F.R. Rosendaal, Department of Clinical Epidemiology, Leiden University Medical Center, C9-P, PO Box 9600, NL-2300 RC Leiden, the Netherlands. Tel.: +31 71 526 4037; fax: +31 71 526 6994; e-mail: email@example.com
Summary. Deep venous thrombosis is a multicausal disease, i.e. more than one risk factor needs to be present to cause the disease. Oral contraceptive use increases the risk of venous thrombosis but since not all women using oral contraceptives develop thrombosis, the presence of additional risk factors in patients is likely. The aim of this study was to assess the joint effect of oral contraceptive use and the levels of procoagulant factors (F)(FII, FV, FVII, FVIII, FIX, FX, FXI, FXII, FXIII and fibrinogen). Data of premenopausal women were re-analyzed in the Leiden Thrombophilia Study. The highest relative risks were observed for the combination of oral contraceptive use and high levels (>90th percentile) of FII (Odds Ratio [OR]OC+FII 10.1; 95% confidence interval [CI] 3.5–29.0), FV (OROC+FV 12.6; 95% CI 3.8–41.5), and FXI (OROC+FXI 11.9; 95% CI 3.6–39.2) and low levels (< 10th percentile) of FXII (OROC+FXII 12.3; 95% CI 2.4–63.0). No interaction was observed between oral contraceptive use and high levels of the other coagulation factors, i.e. the joint effect of these risk factors did not exceed the sum of the separate effects. The results of this study indicate that the risk for the joint effects of oral contraceptive use and coagulation factor levels are minor compared with the joint effect of oral contraceptive use and the FV Leiden mutation (RR > 30).
The use of oral contraceptives is associated with a 3–6-fold increased risk of venous thrombosis (VT) [1–4]. Risk factors for VT can be divided into acquired and genetic risk factors. In addition to oral contraceptive (OC) use, examples of acquired risk factors include surgery, immobilization, trauma, pregnancy, puerperium, and lupus anticoagulant. Genetic factors that predispose to VT are protein C , protein S  and antithrombin deficiency , factor (F)V Leiden  and the prothrombin 20210A mutation .
The multicausal character of deep venous thrombosis (DVT) has been illustrated previously [10–12]. In order to trigger a venous thrombotic event, several risk factors need to be present simultaneously within one individual. Not all women using OCs develop VT, which implies that women using OCs who develop VT are likely to have an additional risk factor. This became clear in women who used OCs and carried the FV Leiden mutation. These women have a > 30-fold increased risk of VT, which is considerably higher than would be expected based on the effects of FV Leiden and OC use separately [1,13,14]. A synergistic effect, but less striking, was also reported between OC use and the prothrombin 20210A mutation [13,14].
More recently it was shown that, besides high levels of prothrombin, high levels of coagulation FVIII, FIX and FXI were associated with an increased risk of VT [15–17]. High levels of FX were associated with a mildly increased risk of VT mainly in women not using OCs . The molecular basis of these abnormalities is still unknown. The prevalence in the general population of high levels of these coagulation factors depends on the cutoff points that are applied. Levels of FVIII, FIX and FXI in the upper 10% of the distribution (by definition present in 10% of the population) are associated with a 2–3-fold increased risk of VT [15–17].
Risk factors that are common in the general population enable us to study gene–environment interactions. So far, no information is available on the joint effect of OC use and high levels of the procoagulant factors. The aim of the present study was therefore to assess the joint effect of OC use and levels of the procoagulant FII (prothrombin), FV, FVII, FVIII, FIX, FX, FXI, FXII, FXIII and fibrinogen.
