High risk of pregnancy-related venous thromboembolism in women with multiple thrombophilic defects
Dr Nienke Folkeringa, Division of Haemostasis, Thrombosis and Rheology, Department of Haematology, University Medical Centre Groningen, Hanzeplein 1, 9713 GZ Groningen, The Netherlands. E-mail: email@example.com
Pregnancy is associated with an increased risk of venous thromboembolism, which probably varies according to the presence of single or multiple thrombophilic defects. This retrospective family cohort study assessed the risk of venous thromboembolism during pregnancy and puerperium, and the contribution of concomitant thrombophilic defects in families with hereditary antithrombin, protein C or protein S deficiencies. Probands were excluded. Of 222 female relatives, 101 were deficient and 121 non-deficient. Annual incidences of venous thromboembolism were 1·76% in deficient women versus 0·19% in non-deficient women [adjusted relative risk (RR) 11·9; 95% confidence interval (CI), 3·9–36·2]. Other single and multiple thrombophilic defects increased the risk in deficient women from 1·55% to 2·14% and 2·92%, and in non-deficient women from 0·16% to 0·09% and 0·54% respectively. Deficient women were at lower risk (1·37%; 0·80–2·19) than deficient women that had never been pregnant (2·96%; 1·53–5·18); RR 0·5 (0·2–0·99). This difference was due to the predominance of events related to oral contraceptives in deficient women that had never been pregnant (75%), while 71% of events in deficient women that had had at least one pregnancy were pregnancy-related. In conclusion, women with hereditary deficiencies of antithrombin, protein C or protein S are at high risk of pregnancy-related venous thromboembolism. This risk is increased by multiple additional thrombophilic defects.
Venous thromboembolism in healthy women during pregnancy or puerperium is rare, but it is still an important cause of maternal morbidity and mortality in the Western world (Rochat et al, 1988). Its incidence in the general female population ranges from 0·5 to 1 in 1000 deliveries (Toglia & Weg, 1996; Heit et al, 2005). The risk of venous thromboembolism in pregnant women is five times higher than in non-pregnant woman of the same age. It predominantly occurs in the third trimester and puerperium (Ray & Chan, 1999).
It is plausible that the increased risk of venous thromboembolism is related to venous stasis and a state of hypercoagulability, as a result of physiological changes in blood coagulation and fibrinolysis during pregnancy (Stirling et al, 1984). Consequently, a higher risk has been found in pregnant women with pre-existent thrombophilic defects (Hellgren et al, 1995; Sanson et al, 1999; Robertson et al, 2006). Of these, hereditary deficiencies of antithrombin, protein C and protein S are strong risk factors for venous thromboembolism (Conard et al, 1990; De Stefano et al, 1994; Friederich et al, 1996; Pabinger & Schneider, 1996). Recently, we demonstrated that the risk in deficient subjects depends on the frequently observed concomitance of other, milder thrombophilic defects (Brouwer et al, 2006). It is likely that multiple deficiencies or defects, in addition to physiological changes, will particularly enhance the risk of venous thromboembolism during pregnancy and puerperium. To test this hypothesis, a sufficient number of deficient women is required to assess interactions between rare deficiencies and more prevalent thrombophilic defects. Therefore, we analysed data from a retrospective family cohort study, which was designed to assess the contribution of currently known thrombophilic defects to the absolute risk of venous thromboembolism associated with hereditary deficiencies of antithrombin, protein C or protein S.
Materials and methods
The original study comprised three cohorts of families with hereditary deficiencies of antithrombin, protein C and protein S, respectively (Brouwer et al, 2006). Probands were consecutive patients with documented venous thromboembolism in whom one of these deficiencies was demonstrated. They were referred with clinically suspected venous thromboembolism to the thrombosis out-patient clinic of our hospital over a period of 12 years. First-degree relatives that were 15 years of age or older were identified by pedigree analysis. As the number of antithrombin-deficient probands was small, second degree relatives from a deficient parent were also identified. Written informed consent was obtained from all participants. Detailed information about episodes of venous thromboembolism, exposure to external risk factors for thrombosis, obstetric history and anticoagulant treatment was collected by the physicians of our out-patient clinic using a validated questionnaire and reviewing medical records (Frezzato et al, 1996). Blood samples for thrombophila testing were taken after clinical data had been collected. In addition to the index deficiency, tests included prothrombin G20210A, factor V Leiden, increased levels of factors VIII, IX and XI, lupus anticoagulant and hyperhomocysteinaemia. For convenience, additional abnormalities are termed hereafter as ‘thrombophilic defects’. The study was approved by the institutional review board of our hospital. The present study analysed the overall risk of venous thromboembolism, particularly the risk of pregnancy-related venous thromboembolism in female relatives of reproductive age.
