Overall Venous Thromboembolism
Recurrence Among Olmsted County, Minnesota,
We determined VTE recurrence in an inception cohort
Residents With a First Lifetime Venous Thromboembolism of patients from a well-defined geographic population18
Diagnosed From 1966 Through 1990* that included the full spectrum of disease in all clinical
settings (the community, nursing home, and hospital)
Hazard Ratio where VTE may occur. In addition to our access to both
Baseline Characteristic (95% CI) outpatient and inpatient medical records, we used infor-
Age† 1.17 (1.11-1.24) mation from autopsy findings and death certificates to
Body mass index‡ 1.24 (1.04-1.47) ensure essentially complete ascertainment of all recog-
Neurologic disease with extremity paresis 1.87 (1.28-2.73) nized disease. Moreover, we stratified our analysis by ei-
Malignant neoplasm ther first overall recurrence or first probable/definite re-
None 1.00
currence alone to provide estimates of the respective
Malignant neoplasm with chemotherapy 4.24 (2.58-6.95)
Malignant neoplasm without chemotherapy 2.21 (1.60-3.06) maximum and minimum true recurrence rates. Finally,
Neurologic surgery and event 5-year interval we systematically evaluated a large number of baseline
1965 characteristics as potential predictors of recurrence. Our
No surgery 1.00 results demonstrate two important findings. First, VTE
Surgery 7.95 (2.67-23.72) recurs frequently and continues to recur for at least 10
1970
years after the incident event and possibly longer. The
No surgery 1.00
Surgery 5.12 (2.34-11.22)
hazard of recurrence was particularly high in the first 6
1975 to 12 months after the initial event but never fell to zero.
No surgery 1.00 Second, patients can be stratified into higher- and lower-
Surgery 3.30 (1.89-5.77) risk categories for recurrence according to several per-
1980 sistent or irreversible baseline characteristics.
No surgery 1.00 In general, our observed VTE recurrence rates are
Surgery 2.13 (1.25-3.62)
1985
higher than most previously reported rates. We believe
No surgery 1.00 our higher rates are best explained by more complete case
Surgery 1.37 (0.66-2.84) ascertainment and the inclusion of patients with the full
1990 spectrum of disease. Other cohort studies and clinical tri-
No surgery 1.00 als included only patients with DVT alone2,8,14,16 or pa-
Surgery 0.88 (0.32-2.46) tients with PE alone (with or without DVT),1,13,22 failed
to separate incident from recurrent events,1,10,12,13,16,22,23
*CI indicates confidence interval.
†Per decade increase in age.
or only reported PE recurrence rather than all VTE re-
‡Per 10-point increase in body mass index. currence.1,11 Since DVT or PE tends to recur in the same
form as the initial clinical manifestation,6,15,23 studies that
did not include the full spectrum of disease may not es-
years. Older patients without malignant neoplasm had a timate recurrence rates accurately. Moreover, many clini-
higher risk of probable/definite recurrence than younger cal trials excluded patients with the highest risk of re-
patients for both sexes. currence (eg, patients with familial thrombophilia, cancer,
Patients with malignant neoplasm had a more than paresis, or prolonged immobility).6,10,24 Our higher rates
2-fold increased risk of overall recurrence, and patients with cannot be explained by inadequate therapy. Over the
malignant neoplasm receiving concurrent chemotherapy course of our study, the standard of care for both DVT
had a more than 4-fold increased risk of overall recur- and PE was intravenous unfractionated heparin therapy
rence (Table 3). While the risk of probable/definite recur- overlapping with oral anticoagulation (eg, warfarin so-
rence also was increased among patients with malignant dium) therapy for 5 to 7 days, followed by at least 3
neoplasm, the risk varied by concurrent chemotherapy, age, months of warfarin anticoagulation therapy. The target
