Familial Clustering of High Factor VIII Levels in Patients With Venous Thromboembolism
Abstract—High levels of factor VIII (FVIII) but not von Willebrand factor (vWF) are known to increase the risk for venous thromboembolism. Whether high FVIII levels originate from hereditary defects or from acquired conditions remains unanswered. The objective of our study was to investigate whether there is evidence for familial clustering of elevated FVIII levels in families in which ≥1 member has been affected by a thromboembolic event and had reproducibly high FVIII levels. We investigated FVIII levels in 361 patients with previous venous thromboembolism. FVIII levels were measured by a chromogenic assay; the cutoff value was defined as the 98th percentile of FVIII plasma levels of 266 blood donors. vWF levels were determined by an enzyme immunoassay. After exclusion of known causes of FVIII elevation, such as the acute thrombotic event itself; inflammation; malignancy; liver, renal, or vascular disease; surgery; or pregnancy, we included 17 patients with unexplained, reproducibly high FVIII levels. The investigation was also extended to these patients’ relatives. Multiple regressive analysis of blood donors and asymptomatic family members showed that the affiliation with a family in which 1 member suffered from venous thromboembolism and had reproducibly high FVIII levels is the second most important predictor for FVIII levels. Familial clustering was analyzed by the Houwing-Duistermaat familial aggregation test. After adjustment for the influence of age, sex, blood group, and vWF, FVIII levels were significantly (P=0.038) clustered within families. In conclusion, FVIII levels seem to be familially determined in families in which a member showed high FVIII levels after previous venous thromboembolism.
- Received April 10, 2000.
- Accepted November 3, 2000.
Familial clustering of venous thromboembolic events was observed at the beginning of the 20th century, but it was not until the 1960s that we began to gain insight into the hemostatic defects associated with familial thrombosis. The past 5 years brought a major breakthrough in the study of familial thrombosis. First, the investigation of activated protein C (APC) resistance and of the underlying factor V (FV) Leiden mutation was established1 2 ; second, a mutation in the prothrombin gene was identified.3 Both genetic defects represent the most frequent hereditary risk factors for venous thromboembolism. APC resistance is caused predominantly by the FV Leiden mutation. Other conditions, such as increased factor VIII (FVIII) activity, also promote APC resistance.4
High levels of FVIII but not von Willebrand factor (vWF) were recently determined to be a possible independent risk factor for deep-vein thrombosis. Patients with FVIII levels of >150 IU/dL had an adjusted odds ratio of 4.8 (95% CI, 2.3 to 10.0) compared with patients with FVIII levels <100 IU/dL.5 Hence, high FVIII concentrations represent a risk for thrombosis similar to those of deficiencies of inhibitors like proteins C and S and APC resistance.6 Whether high FVIII levels are heritable has not been reported, and no molecular defects in the FVIII gene associated with venous thromboembolism have been discovered so far.7
The objective of our study was to demonstrate by stochastic methods whether elevated FVIII levels are heritable. To accomplish this objective, FVIII levels were analyzed in families of patients who had suffered from thrombotic episodes and who were shown to have unexplained high FVIII levels.
Between 1990 and 1997, 361 consecutive patients with a past history of deep venous thrombosis or pulmonary embolism were referred to our outpatient department for thrombophilia screening. In all cases, the diagnosis had been confirmed by objective tests (venography, Doppler ultrasonography, ventilation-perfusion lung scans, or pulmonary angiography). Patients in whom FVIII levels exceeded the 98th percentile of levels determined in blood donors were considered to have high FVIII levels. Because the patients’ blood groups were not known initially, we chose a cutoff value that guaranteed a high specificity. This largely eliminated the impact the blood groups would have on patient selection. Patients were excluded from this group if an acute thromboembolic event had occurred <2 months before they entered our outpatient department; if an acute-phase reaction was detected (increase of C-reactive protein [CRP] or of α1-acid glycoprotein); if liver, renal, or inflammatory vascular disease or malignancy had been diagnosed; or if patients had undergone surgery within the previous 2 months. Further exclusion criteria were pregnancy or delivery within 2 months and treatment with any agents with a known effect on FVIII levels (eg, steroids, estrogens). We did not exclude patients on phenprocoumon, because FVIII levels during and after withdrawal of oral anticoagulation with phenprocoumon differ only very slightly.8 The shortest interval between withdrawal of oral anticoagulation and measurement of FVIII was 4 weeks. Venipuncture was repeated in all patients with unexplained high FVIII levels.