Venous thromboembolism was considered established if proximal deep vein thrombosis was confirmed by compression ultrasound or venography (not in pregnancy), and pulmonary embolism by ventilation/perfusion lung scanning, spiral computed tomography (CT) scanning or pulmonary angiography (not in pregnancy), or when the patient had received full dose unfractionated heparin and a vitamin K antagonist for at least 3 months without objective testing at a time when current techniques were not available. Venous thromboembolism was classified as provoked if it had occurred within 3 months of exposure to one or more external risk factors, including surgery, trauma, immobilisation for more than 7 d, use of oral contraceptives, pregnancy, puerperium and malignancy. In the absence of these risk factors, it was considered spontaneous.
Antithrombin activity (Coatest; Chromogenix, Mölndal, Sweden) and protein C activity (Berichrom Protein C; Dade Behring, Marburg, Germany) were measured by chromogenic substrate assays. Protein C and total protein S and antigen levels were measured by enzyme-linked immunosorbent assay (ELISA) (DAKO, Glostrup, Denmark). Normal ranges (mean ± 2 SD) were determined in 393 healthy blood donors, who had no (family) history of venous thromboembolism and were not pregnant, and had not used oral contraceptives for at least three months. Antithrombin deficiency was defined by levels of antithrombin activity below the lower limit of its normal range (<74 IU/dl), protein C deficiency type I and type II were defined by reduced levels of either protein C antigen (<63 IU/dl) and/or activity (<64 IU/dl). Protein S deficiency type I was defined by lowered total protein S antigen levels (<67 IU/dl). Deficiencies were considered inherited if they were confirmed by measuring a second sample that was collected 3 months later and were demonstrated in at least two family members. Relatives with acquired conditions were excluded. If there was a discrepancy between the results of the two tests, a third sample was tested. A deficiency was considered acquired through use of oral contraceptives or pregnancy, unless it was confirmed at least 3 months after withdrawal of oral contraceptives or delivery, respectively. Relatives with abnormal liver function tests were only evaluable if deficiencies were established at repeated measurements after recovery.
Factor V Leiden and the prothrombin G20210A mutation were demonstrated by polymerase chain reaction (Bertina et al, 1994; Danneberg et al, 1998). Coagulant activity of Factors VIII (FVIII:C), IX (FIX:C) and XI (FXI:C) were measured by one-stage clotting assays (Amelung GmbH, Lemgo, Germany) and were considered increased at levels above 150 IU/dl (van Hylckama Vlieg et al, 2000; Meijers et al, 2000; Bank et al, 2005). Fasting and post-methionine-loading levels of homocysteine were measured by high performance liquid chromatography (Den Heijer et al, 1995). Hyperhomocysteinaemia was defined as a fasting level above18·5 μmol/l and/or a post-loading level above 58·8 μmol/l, as described in the Dutch population (Den Heijer et al, 1996).
In probands and relatives with venous thromboembolism, blood samples were collected at least 3 months after this event had occurred. If, at that time, subjects that were still receiving acenocoumarol, a short acting vitamin K antagonist, samples were taken after this therapy had been interrupted for at least 2 weeks, meanwhile nadroparin was given subcutaneously.
