The epidemiology of venous thromboembolism in Caucasians and African-Americans: the GATE Study1


  • N. F. Dowling,

    1. *Haematologic Diseases Branch, Division of AIDS, STD, and TB Laboratory Research, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA, USA; †Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA, USA; and ‡Mount Sinai School of Medicine, New York, NY, USA
    Search for more papers by this author
  • H. Austin,

    1. *Haematologic Diseases Branch, Division of AIDS, STD, and TB Laboratory Research, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA, USA; †Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA, USA; and ‡Mount Sinai School of Medicine, New York, NY, USA
    Search for more papers by this author
  • A. Dilley,

    1. *Haematologic Diseases Branch, Division of AIDS, STD, and TB Laboratory Research, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA, USA; †Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA, USA; and ‡Mount Sinai School of Medicine, New York, NY, USA
    Search for more papers by this author
  • C. Whitsett,

    1. *Haematologic Diseases Branch, Division of AIDS, STD, and TB Laboratory Research, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA, USA; †Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA, USA; and ‡Mount Sinai School of Medicine, New York, NY, USA
    Search for more papers by this author
  • B. L. Evatt,

    1. *Haematologic Diseases Branch, Division of AIDS, STD, and TB Laboratory Research, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA, USA; †Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA, USA; and ‡Mount Sinai School of Medicine, New York, NY, USA
    Search for more papers by this author
  • W. C. Hooper

    1. *Haematologic Diseases Branch, Division of AIDS, STD, and TB Laboratory Research, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA, USA; †Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA, USA; and ‡Mount Sinai School of Medicine, New York, NY, USA
    Search for more papers by this author

  • 1

    Use of trade names is for identification only and does not constitute endorsement by the U.S. Department of Health and Human Services, the Public Health Service, or the Centers for Disease Control and Prevention.

Dr Nicole F. Dowling, Centers for Disease Control and Prevention, 1600 Clifton Road, N.E., Mailstop E-64, Atlanta, GA 30333, USA.Tel.: + 404 371 5261, fax: + 404 371 5424, e-mail:


Summary.  The aim of this study was to assess, comprehensively, medical and genetic attributes of venous thromboembolism (VTE) in a multiracial American population. The Genetic Attributes and Thrombosis Epidemiology (GATE) study is an ongoing case–control study in Atlanta, Georgia, designed to examine racial differences in VTE etiology and pathogenesis. Between 1998 and 2001, 370 inpatients with confirmed VTE, and 250 control subjects were enrolled. Data collected included blood specimens for DNA and plasma analysis and a medical lifestyle history questionnaire. Comparing VTE cases, cancer, recent surgery, and immobilization were more common in caucasian cases, while hypertension, diabetes, and kidney disease were more prevalent in African-American cases. Family history of VTE was reported with equal frequency by cases of both races (28–29%). Race-adjusted odds ratios for the associations of factor V Leiden and prothrombin G20210A mutations were 3.1 (1.5, 6.7) and 1.9 (0.8, 4.4), respectively. Using a larger external comparison group, the odds ratio for the prothrombin mutation among Caucasians was a statistically significant 2.5 (1.4, 4.3). A case-only analysis revealed a near significant interaction between the two mutations among Caucasians. We found that clinical characteristics of VTE patients differed across race groups. Family history of VTE was common in white and black patients, yet known genetic risk factors for VTE are rare in African-American populations. Our findings underscore the need to determine gene polymorphisms associated with VTE in African-Americans.

Venous thromboembolism (VTE) is a common vascular disease and significant public health problem in the USA, affecting about 1 in 1000 individuals per year [1]. Deep vein thrombosis (DVT), the most frequent presentation of VTE, is associated with significant morbidity and mortality. The most serious complication of DVT, pulmonary embolism (PE), is a life-threatening condition with short-term survival of less than 60% [2].