In addition, relatives of those patients who had high FVIII levels and were not finally excluded were recruited. These relatives were first-, second-, or third-degree relations. Relatives were not evaluated if ≥1 of the exclusion criteria mentioned above were met. Blood was taken only once (at home) and was brought to the laboratory immediately. Plasma was obtained by centrifugation at 2000g for 2×20 minutes; the samples were frozen within 4 hours after venipuncture. If any reason for a transient FVIII increase (eg, an acute-phase reaction) had been identified, the venipuncture was repeated 6 months later.
Three hundred healthy blood donors without any previous thrombotic episodes were recruited. Trained staff obtained a detailed clinical history from all control subjects, comparable to the history from the patients. The goals of this study were explained to both patients and control subjects in depth, and informed consents were obtained from all subjects.
CRP and α1-acid glycoprotein were measured by a nephelometric method (Dade Behring). ABO blood grouping was done by routine methods. Coagulation assays were performed as follows: Thromborel S was used as thromboplastin to determine the prothrombin time and Pathromtin SL as the activator to measure the activated partial thromboplastin time (aPTT) (both assays Dade Behring). FVIII levels were measured by a chromogenic assay (Dade Behring). FVIII concentrations were reported as a percentage; a FVIII concentration of 1000 IU/L corresponds to 100%. vWF was measured by an ELISA (Asserachrom, Roche Diagnostics).
Multiple regression analysis (MEDAS software, University of Würzburg) was used to analyze potentially confounding determinants of FVIII levels, such as vWF, blood group (O and non-O), age (continuous variable), and sex. FVIII and vWF were transformed logarithmically to approximate a normal distribution. The strength of the confounding determinants was calculated on the basis of data obtained from blood donors with normal values of aPTT, prothrombin time, CRP, and α1-acid glycoprotein. The residuals of multiple regression were used to calculate the familial impact on FVIII levels of the family members by a family aggregation test. Furthermore, multiple regression was used to assess the effect on FVIII levels of affiliation with a family in which 1 member suffered from venous thromboembolism and had reproducibly high FVIII levels. For that purpose, asymptomatic relatives of patients with prior thrombotic events were also included. After the effects of the potentially confounding determinants had been considered, the impact of family affiliation on FVIII levels was calculated.
Familial aggregation of FVIII was studied by use of a recently proposed test.9 The applicability of this novel test was recently shown.10 The correlation between pairs of relatives within a pedigree was adopted in the test statistic “Q.” The test for familial aggregation is positive when the calculated Q is significantly larger than the expected Q [E(Q)] value, considering the null hypothesis of no aggregation.
Q was calculated as described in the literature11 by use of correlation matrices and, thereby, consideration of the X-linked inheritance of FVIII. Briefly, for each kind of relationship within families, a correlation matrix M is established. M is calculated from basic matrices of conditional probability I (identity), T (transition matrix), and O (no relation): M=cI I + cT T + co O. For example, the constants cx for siblings are cI=1/4, cT=1/2, co=1/4, representing the probabilities of siblings having 2, 1, or no common genes. For X-linked inheritance, 4 basic matrices each are used. The correlation r between relatives is calculated by replacing the matrix T by the sex-dependent factor rT, I by 1, and O by 0: r=cI + cT × rT .
The effect of a possible genetic defect on the phenotype was taken into account by dominant and linear models.12 The dominant model reflects the case in which a single genetic defect is sufficient for the full expression of the phenotype. In contrast, the linear model considers the case in which genetic defects have an additive effect. q is defined as the probability not to be carrier of a genetic defect in the population investigated. For all tests, a value of P≤0.05 was considered statistically significant.
The present study was performed in accordance with the ethical standards laid down in the current version of the Declaration of Helsinki and was approved by the local medical ethics committee.