We first compared the overall absolute risk of the first episode of venous thromboembolism in deficient and non-deficient female relatives at reproductive age (15–45 years). Probands were excluded from analysis to avoid bias. Annual incidences were calculated by dividing the number of symptomatic relatives by the total number of observation years. Observation time was defined as the fertile period (from age 15 until 45 years), or from age 15 years until the first venous thromboembolic episode, or until the end of study. Second, we compared the risk of venous thromboembolism in deficient and non-deficient relatives that had had at least one pregnancy versus those who had never been pregnant. Third, we compared the incidence of pregnancy-related venous thromboembolism in deficient and non-deficient relatives. In these analyses, pregnancies after a prior episode of venous thromboembolism were not counted and women were considered as never having been pregnant when their first pregnancy occurred after a prior episode of venous thromboembolism, because the outcome of these pregnancies might have been influenced by thromboprophylaxis. Pregnancies that ended in fetal loss or termination, or were ectopic, were not excluded from analysis.
The relative risk of venous thromboembolism was calculated from annual incidences. The 95% confidence intervals (95% CI) and P-values were computed using small sample statistics based on mid-P-value functions from the binomial probability model (Rothman & Greenland, 1998). Random effects logistic regression was used to adjust for clustering of women in families. Continuous variables were expressed as mean values and standard deviations, and categorical data as counts and percentages. Differences between groups were evaluated by the Student's t-test or Mann–Whitney U-test, depending on the normality of data for continuous data, and by Fisher exact test for categorical data. A two-tailed P-value of less than 0·05 was considered to indicate statistical significance. Statistical analyses were performed using SAS software, version 9.1 (SAS-Institute inc., Cary, NC, USA).
Our family cohort consisted of 12 probands with antithrombin deficiency and their 185 relatives, 40 probands with protein C deficiency and their 277 relatives, and 39 probands with protein S deficiency and their 262 relatives. Of all 326 female relatives, 234 were alive and aged 15 years or older. Consent was refused or could not be obtained for geographic reasons in 12 of these relatives. The remaining 222 women were eligible, of whom 101 were deficient and 121 non-deficient. Their characteristics are summarised in Table I. Of 222 relatives, 133 had been pregnant of whom 54 were deficient (162 pregnancies) and 79 were non-deficient (243 pregnancies). Mean parity was comparable in deficient and non-deficient women.
Table I. Characteristics of deficient and non-deficient women.
|Age at enrolment, mean (SD), years||47 (19)||49 (18)||0·26|
|Observation period, mean (SD), years||16 (10)||21 (11)||<0·01|
|Women with at least one pregnancy, n (%)||54 (53)||79 (65)||0·08|
|Pregnancies prior to first venous thromboembolism, n||162||245||0·06|
|Parity, mean (SD)||1·6 (2)||2·0 (2)||0·06|
|Use of oral contraceptives at any time, n (%)||56 (55)||78 (64)||0·21|
|Women with venous thromboembolism, n (%)||29 (29)||5 (4)||<0·01|
|Age at time of venous thromboembolism, mean (SD), years||27 (6)||27 (9)||0·75|
| Pregnancy, n (%)||4*(14)||0||1·0|
| First trimester, n||1||0|| |
| Second trimester, n||1||0|| |
| Third trimester, n||1||0|| |
| Puerperium, n (%)||8 (27)||1 (20)||1·0|
| Oral contraceptives, n (%)||11 (38)||3 (60)||0·63|
| Surgery, trauma or immobility, n (%)||3 (10)||1 (20)||0·49|
|Spontaneous, n (%)||3 (10)||0||1·0|
|Protein levels, mean (SD), IU/dl|
| Antithrombin activity||52 (8)||101 (16)|| |
| Protein C activity||52 (19)||48 (12)|| |
| Protein S antigen||48 (13)||104 (23)|| |
Twenty-nine of 101 deficient women (29%) and five of 121 non-deficient women (4%) had experienced venous thromboembolism before 45 years of age (Table I). Mean age at time of the first episode of venous thromboembolism was 27 years in each group. Of 29 first episodes of venous thromboembolism in deficient women, 12 (41%) were related to pregnancy and 11 (38%) to use of oral contraceptives, compared to one of five and three of five episodes in non-deficient women respectively. Eight of 12 pregnancy-related episodes of venous thromboembolism in deficient women occurred during puerperium, as did the only episode of venous thromboembolism in non-deficient relatives.