VTE is a multifactorial disease, resulting from a complex interaction of genetic and acquired factors. Primary hypercoagulability due to inherited deficiencies in anticoagulant proteins may be present in between 5 and 10% of patients with VTE [3]. Acquired factors such as surgery, malignancy, and immobilization have been associated with increased propensity for thrombosis [4,5]. While both acquired and inherited factors play important roles in the pathogenesis of VTE, risk varies greatly from one individual to another, and the causes for many cases remain unidentified. Moreover, little is known about the importance of interaction between environmental factors and inherited coagulation abnormalities in the development of VTE.

In the past decade, two single nucleotide polymorphisms, factor (F)V Leiden (G1691A) and prothrombin G20210A, have been demonstrated to be risk factors for DVT and PE in European and American Caucasians [6–11]. Despite these advances, the determinants of VTE in African-Americans are not well understood. Few cases of VTE in American blacks can be attributed to the FV Leiden or prothrombin G20210A mutations, because these variants are exceedingly rare among African-Americans. Prevalence of the FV allele, while at least 3% in American whites [11], is about 0.4% in healthy African-Americans [12]. Similarly, birth prevalence of the prothrombin 20210 A allele in African-Americans has been reported to be 0.2% [13], compared with an allele prevalence of 1% for healthy Caucasians [9]. Recent research indicates that incidence of idiopathic VTE in the USA may be higher for African-Americans than for Caucasians, yet inherited factors determining elevated VTE risk among blacks have not been elucidated [14]. Furthermore, the excess VTE risk for blacks has not been explained by elevated prevalence of other medical and surgical conditions.

To date, most epidemiologic studies of VTE risk factors have been conducted within white American and European populations. Moreover, few studies have evaluated interactive effects of these factors on risk of VTE. The goals of the present study are to determine, in an American population, genetic and environmental factors associated with VTE, with a primary emphasis on elucidating race-specific aspects of the etiology and pathogenesis of this complex disease. In this paper, we present a detailed description of the study methods and results of an initial epidemiologic analysis of clinical and lifestyle characteristics of the study population. Additionally, we have investigated the associations of the FV Leiden, prothrombin G20210A, and 5,10-methylenetetrahydrofolate reductase (MTHFR) C677T variants with VTE by using several alternative statistical methods.

Materials and methods

Subject enrollment

The Genetic Attributes and Thrombosis Epidemiology (GATE) study is an ongoing case–control study of risk factors for VTE. Subject enrollment commenced in March 1998 and will continue though 2003. The study protocol was approved by the Institutional Review Boards at the participating institutions.

Patients, aged 18–70 years, hospitalized at two university-owned hospitals in Atlanta, Georgia with recently diagnosed first or recurrent episodes of DVT and/or PE are eligible as cases in the study. Potential cases are identified from a daily review of medical charts of all patients at the two hospitals receiving unfractionated or low-molecular-weight heparin. A DVT is objectively confirmed when diagnosed by Doppler ultrasonography, computed tomography (CT), magnetic resonance imaging, or contrast venography. Diagnosis of PE is made after positive angiogram, ventilation–perfusion lung scan, or CT. Patients with severe illness or with cognitive deficits, who are not able to complete study activities, are considered ineligible.

Control subjects were selected from a list of patients who visited the office of one of 10 physicians at a university-affiliated primary care clinic between January 1, 1997 and December 31, 2000. A patient list was obtained from the clinic's computerized patient accounts database. The master list was sampled to obtain a randomly ordered subset list of potential controls approximately similar to cases in age, sex, and race distributions. Individuals with a history of VTE, currently taking anticoagulant medication, or with a mental or physical problem precluding participation are not eligible to participate in the study.

Data collection

A whole blood sample is drawn from each hospitalized case subject and sent to a Centers for Disease Control and Prevention (CDC) laboratory for genetic analysis. Control subjects have a single blood sample, for genetic and plasma functional analysis, drawn at CDC at the time of enrollment. A second blood sample is obtained from case subjects, during a follow-up appointment at the CDC laboratory, at least 1 month after completion of anticoagulation therapy and at least 3 months after the index thrombotic event. The plasma samples will be used in future analyses to determine plasma levels of clotting factors, anticoagulant proteins, and other components affecting the coagulation system.