sex, and incident event year (Table 4). Concurrent che- intensity of anticoagulation was a 1.5- to 2.5-fold in-
motherapy interacted with sex and also with age; age also crease in the activated partial thromboplastin time and
interacted with event year. Among patients with malig- a 2- to 3-fold increase in the prothrombin time ratio (the
nant neoplasm who were receiving chemotherapy, male pa- international normalized ratio [INR] system was unavail-
tients and older patients had a higher recurrence risk. In able during the period of this study). Nearly all patients
contrast, female patients with malignant neoplasm who were received standard heparin and warfarin anticoagulation
not receiving chemotherapy had a higher risk for probable/ therapy for mean durations of 6 days and 6 months, re-
definite recurrence compared with similar male patients, spectively, which is consistent with the current stan-
for all ages and incident event years. dard of care.7 Moreover, the level of diagnostic certainty
Neurosurgery within the 3 months preceding the in- of the initial VTE event did not interact with other pre-
cident event was the only transient risk factor that re- dictors of either first overall or first probable/definite re-
mained an independent predictor of VTE recurrence (Table currence, suggesting that no bias was incurred by in-
3). However, neurosurgery was an independent predic- cluding possible cases. We believe our observed rates may
tor only for overall recurrence, and the risk varied by in- still underestimate the true recurrence rate because pa-
cident event year such that by 1985, neurosurgery no longer tients with unsuspected recurrent fatal VTE (eg, PE) who
was an independent predictor of recurrence. did not undergo an autopsy were missed.

(REPRINTED) ARCH INTERN MED/ VOL 160, MAR 27, 2000 WWW.ARCHINTERNMED.COM
766
Downloaded from www.archinternmed.com on April 12, 2010
©2000 American Medical Association. All rights reserved.
 Table 4. Multivariate Cox Proportional Hazards Analysis of Predictors of First Probable/Definite Venous Thromboembolism
Recurrence Among Olmsted County, Minnesota, Residents With a First Lifetime Venous Thromboembolism Diagnosed
From 1966 Through 1990*

Baseline Characteristic Hazard Ratio (95% CI)

Probable/definite vs possible incident event 1.56 (1.13-2.14)
Patient location at event onset and neurologic disease
Community
No neurologic disease 1.00
Neurologic disease 0.57 (0.14-2.29)
Nursing home
No neurologic disease 1.00
Neurologic disease 2.01 (0.67-6.02)
Hospital
No neurologic disease 1.00
Neurologic disease 3.60 (1.81-7.16)

Hazard Ratio (95% CI)

Study
Characteristic Age, y Year Female Male
No malignant neoplasm 45 1965 1.00 1.91 (1.43-2.54)
1970 1.28 (1.07-1.52) 2.43 (1.75-3.39)
1980 2.08 (1.22-3.52) 3.96 (2.19-7.16)
1990 3.38 (1.40-8.16) 6.45 (2.58-16.12)
60 1970 2.56 (1.86-3.51) 4.88 (3.21-7.41)
1980 3.23 (1.91-5.45) 6.15 (3.41-11.11)
1990 4.07 (1.88-8.79) 7.76 (3.42-17.57)
75 1970 5.12 (3.14-8.37) 9.77 (5.60-17.06)
1980 5.01 (2.88-8.72) 9.55 (5.13-17.78)
1990 4.90 (2.35-10.21) 9.34 (4.32-21.11)
Malignant neoplasm with chemotherapy 45 1970 6.46 (2.03-20.56) 19.17 (7.39-49.76)
1980 10.52 (3.11-35.57) 31.20 (11.36-85.73)
1990 17.13 (4.35-67.43) 50.79 (15.67-164.56)
60 1970 14.50 (6.04-34.78) 42.99 (18.36-100.66)
1980 18.29 (7.11-47.05) 54.22 (21.88-134.39)
1990 23.06 (7.84-67.87) 68.39 (24.34-192.21)
75 1970 32.51 (11.95-88.44) 96.40 (30.21-307.64)
1980 31.78 (11.77-85.80) 94.23 (30.00-295.97)
1990 31.06 (10.93-92.39) 92.10 (27.73-305.89)
Malignant neoplasm without chemotherapy 45 1970 14.09 (6.73-29.50) 8.14 (3.65-18.18)
1980 22.94 (9.64-54.57) 13.25 (5.17-33.98)
1990 37.33 (12.45-111.91) 21.57 (6.67-69.75)
60 1970 15.09 (7.35-31.00) 8.72 (4.56-16.70)
1980 19.04 (8.43-42.99) 11.00 (5.01-24.13)
1990 24.01 (9.04-63.74) 13.88 (5.22-36.90)
75 1970 16.16 (6.39-40.91) 9.34 (4.39-19.89)
1980 15.80 (6.16-40.49) 9.13 (4.06-20.55)
1990 15.44 (5.47-43.59) 8.92 (3.44-23.18)

*CI indicates confidence interval.