Concomitant thrombophilic defects were demonstrated in 57% of deficient and in 70% of non-deficient relatives (Table II). The prevalence's of these defects were higher in both deficient and non-deficient women than reported in the normal population, except for high levels of factor IX and factor XI in deficient relatives. Lupus anticoagulant was not found. Differences between deficient and non-deficient women were not statistically significant.
Table II. Prevalence of concomitant thrombophilic defects in deficient and non-deficient women.
|Prothrombin G20210A, % (n tested)||6 (89)||9 (109)||0·42|
|Factor V Leiden, % (n tested)||13 (90)||15 (110)||0·84|
|Factor VIII >150 IU/dl, % (n tested)||36 (84)||42 (102)||0·45|
|Factor IX >150 IU/dl, % (n tested)||9 (68)||13 (104)||0·47|
|Factor XI >150 IU/dl, % (n tested)||9 (88)||16 (105)||0·20|
|Hyperhomocysteinaemia, % (n tested)||16 (83)||19 (96)||0·69|
|No concomitant defect, % (n tested)||43 (86)||30 (101)|| |
|1 concomitant defect, % (n tested)||38 (86)||49 (101)||0·18|
|>1 concomitant defect, % (n tested)||19 (86)||21 (101)|| |
The overall annual incidences of venous thromboembolism were 1·76 per 100 person-years (95% CI, 1·18–2·53) in deficient women and 0·19 per 100 person-years (95% CI, 0·06–0·45) in non-deficient women; the relative risk adjusted for clustering of women in families was 11·9 (95% CI, 3·9–36·2) (Table III). In deficient women, these were 1·56 per 100 person-years (95% CI, 0·67–3·07) in the absence of concomitant thrombophilic defects, compared to 2·14 (95% CI, 1·14–3·66) and 2·92 per 100 person-years (95% CI, 1·17–6·01) when one, and more than one concomitant thrombophilic defects were present, respectively. Annual incidences were 0·16 (95% CI, 0·004–0·87), 0·09 (95% CI, 0·002–0·53) and 0·54 (95% CI, 0·11–1·58) per 100 person-years in non-deficient women with none, one, and more than one other defect respectively.
Table III. Absolute risk of venous thromboembolism in deficient and non-deficient fertile women.
|Women with event, n||29||5|
|Observation period, years||1645||2598|
|Annual incidence, % (95% CI)||1·76 (1·18–2·53)||0·19 (0·06–0·45)|
|Adjusted relative risk (95% CI)*||11·9 (3·9–36·2)|
|No concomitant thrombophilic defect, n||37||31|
|Women with event, n||8||1|
|Observation period, years||514||639|
|Annual incidence, % (95% CI)||1·56 (0·67–3·07)||0·16 (0·004–0·87)|
|1 concomitant thrombophilic defect, n||33||50|
|Women with event, n||13||1|
|Observation period, years||608||1059|
|Annual incidence, % (95% CI)||2·14 (1·14–3·66)||0·09 (0·002–0·53)|
|>1 concomitant thrombophilic defect, n||16||22|
|Women with event, n||7||3|
|Observation period, years||240||554|
|Annual incidence, % (95% CI)||2·92 (1·17–6·01)||0·54 (0·11–1·58)|
Deficient women that had had at least one pregnancy had a lower absolute risk of venous thromboembolism (1·37 per 100 person-years; 95% CI, 0·80–2·19) than deficient women that had never been pregnant (2·96; 95% CI, 1·53–5·18); relative risk 0·5 (95% CI, 0·2–0·99) (Table IV). A similar difference was observed in non-deficient women. Of 54 deficient women that had had at least one pregnancy, 24 had never used oral contraceptives, of whom eight (33%) had pregnancy-related thrombosis. Twenty-seven of 47 deficient women that had never been pregnant had used oral contraceptives and in nine of them (33%) venous thromboembolism was related to oral contraceptives.