Each participant is interviewed at the time of enrollment. The 45-min interview takes place in the hospital for cases and at CDC for controls. Questions were designed to elucidate lifestyle, environmental, and medical factors that may be associated with VTE. Questions cover demographics, medical history, personal and family history of thrombosis, smoking and alcohol use, current medications, reproductive history and use of contraceptive and replacement hormones, physical activity, and diet and supplements. Specific questions are asked about life events and conditions occurring within the 4 weeks preceding VTE diagnosis for cases and within 4 weeks preceding enrollment in the study for controls. These life events and conditions include surgery, bed rest of more than 2 days duration, injury requiring medical treatment, travel of more than 8 hours' duration, and confinement to a wheelchair. Finally, a detailed medical records review is conducted for each enrolled case subject.

Enrollment numbers for current analysis

As of 1 March 2001, we identified 886 patients with DVT and/or PE at the two hospitals. Of these 886 patients, 147 were determined to be too ill to participate in the study and 20 others died before we could ask them about enrollment. Thirty-three additional patients were identified but had not yet been invited to participate. Of the 686 remaining patients, 387 (57%) agreed to enroll in the study, while 157 refused and 142 were lost to follow-up.

We sampled 616 control subjects from the patient lists of clinic physicians. Forty-two individuals were excluded because of history of VTE or current use of anticoagulant medications. Twelve were not contacted at the request of the physician. An additional 22 patients had not been contacted as of 1 March 2001. Of the 540 patients eligible to be control subjects in the study, 264 (49%) agreed to participate. A total of 151 patients refused and 125 individuals could not be located.

We excluded 12 cases and four controls with missing questionnaire or DNA data. As the present analysis includes only caucasian and African-American persons, five cases and 10 controls reporting a different racial background were excluded from the analytic file. The analysis for this paper includes 370 cases and 250 controls.

Laboratory methods

Blood samples were collected in 0.109 mol L−1 sodium citrate. DNA was extracted from the whole blood samples according to the manufacturer's protocol using the PuregeneTM kit from Gentra Systems, Inc. (Minneapolis, MN, USA) and then stored at − 20 °C. Allelic discrimination was used for DNA analysis as described by Benson and colleagues [15]. Polymerase chain reaction (PCR) primers and fluorogenic probes were designed and synthesized by PE/Applied Biosystems (Foster City, CA, USA) for the target regions. The probes were fluorescence-labeled with reporter dyes of FAM (6-carboxyfluorescein, 6 FAM) and VICTM on the 5′ ends for sequences determining mutation and wild-type, respectively. The probes were synthesized with a 3′-blocking phosphate, as well as a minor groove binder and a nonfluorescent quencher. PCR amplifications were performed in a GeneAmp PCR System 9600 (PE/Applied Biosystems). Final concentrations of reactant in a 20:l mixture containing 10–100 ng of total DNA were 900 nmol L−1 of each primer, 200 mol L−1 of each probe, and 1X TaqMan® Universal Master Mix. Following an initial cycle of 50 °C for 2 min and 95 °C for 10 min, each cycle consisted of 92 °C for 15 s and 60 °C for 2 min for 40 amplification cycles. The TaqMan® assay was subsequently used for mutation detection.

Statistical methods

Goals of the statistical analyses are to evaluate the associations of gene polymorphisms and environmental factors with VTE and to assess gene-environment and gene–gene interactions. Odds ratios, 95% confidence intervals (CIs) and two-tailed P-values were obtained by large-sample methods (Mantel–Haenszel, unconditional logistic regression) computed by SAS version 8.1 software (SAS Institute, Cary, NC, USA) [16,17]. In cases where a cell expected value was ≤ 5, conditional maximum likelihood estimates for odds ratios and mid-P exact CIs and P-values were used [18]. Student's t-tests were used to test case–control differences between continuous variables.