We have confirmed and extended the observation of
others suggesting that patients with VTE can be stratified
into high- and low-risk categories for recurrence accord-
ing to persistent or transient baseline patient characteris-
tics.2,7,8,10,12,16 For example, we confirmed that active can-
cer is a predictor of recurrence.2,8,23 However, among cancer
patients with VTE, the risk of recurrence can be further
refined based on age, sex, and the presence of concurrent
cytotoxic or immunosuppresive chemotherapy. In gen-
eral, recurrence risk was increased for male cancer pa-
tients receiving chemotherapy, for female cancer patients
not receiving chemotherapy, and for older cancer pa-
tients regardless of sex or therapy. While data regarding
cancer type, histologic characteristics, and stage were not
collected, most male cancer patients receiving chemo-
therapy have cancers unresponsive to hormonal therapy,
while most female cancer patients not receiving chemo-
therapy have breast cancer. Consequently, we speculate
that the risk of recurrence is greatest among male cancer
patients receiving chemotherapy for nonprostate cancer
and among female patients with breast cancer. This hy-
pothesis should be confirmed in studies of patients af-
fected with these particular malignant neoplasms.
Increasing age has been reported as an indepen-
dent predictor of reduced risk of recurrence among pa-
tients with DVT alone.8 In contrast, we found increas-
ing age to be an independent predictor of increased overall
recurrence risk. Aside from the difference in study popu-
lations and our more complete case ascertainment, we
cannot explain this discrepancy.

(REPRINTED) ARCH INTERN MED/ VOL 160, MAR 27, 2000 WWW.ARCHINTERNMED.COM
767
Downloaded from www.archinternmed.com on April 12, 2010
©2000 American Medical Association. All rights reserved.
 Of equal importance, many transient or reversible
baseline characteristics suggested as risk factors for inci-
dent VTE were not predictors of recurrence. In particu-
lar, an incident PE (eg, the most serious initial presenta-
tion of VTE) was not an independent predictor of
recurrence, nor was other evidence of venous disease,
such as previous superficial vein thrombosis or varicose
veins. Moreover, failure of anticoagulation prophylaxis
did not predict recurrence, nor did other reversible base-
line characteristics, such as smoking history, immobili-
zation, surgery, trauma, fracture, hormone therapy, and,
among women, pregnancy or the postpartum period,
oral contraceptive use, tamoxifen therapy, and gyneco-
logic surgery.
Our findings have several important implications.