This study was performed in the Leiden Thrombophilia Study (LETS), a large population based case–control study. The design of this study was described earlier [19,20]. In brief, we included 474 consecutive patients with a first episode of DVT from three anticoagulation clinics in the Netherlands. As control subjects, 474 friends/partners of patients were included. Blood was collected at least 6 months after the thrombotic event using Sarstedt Monovette® tubes, with 0.1 volume of 0.106 mol L−1 trisodium citrate. For the current study we selected only premenopausal women within the age of 15–49 years, who, at the time of thrombosis (or a similar date for the control subjects) (index date), were not pregnant, were not within 30 days postpartum, did not have a recent miscarriage and did not use depot contraceptives. Women still using oral anticoagulation at the time of the venepuncture (n = 6) were also excluded from the current analysis since four of the procoagulant factors studied (FII, FVII, FIX, and FX) are vitamin K-dependent factors. This selection resulted in the inclusion of 149 patients and 169 control subjects. OC use was determined at the time of the index date. For one patient no information on the use of OCs at the time of the thrombotic event was available.
The measurements of the levels of the coagulation factors were described in detail in earlier studies performed in LETS [9,15–18,21–24] In brief, prothrombin activity was measured by a chromogenic assay using Echis carinatus venom as activator . FV:Ag was measured by an in-house-developed sandwich-type enzyme-linked immunosorbent assay (ELISA) with two different monoclonal antibodies, both with a high affinity for the light chain of (activated) FV . The FVII, FVIII and FXII activities were measured by one-stage coagulation assays [15,22,23]. FIX and FX antigen levels were measured by sandwich ELISAs using commercial polyclonal antibodies (Dako A/S, Glostrup, Denmark) [16,18], and FXI antigen levels by an ELISA using a monoclonal anti-FXI antibody as capture antibody and polyclonal anti-FXI as tagging antibody . FXIII A and B subunit levels were measured by sandwich-ELISA [24,25]. The fibrinogen concentration was measured according to the method of Clauss using Dade® thrombin (Baxter, Miami, USA) . All coagulation factors were expressed in U dL−1, where 1 U is the amount of coagulation factor present in 1 mL pooled normal plasma. Fibrinogen levels were expressed in g L−1.
To assess the joint effect of high coagulation factor levels and OC use, we calculated Odds ratios (OR) with 95% confidence intervals (CI). For all analyses, the levels of the coagulation factors were dichotomized using the 90th percentile (P90) measured in the control group (n = 169) as a cutoff point (highest 10%). For FXII, where low levels have been suggested as a risk factor for VT , we dichotomized at the 10th percentile (P10) measured in the control group (lowest 10%).
Firstly, we estimated the crude OR for OC use and for high levels of FII (prothrombin), FV, FVII, FVIII, FIX, FX, FXI, FXIII and fibrinogen (> P90) and low levels of FXII (< P10), respectively. Secondly, the joint effects of OC use and high levels of the coagulation factors (low levels for FXII) were calculated.
The joint effects of OC use and high (or for FXII low) levels of coagulation factors were also estimated using different cutoff values (95th and 75th percentile; for FXII 5th and 25th percentile).
Since OC use influences the levels of many coagulation factors, one would like to determine coagulation factor levels during OC use for those who developed VT, i.e. at the time of thrombosis. Blood collection took place more than 6 months after the venous thrombotic event, and many patients who used OCs at the time of thrombosis had stopped using them on doctor's advice. Using the mean effect of OC use on coagulation factor levels (measured in the control subjects), we estimated the coagulation factor levels in patients prior to the index date, i.e. at the time of the thrombosis. For example, OC use induces a mean increase of FIX levels of 22.6 U dL−1, therefore patients who used OCs at the time of thrombosis but stopped using them at the time of the venepuncture had an approximate FIX level at the time of thrombosis of the measured level at the time of the venepuncture + 22.6.
The mean age of the 149 female patients with a first episode of DVT was 35.4 years (range 16.2–49.4), and for the 169 female control subjects it was 34.7 years (range 15.1–48.9).
Coagulation factor levels
The cutoff points for all coagulation factors measured in the control group are shown in the table. After dichotomization of the coagulation factor levels at the 90th percentile (10th percentile for FXII) we calculated the crude relative risks associated with coagulation factor levels, i.e. coagulation factor levels above the 90th percentile (for FXII below the 10th percentile) (Table 1).