For each gene polymorphism, genotypes were classified as homozygous wild-type or heterozygous or homozygous for the variant allele. In all comparisons, the homozygous wild-type genotype was considered the referent group. Odds ratios were calculated as the odds of being a case for each genotype divided by the odds of being a case for the referent genotype. The interpretation of the odds ratio is the relative risk of VTE for subjects with that genotype compared with subjects with the referent genotype. The χ2 distribution was used to assess differences in allele frequencies between cases and controls and between racial groups. χ2 tests were also used to test the assumption of Hardy–Weinberg equilibrium for each polymorphism.

We evaluated multiplicative, two-way interaction between homozygosity/heterozygosity for FV Leiden, homozygosity/heterozygosity for prothrombin G20210A, and homozygosity for the MTHFR T allele in a case-only analysis [19,20]. GATE data describing interaction of these genes were presented recently in a methods paper of the case-only design (L. Botto et al., submitted for publication). Odds ratios bigger than unity in this analysis indicate more than a multiplicative effect of the two genetic factors. Case-only analysis requires that the exposures are statistically independent. Thus, we tested whether each of the three genetic factors was in linkage disequilibrium by obtaining a disequilibrium coefficient by maximum likelihood estimation [21]. Estimates in the vicinity of zero indicate no linkage disequilibrium and hence justify a critical assumption of the case-only analysis. We evaluated linkage disequilibrium among 4344 pooled control subjects obtained from the membership of a large health plan in California and enrolled at their annual physical examination [CDC controls, Thrombosis and Genes (TAG) study]. Our findings indicate the statistical independence of these genetic factors and support the use of case-only analyses.

We also evaluated two-way, multiplicative interaction between the three genetic factors using standard case–control methodology. However, these analyses were not possible using only GATE controls because the data are too sparse. Thus, we supplemented the GATE controls with the TAG study controls (assayed at the same laboratory as were GATE subjects). First, we evaluated whether or not the prevalence of the three genetic factors was statistically equivalent in the GATE and CDC TAG controls. We considered ‘exposure’ to be homozygosity or heterozygosity for the FV Leiden mutation, homozygosity or heterozygosity for the prothrombin G20210A variant, or homozygosity for the T allele of the MTHFR C677T polymorphism. The odds ratios of the ‘exposed’ genotypes for GATE controls compared with CDC TAG controls are 0.86, 1.6, and 0.75 for FV Leiden, prothrombin G20210A, and MTHFR C677T, respectively. No odds ratio is statistically significant (each P-value greater than 0.20). The similarity of these genetic factors of the GATE and CDC TAG controls provides a rationale for pooling these two groups.


The average age of case subjects (49.2 years) is comparable to the average age of control subjects (49.5 years). A diagnosis of DVT was given to 250 case subjects, while 74 received a diagnosis of PE only, and 46 a diagnosis of PE with concomitant DVT. Of the cases, 255 (69%) were enrolled in the study when diagnosed with a first episode of VTE, while 115 (31%) reported at least one previously diagnosed VTE. Case–control comparisons first were conducted separately for subjects with a first episode of VTE and for those with recurrent VTE. As we found no statistically significant differences in association estimates between these two groups, we have reported results of analyses for all case subjects together.

As displayed in Table 1, cases are less likely than controls to be of caucasian race, to have a college degree (P < 0.001), and to have an annual household income greater than $40 000 (P < 0.001). Cases reported family history of VTE more frequently than control subjects (28% vs. 12%). Current smoking is not associated with risk of VTE, but alcohol consumption of up to seven drinks a week confers a statistically significant reduced risk of VTE. Body mass index (BMI) is higher for cases than for controls. This case–control difference, however, is seen only among Caucasians, with average body BMI for patients with VTE of 28.9 kg m−2 compared with 26.4 kg m−2 for control subjects (P < 0.0001). In logistic regression analysis, adjusting for age, gender, and education, a comparison of Caucasians with BMI of 30 kg m−2 (obese) to a referent group with normal BMI indicates a relative risk of 3.3 (95% CI 1.7, 6.1). In African-American subjects, case and control BMI averages are higher than corresponding values for Caucasians but do not differ by case–control status (29.6 kg m−2 for cases vs. 30.6 kg m2 for controls, P > 0.20).