Of 300 healthy blood donors, 266 (152 men and 114 women; median age 40 years; age range 20 to 67 years; non-O blood group, n=160; O, n=106) revealed normal prothrombin times, normal aPTTs, and no acute-phase reaction. Of these, 260 (98%) had FVIII levels ≤200%. Of the 361 patients, 27 could not be included because of missing data on FVIII or CRP. Applying the exclusion criteria, 61 patients (16.9% of all patients) with high FVIII concentrations (ie, FVIII >200%) remained. Of these 61 patients, 36 were not included because they could not be reached, lived >120 km away from our hospital, did not agree to have the investigation extended to their families, developed chronic inflammatory or malignant diseases, or died in the interim. Finally, 17 of the remaining 25 patients were included because high FVIII levels were confirmed in those patients by the second blood sample (the maximum difference between the first and second samples that was considered acceptable was 30%).
The median age at the time of the first thrombotic event was 41 years (range, 17 to 67 years). The median time between the last thromboembolic event and venipuncture for this study was 7 months (range, 2 to 177 months). Clinical diagnoses of the patients with a history of thrombosis were as follows: 20 occurrences of deep-venous thrombosis of the lower extremity or pelvis and 2 events of pulmonary embolism; in 2 cases, both events occurred. There was 1 case of a more uncommon site, ie, an inferior vena cava thrombosis. Seven patients had experienced >1 thrombotic event. Deep venous thrombosis was considered idiopathic (absence of trigger factors, ie, immobilization, surgery, pregnancy, trauma, or oral contraceptives) in 14 of 25 cases (8 of 17 cases if only the first event was considered). Of the family members, 2 subjects had a personal history of venous thrombosis. A son of 1 patient (FVIII=152%) and a sister of another patient (FVIII=267%) experienced deep venous thrombosis at the ages of 18 and 68 years, respectively. In total, 94 patients and their relatives were included in the family aggregation study.
By multiple regression of the data of blood donors, vWF could be identified as the most powerful confounding variable (β-weight=0.685). Other important variables, such as blood group, age, and sex (β-weight <0.2), did affect FVIII levels far less than vWF levels (Table 1⇓). Multiple regression analysis of data from blood donors and asymptomatic family members revealed that belonging to a family in which 1 member suffered from venous thromboembolism and had reproducibly high FVIII levels was the second most important predictor for high FVIII levels (Table 2⇓). The impact of belonging to families of probands with thrombosis and persistently high FVIII levels (β-weight=0.091) had the same magnitude as that of the blood group (β-weight=0.104) when only 1 randomly chosen first-degree relative of each family was considered. We conclude that FVIII levels have a common determinant in families with members who had thrombotic events and reproducibly high FVIII levels; this suggests heritability of FVIII level in these families.
For the family aggregation test, different values of q were used, because the mode of inheritance is not known. Considering the families in which the high FVIII level of the affected patient could be confirmed by the second blood sample, familial aggregation of FVIII levels could be demonstrated (Table 3⇓). Hence, there was consistent evidence for familial aggregation in either the linear or the dominant model with different q values.
Familial clustering could not be shown by the family aggregation test, however, if only the 8 families were considered in which the high FVIII level of the affected patient could not be confirmed.
High FVIII levels are now known to be a major risk factor for venous thromboembolism. It is still open whether high FVIII levels after previous thromboembolism are acquired or hereditary. Genetic defects that promote high FVIII levels have not been reported so far. Mansvelt and coworkers7 could not detect any sequence polymorphism in the promoter region and the 3′ terminus of the FVIII gene. Our study aimed to demonstrate by stochastic methods whether FVIII levels are familially clustered in families in which ≥1 member has been affected by a thromboembolic event and has shown high FVIII levels.
Multiple regression analysis of data from blood donors and asymptomatic members of affected families demonstrated that family affiliation was a major determinant for FVIII levels. This was the case for families in which high FVIII levels in individual patients were confirmed by a second blood sample. Hence, multiple regression suggests heritability of FVIII levels in affected families, a fact expected to be evident as well when the familial aggregation test is applied.