Table 1. Thrombotic risk associated with OC use and high (for factor XII low) coagulation factors levels
All Odds ratios are adjusted for age. *For all coagulation factors the thrombotic risk of individuals with levels above the 90th percentile measured in the control subjects [> P90] (for FXII, levels below the 10th percentile measured in the control subjects [< P10]) were compared with individuals with levels ≤P90 (for FXII ≥P10) see Methods. †Oral contraceptive use at the time of the thrombotic event (or index date for controls). ‡All risks are relative to non-users of OCs who do not have high (or for FXII low) coagulation factor levels.
After dichotomization, the levels of several coagulation factors were associated with an increased risk of VT. In the crude analysis, high levels of FII, FV, FVIII and FIX were associated with a 2-fold increased risk of VT. High levels of FX and XI were associated with a mildly increased risk of VT in this study population with a 1.8- and 1.5-fold increased risk, respectively. No effect on the risk of VT was seen for high levels of FVII, FXIII and fibrinogen or low levels of FXII. Adjustment for age did not affect the risk estimates (Table 1).
Oral contraceptive use
Oral contraceptive use at the index date was associated with a 4-fold increased risk of VT (OR 3.8; 95% CI 2.4–6.0). Adjustment for age increased the risk of VT associated with OC use to a nearly 6-fold increased risk (OR 5.6; 95% CI 3.2–9.7).
Joint effects of OC use and coagulation factor levels
In Table 1 and Fig. 1 the joint effects of OC use and the coagulation factor levels are shown. The highest relative risks were observed for the combination of OC use and high levels of FII (OROC+FII 10.1; 95% CI 3.5–29.0), FV (OROC+FV 12.6; 95% CI 3.8–41.5), and FXI (OROC+FXI 11.9; 95% CI 3.6–39.2) and low levels of FXII (OROC+FXII 12.3; 95% CI 2.4–63.0). No interaction was observed between OC use and high levels of coagulation FVII, FVIII, FIX, FX, FXIII and fibrinogen, i.e. the joint effect of these risk factors did not exceed the sum of the separate effects (Table 1; Fig. 1).
To assess whether the risk estimates shown in the table are consistent, we dichotomized the coagulation factor levels at several cutoff points and estimated the combined effect of high (or for FXII low) coagulation factor levels and OC use (sensitivity analysis). Coagulation factor levels were also dichotomized at the 95th percentile (> P95 vs. ≤ P95; for FXII < P5 vs. ≥ P5). Furthermore a comparison was made between coagulation factor levels in the highest quartile compared with the lowest quartile (for FXII < P25 vs. > P75) and between coagulation factor levels above 90th percentile compared with coagulation factor levels below 50th percentile (for FXII < P10 vs. > P50). Variations in risk estimates occurred due to the small size of some of the strata. The range of risk estimates for the various analyses was: FII (OROC+FII 8.4–12.1), FV (OROC+FV 7.1–13.1), FVII (OROC+FVII 6.0–10.3), FVIII (OROC+FVIII 6.8–18.2), FIX (OROC+FIX 6.9–17.0), FX (OROC+FX 5.8–13.5), FXI (OROC+FXI 9.4–13.8), FXII (OROC+FXII 7.1–14.5), FXIII A subunit (OROC+FXIIIA 5.8–9.3), FXIII B subunit (OROC+FXIIIB 4.3–15.7), and fibrinogen (OROC+FIB 2.2–7.0).
Recalculating the relative risks for the combination of OC use and high coagulation factor levels (i.e. above the 90th percentile or for FXII below the 10th percentile) using the estimated coagulation factor levels at the time of thrombosis, all risks were within or very close to the range provided by the sensitivity analysis. The only exceptions were the thrombotic risk for the combination of OC use and high levels of FII, which was slightly higher (OR 12.4; 95% CI 4.7–32.3), and for the combination of OC use and low levels of FXII, which was lower (OR 3.7; 95% CI 0.2–63.5).
In this study we investigated the joint effects of OC use and the levels of coagulation FII, FV, FVII, FVIII, FIX, FX, FXI, FXII, FXIII, and fibrinogen. To our knowledge, the joint effects of OC use and high levels of the various coagulation factors have not been reported previously.