Table 1.  Characteristics of study participants
CharacteristicCases (%)
(N = 370)
Controls (%)
(N = 250)
OR (95% CI)
  • *

    Odds ratios and 95% confidence intervals.

  • †Income missing for one case subject.

  • ‡Odds ratios, adjusted for age, gender, race, and education, and 95% confidence intervals.

  • §

    Calculated excluding 76 subjects with ‘unknown’ family history.

 Male181 (49)125 (50)1.0*
 Female189 (51)125 (50)1.0 (0.8, 1.4)*
 African-American174 (47)91 (36)1.0*
 Caucasian196 (53)159 (64)0.6 (0.5, 0.9)*
Annual income
 <$24 999134 (37)29 (12)1.0*
 $25 000 – $39 99979 (22)47 (19)0.4 (0.1, 0.9)*
 $40 000 – $54 99938 (10)36 (14)0.2 (0.1, 0.7)*
 $55 000 – $70 00039 (11)39 (16)0.2 (0.1, 0.6)*
 >$70 00073 (20)*98 (39)0.2 (0.1, 0.4)*
 ≤High school graduate158 (43)37 (15)1.0*
 Some college96 (26)60 (24)0.4 (0.2, 0.6)*
 Junior college degree19 (5)14 (6)0.3 (0.1, 0.7)*
 Four-year college degree47 (13)58 (23)0.2 (0.1, 0.3)*
 Postgraduate work50 (13)81 (32)0.1 (0.09, 0.2)*
 Not current295 (80)209 (84)1.0
 Current75 (20)41 (16)1.0 (0.6, 1.5)
Alcohol use
 Rare/never119 (32)46 (18)1.0
 Light drinker177 (48)168 (67)0.6 (0.4, 0.9)
 Moderate drinker43 (12)22 (9)0.9 (0.4, 1.7)
 Heavy drinker30 (8)14 (6)0.6 (0.3, 1.4)
Family history of VTE§
 No220 (72)208 (88)1.0
 Yes87 (28)29 (12)2.3 (1.4, 3.8)
Body mass index (kg m−2)
 <18.510 (3)7 (3)0.7 (0.2, 2.1)
 18.5–24.991 (24)81 (32)1.0
 25.0–30.0126 (34)89 (36)1.3 (0.8, 2.0)
 ≥30.0143 (39)73 (29)1.5 (0.9, 2.3)

In Table 2, medical history factors are compared for cases and controls; results are calculated, adjusting for age, gender, race, and education. History of malignancy and of congestive heart failure are more frequent among subjects with VTE than among control subjects. History of diabetes and kidney failure are associated with elevated risk of VTE, but the results are not statistically significant. Case subjects are more likely than control subjects to report recent surgery, bed rest, injury, and confinement to a wheelchair.

Table 2.  Medical history of study participants
 Cases (%)
N = 370
Controls (%)
N = 250
OR*95% CI*
  • *

    Odds ratios, adjusted for age, gender, race, and education, and 95% confidence intervals.

Chronic diseases
 Cancer87 (24)33 (13)2.5(1.6, 4.1)
 Diabetes64 (17)23 (9)1.5(0.8, 2.5)
 Hypertension160 (43)94 (38)1.0(0.7, 1.5)
 Kidney disease27 (7)6 (2)2.3(0.9, 5.8)
 Heart failure31 (8)2 (1)9.2(2.1, 39.8)
Conditions in 4 weeks preceding VTE diagnosis or preceding enrollment in study (controls)
 Surgery140 (38)3 (1)50.8(15.8, 163.4)
 Bed rest > 2 days211 (57)10 (4)29.4(14.9, 58.1)
 Injury36 (10)6 (2)3.9(1.6, 9.8)
 Travel > 8 h50 (14)33 (13)1.1(0.7, 1.9)
 Wheelchair26 (7)2 (1)9.4(2.2, 40.9)

We evaluated differences in medical history and clinical characteristics between caucasian and African-American case subjects (Table 3). While history of cancer is more prevalent among white cases (31% vs. 15%), kidney disease, hypertension, and diabetes (though not statistically significant) were reported more frequently by black VTE cases. Recent surgery and immobilization (bed rest and wheelchair use) are more common among white subjects with VTE. Family history of VTE was reported with equal frequency by white (29%) and black (28%) cases. The average age for black case subjects at time of index event diagnosis is 47.5 years, a value that is significantly lower than the average age of 50.7 years for white cases (P = 0.02).

