Pregnancy-related venous thrombosis is a major cause of maternal morbidity and mortality worldwide.1–3 Clinical diagnoses of deep venous thrombosis are often inaccurate,4 and it has been estimated that if diagnosis is based on clinical criteria alone, two of three cases will receive unnecessary anticoagulant treatment.5 However, there have been very few reports of clinically significant thrombosis verified by objective methods. In a prospective Swedish study in which diagnosis was confirmed by phlebography and plethysmography, the incidence of antepartum venous thromboembolic events was 7 per 10,000.6
Normal pregnancy is associated with a manifest shift of coagulation and fibrinolytic systems towards hypercoagulability. Although these changes are of physiologic importance in minimizing the risk of blood loss during delivery, they also increase the risk of thrombosis.7 Other predisposing factors are advanced maternal age, higher parity, operative delivery, immobilization, obesity, heart disease, malignancy, white race, history of thrombosis, thrombophilia, or familial thrombosis.1,8
In Sweden, there is an established pretreatment routine for verifying clinically suspected thrombosis related to pregnancy. A diagnosis of deep venous thrombosis is almost always based on findings at phlebography or ultrasonography. Pulmonary embolism is diagnosed by perfusion and ventilation lung scan or by pulmonary angiography. Treatment is routinely an inpatient procedure. Births are registered in the national birth registry and hospitalizations in the national patient registry. By merging the two registries, it was possible to make reliable estimates of the national incidence of thrombosis and of the magnitude of several risk factors.
The purpose of the present retrospective population-based study was to determine the incidence of pregnancy-related thrombosis in Sweden and its relationship to selected risk factors.
Over 99% of deliveries in Sweden are registered in the national birth registry,9 and all patients who have been hospitalized (even if only overnight) are registered in the national patient registry. The national birth and patient registries were used to identify all women with pregnancy-related thromboses during the 4-year period 1990–1993. Pregnancy-related thrombosis was defined as deep-vein thrombosis or pulmonary embolism. The diagnosis numbers classified as thrombosis (according to International Classification of Diseases, ninth revision) were deep-vein thrombosis (671D, 671E, and 671F) or pulmonary embolism (673C) related to pregnancy or the corresponding nonpregnant diagnosis numbers (451B, 452, 453 [C, D, W, or X]) or (415B), respectively, when in conjunction with pregnancy (ie, from 240 days before delivery to 6 weeks postpartum). Age, parity, weight, and height also were collected from registry data. Complete information on weight and height was available for less than 30% of the women, so those two variables were not included in analysis. Data were also collected on occurrence of cesarean delivery, multiple pregnancy, and preeclampsia.
The risk patterns before and after delivery might differ, so thromboses were divided into antepartum and postpartum subgroups. From 479,422 deliveries, a subgroup of 608 women accounted for 625 cases of thromboses. Of 44 women who gave birth more than once during the 4-year study period, 16 had recurrent thrombosis (one woman had thrombosis three times), 15 had one pregnancy-related thrombosis but a thrombosis-free pregnancy in 1993, and 13 women delivered twice during 1993. To avoid the same patient being included twice, only data on first pregnancies were included in logistic regression analysis. Of the 608 women with thromboses (308 antepartum, 300 postpartum) 90 had pulmonary embolisms and 518 had deep-vein thromboses (of which five had cerebral thromboses). Using logistic regression analysis, the women with thrombosis (n = 608) were compared with all thrombosis-free pregnant women in the country during 1993 (n = 114,940).
Data on smoking habits, routinely recorded at initial maternity unit appointments, were available in 96% of cases. Smoking was classified in terms of daily cigarette consumption, ie, at least 10 (heavy smokers), 1–9 (moderate smokers), or 0 (nonsmokers or not regular smokers) used as the reference class. Preeclampsia was defined as the combination of blood pressure higher than 139/89 mm Hg and albuminuria (at least 0.3 g/L). Parity was classified as para 0, para 1, para 2, or at least para 3, para 1 chosen as the reference class because bivariate analysis showed it to be associated with the lowest odds ratio (OR) for thrombosis. Maternal age was first classified according to five age groups (under 20, 20–24, 25–29, 30–34, and at least 35 years of age), but because bivariate analysis found no significant differences in ORs among the 20–24-, 25–29-, and 30–34-year-old age groups, in subsequent analysis they were combined as a 20–34-year-old age group and used as the reference class.