High levels of coagulation FII, FVIII, FIX, FX and FXI were associated with an increased risk of VT as we described previously [9,15–18]. We found in this subgroup of premenopausal women an increased risk associated with high levels of FV (OR 2.2; 95% CI 1.1–4.3). In the total study population of the LETS consisting of men and women below 70 years no significant increase in the risk of thrombosis associated with high levels of FV was found, i.e. individuals with FV levels > 150 U dL−1 had only a 1.3-fold (95% CI 0.9–1.8) increased risk of VT compared with individuals with FV levels < 110 U dL−1, and no dose–response relationship was observed . This difference in thrombotic risk estimated for high levels of FV may be due to chance since the confidence intervals of both risk estimates show a fairly large overlap. Alternatively, high levels of FV may be a risk factor predominantly in premenopausal women, as was also found for high levels of FX . In this selection of women, using the 90th percentile as a cutoff point, high levels of fibrinogen were not associated with an increased risk of VT. Earlier analysis of the LETS, i.e. in the first 199 participants of this study, showed a mild increase of thrombotic risk associated with fibrinogen levels > 4.0 g L−1 when compared with levels < 3.0 g L−1. High levels of FVII, and FXIII A and B subunits, or low levels of FXII were not associated with an increased risk of VT as reported previously in the total population of the LETS study [22–24].
The joint effect on the risk of VT was highest for the combination of OC use and high levels of FII, FV, FXI and for OC use and low levels of FXII (Fig. 1). For women who use OCs and who have high levels of FII, FV, FXI or low levels of FXII, a mild additional effect on the risk of VT was found. OC use combined with abnormal levels of these coagulation factors led to an increased risk beyond what was expected based on the separate risks. For FVIII levels, as reported earlier for this patient and control group, risks were merely additive with OC use .
The sensitivity analysis with various cutoff points showed that variations in risk estimates of the combined effect between high (or for FXII low) coagulation factor levels and OC use occurred. Only the thrombotic risks associated with the combined effect of OC use and high levels of coagulation FII and FXI seemed consistently high. Recalculation of the thrombotic risk associated with the joint effect of high coagulation factor levels and OC use, using estimated coagulation factor levels at the time of the thrombosis, led to risk estimates that fell within the ranges described in the sensitivity analysis. A disadvantage of this recalculation, however, is that this does not take into account the presence of so-called hyper-responders to OCs . Nevertheless, the result of this analysis shows our risk estimates are fairly robust.
These findings are in agreement with the multicausal interpretation of DVT, i.e. in some instances the combined presence of risk factors leads to elevated risks. However, the thrombotic risks associated with the combinations we reported here were not very high. The risk of VT associated with the combination of OC use and the FV Leiden mutation is approximately 30-fold increased compared with non-carrier non-OC users [1,13,14]. This risk estimate is considerably higher than the relative risks presented here for the combination of OC use and high (or for FXII low) levels of the procoagulant factors of the coagulation system.
The results of this study indicate that the joint effects of OC use and coagulation factor levels are minor compared with the joint effect of OC use and the FV Leiden mutation. Therefore the interaction between OC use and high coagulation factor levels will currently have little relevance for clinical care.
In this study we did not assess differences between second- and third-generation OCs since the paucity of individuals did not allow these analysis. In a crossover study by Middeldorp et al., it was shown that the effects of second- and third-generation OCs on the hemostatic system differ . In that study, compared with levonorgestrel, OCs containing desogestrel were associated with an increase of coagulation FII and FVII and a decrease of coagulation FV. Therefore one must keep in mind that the synergistic effect may be different between types of OCs. Since this is the first study reporting interaction between OC use and high levels of several coagulation factors, more studies are needed to confirm these findings.
We thank Dr T. Koster for collecting blood samples of patients and control subjects, and Ank Schreijer and Ingeborg de Jonge for performing datamanagement. We also thank all participants of the Leiden Thrombophilia Study for their cooperation. The LETS study was financially supported by the Netherlands Heart Foundation (grant no. 89.063).