Table 3.  Comparison of characteristics of patients with VTE, by race
 Caucasians (n = 196)African-Americans (n = 174) P-value
Family history of VTE48293928>0.20
Diagnosis of PE66345431>0.20
First VTE1306612572>0.20
Chronic diseases
 Kidney disease8419110.01
 Heart failure1910127>0.20
Conditions in 4 weeks preceding VTE diagnosis
 Bed rest > 2 days1266485490.003
 Travel > 8 h321618100.09
Mean age (standard deviation)50.7 years(12.7)47.5 years(12.7)0.02

We calculated race-specific and race-adjusted odds ratios to assess the relationships between case–control status and FV Leiden, prothrombin G20210A, and MTHFR C677T genotypes (Table 4). Caucasians with the FV mutation have a three-fold increased risk of VTE; the relative risk for African-Americans is elevated but is not statistically significant. For whites, the prevalence of the prothrombin gene mutation is higher for cases than for controls (8.2% vs. 5.0%) but the difference does not reach statistical significance. Similarly, a race-adjusted odds ratio is above, though not statistically different from, unity; a relative risk among blacks could not be estimated because of the rarity of the mutation. The MTHFR variant is not associated with VTE among caucasian or African-American subjects. Frequencies of the FV Leiden A allele are 2.8% and 0.5% for whites and blacks, respectively. The prothrombin 20210 A allele is present in 2.5% of white control subjects but is absent in black controls. The T allele (MTHFR C6777T) prevalence also is highest in whites (35.5% vs. 11.5%).

Table 4.  Case–control comparison of genotype frequencies for factor V Leiden, prothrombin G20210A, and MTHFR C677T polymorphisms
N = 194N = 159OR*95% CI*N = 174N = 91OR*95% CI*
  1. T/T vs. C/C + C/T: Race-adjusted OR = 1.2 (0.6, 2.5)95%P-value for homogeneity > 0.20. *Odds ratios and 95% confidence intervals.

Factor V Leiden
 G/A + A/A14.5%5.0%3.2(1.4, 7.2)2.9%1.1%2.7(0.4, 64.0)
Race-adjusted OR = 3.1 (1.5, 6.7)95%P-value for homogeneity >0.20:
Prothrombin G20210A
 G/A + A/A8.2%5.0%1.7(0.7, 4.1)1.1%0%Inf.
Race-adjusted OR = 1.9 (0.8, 4.4)95%P-value for homogeneity > 0.20:
 C/T47.4%53.5%0.8(0.5, 1.2)13.2%23.1%0.5(0.3, 1.0)
 T/T9.3%8.8%0.9(0.4, 2.0)1.7%0%Inf.

The case-only analysis suggests a multiplicative, though not statistically significant, interaction between FV Leiden and prothrombin G20210A mutations on VTE risk (Table 5). There is evidence of interaction between prothrombin and MTHFR C677T variants on disease risk, but the effect is considerably smaller than that for FV Leiden and prothrombin. The joint effect of FV Leiden and MTHFR is smaller than the product of the marginal effects. The inclusion of the large number of CDC TAG study controls allows for direct evaluation of multiplicative interaction in a case–control analysis. This analysis supports the findings of the case-only analysis. The measure of multiplicative interaction between FV Leiden and the prothrombin mutation is somewhat increased in the case–control analysis, although the finding is not statistically significant (P = 0.11).

Table 5.  Case-only and case–control analysis of two way multiplicative interaction between VTE, factor V Leiden, prothrombin mutation, and MTHFR polymorphisms among caucasian subjects
 Case-only analysisCase–control analysis*
ORSE log(OR)P-valueORSE log(OR)P-value
  • *

    Pooled control subjects from GATE and TAG studies.