First, among a subset of patients, VTE is a chronic dis-
ease with acute exacerbations. For these patients, long-
term anticoagulation therapy after an initial VTE epi-
sode may offer significant benefit. Because the hazard of
recurrence is especially high during the first 6 to 12
months after an incident event, we speculate that a stan-
dard intensity of anticoagulation therapy (eg, target INR,
2.5; INR range, 2.0-3.0) may provide optimal benefit dur-
ing this period despite the associated increase in bleed-
ing risk. After 12 months, the hazard of recurrence de-
creases such that a lower intensity of anticoagulation
therapy (eg, target INR, 1.85)7 may be equally effective
with less risk of anticoagulant-related bleeding. Second,
among a different subset of patients, VTE is a self-
limited disease precipitated by a transient exposure. For
these patients, a shortened course of anticoagulation
therapy may offer similar benefit with less risk of bleed-
ing. Additional clinical trials are warranted to address these
hypotheses, and we have identified predictors of recur-
rence that will be useful in the design of such trials. Fi-
nally, additional studies addressing familial25-27 or ac-
quired28 procoagulant disorders (thrombophilia) as
predictors of recurrence are needed. We believe such stud-
ies will allow physicians to better stratify recurrence risk
and target long-term anticoagulation therapy to those at
highest risk for recurrence.
Accepted for publication July 14, 1999.
This study was funded in part by grants HL46974 and
AR30582 from the National Institutes of Health, Bethesda
Md; by a grant from the US Public Health Service, Wash-
ington, DC; and by a grant from the Mayo Foundation, Roch-
ester, Minn.
The authors thank C. Mary Beard, RN, for assistance
in managing the study; Janet Ebersold, RN, Kay Traverse,
RN, Mary Lou Notermann, RN, and Susan Stotz, RN, for
untiring medical record review; Randall Stick, BS, for pro-
gramming; Diana Rademacher, BS, and Christine Lohse, BS,
for assistance with data analysis; and Stephanie Wellik for
secretarial support.
Reprints: John A. Heit, MD, Hematology Research,
Plummer 549, Mayo Clinic, 200 First St SW, Rochester, MN
55905 (e-mail: heit.john@mayo.edu).
(REPRINTED) ARCH INTERN MED/ VOL 160, MAR 27, 2000
768
Downloaded from www.archinternmed.com on April 12, 2010
©2000 American Medical Association. All rights reserved.
REFERENCES
1. Carson JL, Kelley MA, Duff A, et al. The clinical course of pulmonary embolism.
N Engl J Med. 1992;326:1240-1245.
2. Prandoni P, Lensing AWA, Cogo A, et al. The long-term clinical course of acute
deep venous thrombosis. Ann Intern Med. 1996;125:1-7.
3. Bergqvist D, Jendteg S, Johansen L, Persson U, Ödegaard K. Cost of long-term
complications of deep venous thrombosis of the lower extremities: an analysis
of a defined patient population in Sweden. Ann Intern Med. 1997;126:454-457.
4. Hull R, Delmore T, Genton E, et al. Warfarin sodium versus low-dose heparin in
the long-term treatment of venous thromboembolism. N Engl J Med. 1979;301:
855-858.
5. Brandjes DP, Heijboer H, Büller H, deRijk M, Jagt H, ten Cate JW. Acenocouma-
rol and heparin compared with acenocoumarol alone in the initial treatment of
proximal vein thrombosis. N Engl J Med. 1992;327:1485-1489.
6. Hirsh J. Duration of anticoagulant therapy after first episode of venous thrombosis
in patients with inherited thrombophilia. Arch Intern Med. 1997;157:2174-2177.
7. Schulman S, Rhedin A-S, Lindmarker P, et al. A comparison of six weeks with
six months of oral anticoagulant therapy after a first episode of venous throm-
boembolism. N Engl J Med. 1995;332:1661-1665.
8. White RH, Zhou H, Romano PS. Length of hospital stay for treatment of deep
venous thrombosis and the incidence of recurrent thromboembolism. Arch In-
tern Med. 1998;158:1005-1010.
9. Anderson FA Jr, Wheeler HB, Goldberg RJ, et al. A population-based perspective of
the hospital incidence and case-fatality rates of deep vein thrombosis and pulmo-
nary embolism: the Worchester DVT Study. Arch Intern Med. 1991;151:933-938.
10. Research Committee of the British Thoracic Society. Optimum duration of anti-
coagulation for deep-vein thrombosis and pulmonary embolism. Lancet. 1992;
340:873-876.