Table IV. Venous thromboembolism in women with at least one pregnancy versus women that had never been pregnant.
|Women with event, n||17||12||2||3|
|Observation period, years||1241||405||2169||430|
|Annual incidence, % (95% CI)||1·37 (0·80–2·19)||2·96 (1·53–5·18)||0·09 (0·01–0·33)||0·70 (0·14–2·04)|
|Relative Risk (95% CI)||<—0·5 (0·2–0·99)—>||<—0·1 (0·02–0·9)—>|
| ||<————————————14·9 (4·0–95·1)–————————–>|| |
| || ||<——————————–4·2 (1·3–18·8)——————————–>|
|Pregnancy-related events, n (%)||12 (71)||0||1 (50)||0|
|Oral contraceptives related events, n (%)||2 (12)||9 (75)||0||3 (100)|
Table V shows the distribution of pregnancy-related venous thromboembolism in the separate antithrombin, protein C and protein S cohorts. Of all the deficient women that had had at least one pregnancy, 22% experienced pregnancy-related venous thromboembolism, corresponding to 7% of pregnancies. In non-deficient women that had had at least one pregnancy, 1% suffered a pregnancy-related venous thromboembolism and 0·4% of these pregnancies were associated with this complication.
Table V. Pregnancy-related venous thromboembolism in antithrombin, protein C and protein S deficient and non-deficient women.
|Women with at least one pregnancy, n (%)||54 (53)||79 (65)||18 (55)||19 (63)||19 (50)||32 (64)||17 (57)||28 (68)|
|Pregnancy-related event, n (%)||12 (22)||1 (1)||8 (24)||0||3 (16)||0||1 (6)||1 (2)|
|Total pregnancies, n||162||245||45||62||63||109||54||74|
|Pregnancy-related event, n (%)||12 (7)||1 (0·4)||8 (18)||0||3 (5)||0||1 (2)||1 (1)|
|Venous thromboembolism during pregnancy, n||4||0||3||0||1||0||0||0|
|Venous thromboembolism during puerperium, n||8||1||5||0||2||0||1||1|
This study showed a high absolute risk of venous thromboembolism in deficient women (1·76 per 100 person-years). It was 59-fold higher than reported in the normal female population of comparable age (0·03 per 100 person-years) (Anderson et al, 1991, Nordström et al, 1992). Even female non-deficient relatives had a sixfold higher risk (0·19 vs. 0·03 per 100 person-years). The main risk modulators for venous thromboembolism were the presence of other thrombophilic defects, pregnancy and the use of oral contraceptives.
Concomitance of other single and multiple thrombophilic defects increased the absolute risk of venous thrombosis from 1·55 to 2·14 and 2·92 per 100 person-years, respectively, in deficient women. In non-deficient women, the absolute risk was 0·16 in the absence of any thrombophilic defect, compared to 0·09 and 0·54 per 100 persons-years when a single other defect and other multiple defects were demonstrated respectively. This finding suggests that the risk in non-deficient women was also increased by other multiple thrombophilic defects. Why other thrombophilic defects aggregated in our families with hereditary deficiencies of antithrombin, protein C or protein S remains to be clarified. Previously, we observed a similar aggregation in families with factor V Leiden (Libourel et al, 2002). It may be that aggregation contributes to the thrombophilic phenotype of symptomatic subjects. However, such selection bias is unavoidable, considering that asymptomatic subjects will not be tested for thrombophilic deficiencies or defects neither in studies, except family studies (with a symptomatic proband), nor in clinical practice.
Remarkably, women that had had a least one pregnancy were at lower risk of venous thromboembolism than women that had never been pregnant, whether they were deficient (1·37 vs. 2·96 per 100 person-years) or non-deficient (0·09 vs. 0·70 per 100 person-years). These differences can be attributed to the use of oral contraceptives. Of the first episodes of venous thromboembolism in deficient women that had had a least one pregnancy, 71% were related to pregnancy, and 12% were related to the use of oral contraceptives, while 75% of these episodes in deficient women that had never been pregnant were related to the use of oral contraceptives. This finding is explained by the planning of pregnancies. Deficient women who used oral contraceptives had mostly already suffered their first episode of venous thromboembolism before they became pregnant. Pregnancy-related venous thromboembolism predominated in deficient women who did not use oral contraceptives. Unfortunately, small numbers did not enable the proper analysis of the contribution of other thrombophilic defects in these subgroups of women.