  • †Odds ratio for multiplicative interaction.

  • ‡Standard error of the logarithm of interaction odds ratio.

 Factor V Leiden *prothrombin2.370.62550.173.330.75310.11
 Factor V Leiden* MTHFR0.690.7852>0.200.670.8050>0.20

In addition, we evaluated the main effect of each genetic factor using the GATE cases compared with a combined group of GATE and CDC TAG study controls. The prothrombin G20210A mutation-VTE association is not statistically significant in the GATE-only analysis (Table 4). However, inclusion of the CDC TAG study controls increases the odds ratio to 2.5, and the finding is statistically significant (P = 0.001).


The GATE study is the first large case–control study designed to evaluate genetic, environmental, and medical factors related to VTE in an American population. Our initial findings suggest that clinical characteristics of patients hospitalized with VTE differ significantly by race. Among caucasian participants, history of cancer, recent surgery, immobilization, injury, and heart failure are more prevalent among individuals with VTE, compared with control subjects. Heit and colleagues [4] noted that these factors are important risk factors for VTE within a predominantly caucasian population in Minnesota. For African-American VTE patients in our study, another set of characteristics is more common. Comparing black and white cases, we found higher frequencies of diabetes, hypertension, and kidney disease among blacks yet lower frequencies than among whites of conditions such as surgery, cancer, immobilization, and injury. For blacks, frequencies of diabetes and hypertension were nearly as high in controls as in cases, probably reflecting a higher rate of these chronic diseases in African-Americans overall. While we do not have sufficient evidence to conclude that these observed differences are related to the etiology of VTE, we can state that two different sets of clinical characteristics are present in white and black hospitalized VTE patients in the GATE study.

The results of our comparison, by race, of surgical history of VTE patients differ from the findings of another recent study of VTE. White and colleagues used California hospital discharge data from 1991 to 1994 to compare, across ethnic groups, incidence rates of VTE [14]. Within the California population, African-American patients with secondary VTE, compared with Caucasians with the same discharge diagnosis, had higher rates of colon and hip surgery preceding VTE diagnosis. In contrast, we noted the prevalence of all surgery to be significantly higher for white VTE patients (45%) than for black VTE patients (29%). However, we have not stratified our data by specific surgical procedures. Another source of differing results may be White's use of hospital discharge data, information subject to recording errors and misclassification. In the GATE study, diagnosis of VTE for case subjects was carefully ascertained by systematic review of hospital radiology reports.

Our study currently includes inpatients with VTE at two large hospitals. We did not enroll individuals diagnosed with VTE and treated as outpatients. Although not including outpatient VTE cases may limit the generalizability of our study findings, we do believe that the number of individuals diagnosed with VTE and treated outside of the hospital between 1998 and 2001 in our study area is quite small. Moreover, we have no reason to conclude that type of treatment for VTE differed substantially by race during this period. Another possible limitation of our study design is our use of outpatient clinic controls. Controls were, on average, more educated and had higher annual income than case subjects. We do not believe these differences will impact on the genetic findings of our study. For all other analyses of clinical and environmental factors, we have adjusted for education. Therefore, potential for bias is minimized.

We report associations for the FV Leiden, prothrombin G20210A, and MTHFR C677T polymorphisms that are similar to estimates reported in the recent literature. In the past decade, epidemiologic studies of caucasian populations have identified the FV Leiden mutation as a cause of primary [6,11,22] and recurrent VTE [23]. The A allele of the prothrombin G20210A polymorphism has been associated with elevated prothrombin levels and with a threefold increased risk of VTE in whites [9,10]. Studies of the C677T variant of the MTHFR gene, a polymorphism associated with blood levels of homocysteine [24], have yielded conflicting results [25,26]. In caucasian subjects in the GATE study, FV Leiden is strongly associated with VTE, while prothrombin 20210 A confers a nonsignificant elevation of VTE risk. We found both mutations to be rare and thus not important determinants of VTE among blacks. The MTHFR 677T allele is not related to VTE in participants of either race.