11. Kniffin WD JR, Baron JA, Barrett J, Birkmeyer JD, Anderson FA Jr. The epide-
miology of diagnosed pulmonary embolism and deep venous thrombosis in the
elderly. Arch Intern Med. 1994;154:861-866.
12. Beyth RJ, Cohen AM, Landefeld CS. Long-term outcomes of deep-vein throm-
bosis. Arch Intern Med. 1995;155:1031-1037.
13. van Beek EJR, Kuijer PMM, Büller HR, Brandjes DPM, Bossuyt PMM, ten Cate
JW. The clinical course of patients with suspected pulmonary embolism. Arch
Intern Med. 1997;157:2593-2598.
14. Hull RD, Raskob GE, Brant RF, Pineo GF, Valentine KA. The importance of initial
heparin treatment on long-term clinical outcomes of antithrombotic therapy: the
emerging theme of delayed recurrence. Arch Intern Med. 1997;157:2317-2321.
15. Douketis JD, Kearon C, Bates S, Duku EK, Ginsberg JS. Risk of fatal pulmonary
embolism in patients with treated venous thromboembolism. JAMA. 1998;279:
458-462.
16. Levine MN, Hirsh J, Gent M, et al. Optimal duration of oral anticoagulant therapy:
a randomized trial comparing four weeks with three months of warfarin in patients
with proximal deep vein thrombosis. Thromb Haemost. 1995;74:606-611.
17. Silverstein MD, Heit JA, Mohr DN, Petterson TM, O’Fallon WM, Melton LJ III.
Trends in the incidence of deep vein thrombosis and pulmonary embolism: a
25-year population-based study. Arch Intern Med. 1998;158:585-593.
18. Melton LJ. History of the Rochester Epidemiology Project. Mayo Clinic Proc. 1996;
71:266-274.
19. Kurland LT, Molgaard DA. The patient record in epidemiology. Sci Am. October
1981;245:54-63.
20. National Institutes of Health. Clinical guidelines on the identification, evaluation,
and treatment of overweight and obesity in adults: the evidence report [pub-
lished correction appears in Obes Res. 1998;6:464]. Obes Res. 1998;6(suppl 2):
51S-209S.
21. Saerbrei W, Schumacher M. A bootstrap resampling procedure for model build-
ing: application to the Cox regression model. Stat Med. 1992;11:2093-2109.
22. Simonneau G, Sors H, Charbonnier B, et al. A comparison of low-molecular-
weight heparin with unfractionated heparin for acute pulmonary embolism. N Engl
J Med. 1997;337:663-669.
23. The Columbus Investigators. Low-molecular-weight heparin in the treatment of
patients with venous thromboembolism. N Engl J Med. 1997;337:657-662.
24. Schulman S, Granqvist S, Holmström M, et al. The duration of oral anticoagu-
lation therapy after a second episode of venous thromboembolism. N Engl J Med.
1997;336:393-398.
25. Ridker PM, Miletich JP, Stampfer MJ, Goldhaber SZ, Lindpaintner K, Henne-
kens CH. Factor V Leiden and risks of recurrent idiopathic venous thromboem-
bolism. Circulation. 1995;92:2800-2802.
26. Simioni P, Prandoni P, Lensing AWA, et al. The risk of recurrent venous throm-
boembolism in patients with an Arg506-Gln mutation in the gene for Factor V
(Factor V Leiden). N Engl J Med. 1997;336:399-403.
27. van den Belt AGM, Sanson B-J, Simioni P, et al. Recurrence of venous throm-
boembolism in patients with familial thrombophilia. Arch Intern Med. 1997;157:
2227-2232.
28. Schulman S, Svenungsson E, Granqvist S, and the Duration of Anticoagulation
Study Group. Anticardiolipin antibodies predict early recurrence of thromboem-
bolism and death among patients with venous thromboembolism following an-
ticoagulant therapy. Am J Med. 1998;104:332-338.
WWW.ARCHINTERNMED.COM