Bivariate and multiple logistic regression analyses were used to determine relationship between the outcome variable (the occurrence of thrombosis) and the explanatory variables (smoking, parity, maternal age, multiple pregnancy, preeclampsia, and cesarean delivery). All explanatory variables were included in analysis of postpartum thrombosis, and all except cesarean delivery in analysis of antepartum thrombosis. No significant interactions were found between any of the explanatory variables. Relative risk was determined in terms of ORs and 95% confidence interval (CI). The degree of linear association was calculated with the Mantel-Haenszel χ2 test. All statistical calculations were done by computer, using SPSS software (Statistical Package for the Social Sciences; SPSS Inc, Chicago, IL), P <.05 was considered statistically significant.
The incidence of pregnancy-related thrombosis was 13 per 10,000 pregnancies. The results of bivariate analysis of explanatory variables are shown in Table 1. The results of logistic regression analysis of antepartum thrombosis are shown in Table 2. Only parity differed significantly in regard to risk of thrombosis, and neither preeclampsia nor advanced age (at least 35 years of age) was associated with increased risk of antepartum thrombosis.
The results of multiple logistic regression analysis of postpartum thrombosis are shown in Table 3. Cesarean delivery was associated with a fivefold increased risk of postpartum thromboses. There was an increase in the rate of cesarean deliveries with increasing age (9% in the age group below 20 years, 10% in the 20–34-year-old group, and 18% among those 35 years or older). The risk of postpartum thrombosis was twice as great in the para 2 and para 3 subgroups compared with the para 1 (reference) subgroup, and threefold greater in the preeclampsia subgroup and the youngest age group (under 20 years old) than in their respective reference classes.
In the thrombosis series as a whole, (ie, antepartum and postpartum subgroups), smoking was associated with a significantly increased risk of thrombosis (OR 1.24; 95% CI 1.02, 1.51). There was a statistically significant association between increased risk of thrombosis and higher tobacco consumption (P = .007), but the difference between nonsmokers and smokers was significant only for heavy smokers (OR 1.41; 95% CI 1.04, 1.82 versus 1.11; 95% CI 0.87, 1.41 for moderate smokers).
A notable finding was that preeclampsia was associated with increased risk of thrombosis postpartum but not antepartum. The increased thrombin generation in preeclampsia has been assumed to be evidence for a prothrombotic state,10 but it has also been suggested that fibrinolysis is more pronounced than fibrin formation in women with severe preeclampsia.11 The routine recommendation of bed rest before and after delivery for women with preeclampsia might add to the risk of postpartum thrombosis. We noted that smoking was a significant risk factor for pregnancy-related thrombosis. The reason for the cigarette consumption-dependent increase in the risk of thrombosis is not known, but inhibited or defective fibrinolysis during pregnancy among smokers might be one explanation.12
Advanced maternal age has been reported as a risk factor for pregnancy-related thrombosis,8,13 but our results do not support that (Tables 2 and 3). Instead, the youngest women (under 20 years of age) were at threefold higher risk of postpartum thrombosis (Table 3). The higher prevalence of thrombosis among women above 35 years of age in bivariate analysis (Table 3) was not significant after adjustment for other variables, mainly because of an increased rate of cesarean delivery with increasing maternal age. When plethysmography was used to screen for thrombosis, Bergqvist et al14 found an association between cesarean delivery and a 1.8% incidence of thrombosis. In the present series, the prevalence of clinically significant thromboses among those delivered by cesarean was 0.9%, which was fivefold greater than among those who delivered vaginally. The proportion of postpartum thromboses associated with cesarean delivery was high (41%). Among women who died after delivery due to postpartum pulmonary embolism, the figure was even higher (76%).1
Women with postpartum thrombosis might be treated in internal medicine wards, so that group is difficult to identify because such cases tend to be inadequately classified (eg, all thromboses during pregnancy, or during the first 42 days postpartum, should be classified as 671 [D, E, or F], or 673C, but they often have nonpregnant thrombosis diagnosis numbers). By merging the two national registries we were able to identify women with thrombosis, irrespective of where they were treated. Previously, postpartum thromboses have been far more common than antepartum thromboses,15 which is in contrast to the predominance of antepartum thrombosis in a study by Macklon and coworkers,13 and with the even distribution found in our study. The lower proportion of postpartum thromboses in the latter two studies might be attributable to a change in obstetric practice that minimizes the duration of hospitalization and postpartum immobilization. Our findings are consistent with those of the Confidential Enquiry into Maternal Death during the period 1985–1993,1 showing deaths due to pulmonary embolism to be equally distributed between antepartum and postpartum subgroups.