Pregnancy-related venous thromboembolism in 22% of deficient women corresponded to 7% of pregnancies, compared with 1% of non-deficient women, which corresponded to 0·4% of pregnancies. Considering that 0·5–1 of 1000 pregnancies in the general female population is complicated by venous thromboembolism, deficient women had a 70- to 140-fold higher risk and non-deficient women a four- to eightfold higher risk. The highest risk was demonstrated in antithrombin-deficient women (18% of pregnancies), compared with protein C deficient women (5%) and protein S deficient women (2%).
The risk of pregnancy-related venous thromboembolism in our study was less pronounced than in three of four previous studies addressing this issue (Conard et al, 1990; De Stefano et al, 1994; Friederich et al, 1996; Pabinger & Schneider, 1996). Venous thromboembolism was reported in 37–47% of pregnancies in antithrombin-deficient women, 12–19% in protein C-deficient women and 13–27% in protein S-deficient women (Conard et al, 1990; De Stefano et al, 1994; Pabinger & Schneider, 1996) Differences in risk between our study and previous studies are probably due to differences in design and analysis. The higher risk in previous studies can be explained by the inclusion of probands, recurrences of venous thromboembolism and superficial phlebitis. The fourth study had a design comparable to our study and reported venous thromboembolism in 3·0%, 1·7% and 6·6% of the pregnancies in women with deficiencies of antithrombin, protein C and protein S respectively (Friederich et al, 1996).
The present study addressed the multicausality of pregnancy-induced venous thromboembolism, including all currently known thrombophilic deficiencies and defects. It comprised families with hereditary deficiencies of antithrombin, protein C and protein S and, frequently, concomitance of other thrombophilic defects. Hence, we had the opportunity to assess interactions between rare deficiencies that are recognised as strong risk factors for venous thromboembolism, and more prevalent but mild thrombophilic defects. However, as a result of numerous combinations of deficiencies and defects, the numbers of women were too small to obtain reliable risk estimates for specific combinations. Therefore, deficient and non-deficient women were classified according to the numbers of additional thrombophilic defects. It should be noticed that risk estimates were unstable and confidence intervals were wide, due to the small size of the subgroups in this approach. Moreover, a number of subjects were not tested for all thrombophilic defects, because of insufficient plasma samples. Our study had several other limitations because of its retrospective design. In a number of subjects, venous thromboembolism was not established by objective techniques because these were not available at the time of occurrence. Of 34 events, six episodes of proximal deep vein thrombosis (five in deficient women, one in a non-deficient woman) had occurred before 1974 and were not objectively confirmed. These were not excluded from analysis because five of six women experienced a recurrence, which was considered as indirect confirmation of the first episode or could have been classified as the first, confirmed event. Referral bias cannot be excluded, considering the university hospital setting of the study, but this was probably limited by testing all patients with venous thromboembolism for hereditary deficiencies. Of 1600 tested patients, 91 were deficient, corresponding to a prevalence of 5·7%, which agrees with previous reports on unselected patients (Heijboer et al, 1990). Selection bias was reduced by testing consecutive patients with venous thromboembolism and by an extraordinarily high response rate (95%) of eligible female relatives. Recall bias was reduced as much as possible by collecting clinical data prior to laboratory testing for thrombophilic deficiencies and defects. Inherent in its design as a family cohort study, controls (i.e. non-deficient relatives) were at higher risk of venous thromboembolism than the general population. Therefore, relative risks may have been underestimated. For this reason, we also compared absolute risks in deficient and non-deficient relatives with available data from previous population studies.
Our data supports the concept of a multicausal pathogenesis of venous thromboembolism. Both pregnancy and puerperium, and the use of oral contraceptives were shown to be the main determinants of venous thromboembolism in deficient women of reproductive age. A possible clinical implication of our findings might be that thromboprophylaxis could be considered during pregnancy and puerperium, and oral contraceptives discouraged in these women.
In conclusion, women of reproductive age with hereditary deficiencies of antithrombin, protein C or protein S are at high risk of pregnancy-related venous thromboembolism. Our data provides evidence that this risk is increased by multiple additional thrombophilic defects.
J. van der Meer conceived the study idea and all authors contributed to the study design, data abstraction and interpretation. N. J. G. M. Veeger did the statistical analysis. N. Folkeringa wrote the manuscript and all authors took part in its revision and approved the final version.