To test familial aggregation, a suitable test for pedigrees that are selected because of the response of a subject would be based on conditional likelihood. Unfortunately, such mathematical tests do not exist. For our study, we used the novel familial aggregation test of Houwing-Duistermaat et al.9 According to the authors, their test should be applied for randomly chosen pedigrees to yield generally applicable evidence for a whole population. Family aggregation studies based on randomly chosen pedigrees, however, are hardly feasible. Kamphuisen et al10 studied selected families of hemophilia patients and countered this problem by considering a certain genetic defect as determined by DNA analysis. In our opinion, “carriership” does not appear to be the appropriate way to bridge the gap between feasibility of familial aggregation studies and a general statement on the heritability of FVIII levels. The test is probably valid for the population studied, but the power of this test cannot be generalized by correcting for genetic defects. The selected population may not be assumed to be representative of the normal population even if carriership is considered. Environmental and genetic factors may have different effects on FVIII levels. FVIII levels can be, but do not always have to be, subject to heredity. For our purposes, the test may be suitable, because we are looking only at familial clustering of a certain feature in families with a certain clinical entity.
Also, the use of this familial aggregation test is justified in studies of families in which a weak genetic effect, ie, a slight impact on FVIII levels, can be expected. This seems to hold true for our study group, in which moderately elevated FVIII levels are frequently observed in patients with a history of thrombotic events.6
In contrast to the study by Kamphuisen et al, the power of confounding variables was derived from a different collective. Therefore, the bias of additional interrelations between the family members was avoided.
In families with a patient who had had a thrombotic episode and an associated, reproducibly high FVIII level, FVIII levels were shown to be also familially determined. The familial aggregation test emphasizes the findings obtained by multiple regression. The lack of familial clustering in the families in which the affected patient had an only temporarily high FVIII level might strengthen the findings of our study. However, these calculations have to be interpreted very cautiously, because the number of families might not be sufficient.
It is conceivable that a genetic polymorphism of the FVIII gene is linked with high FVIII levels in patients with prior thrombosis. Research will also have to focus on other relevant genes that may lead to high FVIII levels, however. For instance, a chaperone molecule has been detected in association with FVIII but not with FV. Chaperones either facilitate or prohibit protein folding and transport out of the endoplasmic reticulum. Hence, FVIII is 10-fold less efficiently secreted than the homologous FV.13 14 15 Moreover, secretion can be influenced by mutagenesis in the chaperone binding region of the FVIII A1 domain.16
In conclusion, a familial component of FVIII levels can be detected in families in which patients showed unexplained high FVIII levels after previous thromboembolism. The underlying mechanisms and the contribution to an increased risk for venous thromboembolism remain to be investigated.
We express our thanks to Martina Misoph, whose critique and very helpful comments are highly appreciated. We are indebted to the Bavarian Red Cross for the opportunity to collect samples from healthy persons.
Dahlbäck B, Carlsson M, Svensson PJ. Familial thrombophilia due to a previously unrecognized mechanism characterized by poor anticoagulant response to activated protein C: prediction of a cofactor to activated protein C. Proc Natl Acad Sci U S A. 1993;90:1004–1008.
Poort SR, Rosendaal FR, Reitsma PH, Bertina RM. A common genetic variation in the 3′-untranslated region of the prothrombin gene is associated with elevated plasma prothrombin levels and an increase in venous thrombosis. Blood. 1996;88:3698–3703.
Elandt-Johnson RC. Probability Models and Statistical Methods in Genetics. New York, NY: John Wiley; 1971.
Pittman DD, Tomkinson KN, Kaufman RJ. Post-translational requirements for functional factor V and factor VIII secretion in mammalian cells. J Biol Chem. 1994;269:17329–17337.
Marquette KA, Pittman DD, Kaufman RJ. A 110-amino acid region within the A1-domain of coagulation factor VIII inhibits secretion from mammalian cells. J Biol Chem. 1995;270:10297–10303.
Swaroop M, Kaufman RJ. Mutation of Phe309Ser, a putative BIP binding site, enhances secretion of coagulation factor VIII. Blood. 1996;88(suppl):441a.