Only pregnancies that resulted in birth were included; extrauterine pregnancies and spontaneous or legal abortions were excluded. We relied on diagnoses recorded in the national birth and patient registries, so clinically insignificant thromboses, or cases treated in outpatient settings were not included. In Sweden, thrombosis is routinely diagnosed on the basis of objective criteria, and it is unlikely that many thromboses were treated in outpatient clinics. The design of the study, ie, including 4 years of cases and 1 year of controls, and the inclusion of only the first pregnancy for thrombosis cases with multiple pregnancies during the 4-year study period might have introduced a selection bias; however, we believe that bias is small compared with the gain of power in analysis by using 4 years of cases.
Our results are supported to some extent by the findings of Bergqvist and coworkers5 whose prospective study in Sweden found the incidence of antepartum thrombosis to be 7/10,000. The relative homogeneity of the Swedish population and the national uniformity of obstetric practices made confounding of the results, eg, due to demographic factors, likely to be minimal. Details of cigarette consumption are routinely recorded at initial appointments at maternity units throughout the country, so any recall bias in that respect was precluded. The design of the study did not allow analysis of other genetic or acquired risk factors of thrombosis.
1. Department of Health. Welch Office. Scottish Office Department of Health. Department of Health and Social Services, Northern Ireland. Report on confidential enquiries into maternal deaths in the United Kingdom 1991–1993. London: HMSO, 1996.
2. Atrash HK, Koonin LM, Lawson HW, Franks AL, Smith JC. Maternal mortality in the United States, 1979–1986. Obstet Gynecol 1990;76:1055–60.
3. Högberg U. Maternal deaths in Sweden, 1971–1980. Acta Obstet Gynecol Scand 1986;65:161–7.
4. Genton E, Turpie AG. Venous thromboembolism associated with gynecologic surgery. Clin Obstet Gynecol 1980;23:209–41.
5. Ramsay LE. Impact of venography on the diagnosis and management of deep vein thrombosis. Br Med J Clin Res Ed 1983;286:698–9.
6. Bergqvist Å, Bergqvist D, Hallbook T. Deep vein thrombosis during pregnancy. A prospective study. Acta Obstet Gynecol Scand 1983;62:443–8.
7. Nordström M, Lindblad B, Bergqvist D, Kjellström T. A prospective study of the incidence of deep-vein thrombosis within a defined urban population. J Intern Med 1992;232:155–60.
8. de Swiet M. Thromboembolism. In: James D, Steer P, Weiner C, Gonic B, eds. High risk pregnancies. London: Saunders, 1996:597–603.
9. Cnattingius S, Ericson A, Gunnarskog J, Källen B. A quality study of a medical birth registry. Scand J Soc Med 1990;18:143–8.
10. de Boer K, ten Cate JW, Sturk A, Borm JJ, Treffers PE. Enhanced thrombin generation in normal and hypertensive pregnancy. Am J Obstet Gynecol 1989;160:95–100.
11. Borok Z, Weitz J, Owen J, Auerbach M, Nossel HL. Fibrinogen proteolysis and platelet alpha-granule release in preeclampsia/ eclampsia. Blood 1984;63:525–31.
12. Mercelina Roumans PE, Ubachs JM, van Wersch JW. Coagulation and fibrinolysis in smoking and nonsmoking pregnant women. Br J Obstet Gynaecol 1996;103:789–94.
13. Macklon NS, Greer IA. Venous thromboembolic disease in obstetrics and gynaecology: The Scottish experience. Scott Med J 1996; 41:83–6.
14. Bergqvist A, Bergqvist D, Hallbook T. Acute deep vein thrombosis (DVT) after cesarean section. Acta Obstet Gynecol Scand 1979;58:473–6.
15. Aaro LA, Juergens JL. Thrombophlebitis associated with pregnancy. Am J Obstet Gynecol 1971;109:1128–36.