The case-only analysis suggests a multiplicative interaction between FV Leiden and the prothrombin G20210A variant on VTE risk. Typically, the evaluation of interaction between rare genetic traits in case–control studies is impossible because of lack of statistical power. In the present study, this interaction model failed using just GATE controls because none of these controls had both the FV Leiden and the prothrombin mutations. With respect to the evaluation of the interaction odds ratio, the case-only method substitutes the need for a control group with the assumption that the two exposures are independent. For genetic traits, this assumption can be evaluated easily by consideration of linkage disequilibrium, so that genetic studies are good candidates for case-only analyses. Through our use of the historical CDC controls, we were able to increase the number of controls considerably and could evaluate statistical interaction between the two genetic mutations using standard case–control methodology. This analysis provided support for the validity of our case-only analysis. However, we note that no matter how large the control group, the case-only analysis will provide a more precise estimate of the interaction odds ratio if the two genetic traits are independent. This fact is evidenced by the slightly larger standard errors of the interaction odds ratios for the case–control analysis compared with the case-only analysis in Table 5.

We believe that the use of historic (external) controls in genetic studies is under-utilized. In the present study, our pool of CDC controls enabled us to evaluate linkage disequilibrium with high statistical power. At the least, this evaluation provided a strong justification for the case-only study. Additionally, the observation that the GATE and CDC control groups did not differ significantly with respect to the distribution of these genetic factors provides justification for pooling control groups. The odds ratio for the prothrombin mutation based on the GATE study becomes more positive and statistically significant using pooled controls. As we have no reason to suspect that the genetic background of whites in Atlanta is different from that of whites in California, and since our data suggest that the two groups do not differ on these genetic factors, we believe our study provides persuasive evidence that the prothrombin mutation is a cause of VTE.

Several recent studies have examined the effects of combinations of inherited risk factors on VTE risk among Caucasians. An interaction between hyperhomocysteinemia and the FV Leiden mutation has been reported by at least two research groups [27,28]. Cattaneo et al. [29] reported a statistically significant effect modification by the MTHFR C677T mutation on the association between the FV Leiden mutation and VTE. Findings from the Leiden Thrombophilia Study [25], however, did not support a role for the MTHFR variant in VTE risk among individuals with or without the FV Leiden mutation. Similarly, Brown et al. [30] and Alhenc-Gelas et al. [31] noted no significant interaction between the MTHFR C677T and either the FV Leiden or the prothrombin G20210A mutation. One recent study has demonstrated an increased risk for recurrent VTE among patients with both the FV Leiden and prothrombin G20210A mutations compared with the risk of recurrence among carriers of FV Leiden alone [32]. A second study analyzing pooled data from eight European case–control studies of VTE reported an odds ratio of 20.0 for double heterozygotes of FV Leiden and prothrombin mutations, a finding that represents a multiplicative interactive effect [33]. Our results, using data from a single case-series, suggest a multiplicative interaction between the FV Leiden and prothrombin G20210A mutations among Caucasians.

One of the more striking aspects of our analyses is that despite the rarity of known genetic risk factors among African-Americans, we found the prevalence of family history of VTE to be equal for black and white cases. This finding suggests that a strong genetic component exists in the etiology of VTE also among African-Americans. To date, the set of genetic factors responsible for a significant proportion of VTE cases among blacks remains undetermined. These results only underscore the need for research that addresses risk factors and etiologic mechanisms for VTE specific to an African-American population. Among both whites and blacks, an understanding of interactive effects between acquired and inherited factors is key to elucidating causes of VTE. The GATE study, with an ongoing enrollment of a large bi-racial study population, will provide the opportunity for a thorough evaluation of the complex etiology of VTE in Caucasians and African-Americans.


The authors thank Cathy Lally, Michele Beckman, Martha Forrester, Eileen Osinski, Pam Ruby, and Cathy Wood-Siverio (the GATE study staff); Penny Castellano, M.D and Kenneth Smith, M.D; and the CDC Hematologic Diseases Branch laboratory staff.