Risk factors for first venous thromboembolism around pregnancy: a population-based cohort study from the United Kingdom

Venous thromboembolism (VTE) is a rare but serious maternal complication. Despite its low absolute rate of 1 to 2 per 1000 maternities^1,,-4  during pregnancy and the puerperium, it remains a leading cause of maternal mortality in developed countries, as well as being an important source of morbidity in the form of postthrombotic syndrome.⁵  Universal thromboprophylaxis may not be cost effective or safe because of the risk for allergy and bleeding.⁶  Therefore, routine thromboprophylaxis is recommended only for women considered at high risk for VTE on the basis of certain factors such as a previous VTE.⁷  However, there is disagreement and inconsistency regarding the characteristics that put women at higher risk of developing a first VTE during pregnancy or postpartum combined with a lack of data about the relative effect of those risk factors with respect to the absolute risk for VTE. For example, existing estimates of the increase in risk for VTE among pregnant women with high BMI (≥30 kg/m²) compared with those with normal BMI range from 1.5-fold to 5.3-fold higher during antepartum and postpartum periods.^2,8,-10  Similar inconsistencies surround women’s demographic risk factors (eg, maternal age), comorbidities (eg, diabetes), and possible pregnancy complications (eg, mode of delivery, obstetric hemorrhage, and other complications),^{2,3,8,10,,,,-15}  particularly as many studies have inappropriately assumed that these risk factors similarly affect occurrence of VTE in the antepartum and postpartum periods.^2,8,10,14 

Of the studies that have separately assessed antepartum and postpartum VTE risks,^3,9,11,12,15  most have used a case-control design or relied on cross-sectional analysis of hospital discharge records based on risk factor information recorded around the time of delivery. Neither of these options enables estimation of absolute risk (AR) for VTE based on recognized risk factors during the entire period of gestation and immediately after childbirth. As a result, the current Royal College of Obstetricians and Gynecologists (RCOG)⁷  and American College of Chest Physicians¹⁶  guidelines on obstetric thromboprophylaxis are based on suboptimal information to distinguish between women who are at lower or higher risk. Therefore, the aim of this study was to determine population-level ARs and relative risks for VTE according to women’s preexisting and pregnancy-related factors in both antepartum and postpartum periods, with the primary objective of allowing targeted provision of obstetric thromboprophylaxis. We also aimed to estimate specific ARs for VTE for women categorized as having low-, intermediate-, and high-risk pregnancies according to the United Kingdom’s RCOG guidelines on who should receive prophylaxis.

The Health Improvement Network (THIN)¹⁷  is an electronic database of medical records from more than 1500 general practitioners (GPs) and 429 UK practices. Approximately 98% of the UK population is registered with GPs, who are responsible for almost the entirety of a patient’s medical care. All GPs participating in THIN are trained to record information using the general practice Vision software.¹⁸  This software records individual-level data on diagnoses, prescription, lifestyle, and sociodemographic characteristics from all medical consultations. All data are then made anonymous before being added to the THIN database.¹⁹ 

Medical conditions and symptoms reported by patients to their GPs are recorded in the form of medical codes (Read codes), which are very comprehensive and use a hierarchal clinical classification system. These can be cross-referenced to the International Classification of Diseases.²⁰  Details on patients’ secondary care referrals are obtained from hospital discharge forms, which contain details on inpatient and outpatient hospital admissions, including a summary of the main diagnoses and procedures undertaken. Prescription data in primary care are very well recorded, as the computerized system used by GPs to enter medications is also used to print the paper copy of the prescription to be presented at the pharmacy by the patient.

According to the National Institute for Health and Clinical Excellence guideline for antenatal care,²¹  all pregnant women are required to undergo routine antenatal assessment to provide individualized care allowing primary care data to contain a comprehensive record of pregnancies in the United Kingdom. Both diagnoses of VTE²²  and fertility rates²³  have been validated with a high degree of accuracy in THIN or in the General Practice Research Database (GPRD), to which a number of practices in THIN also contribute. At the time of data analysis, THIN had information on 7.7 million patients who were broadly representative of the general UK population.

We identified all incident pregnancies ending in live birth or stillbirth for women aged 15 to 44 years who contributed data to THIN between January 1995 and July 2009. The pregnancy-related person time for each woman during the study period was divided into antepartum (from the date of conception to the pregnancy outcome) and postpartum (up to 12 weeks after the pregnancy outcome). This study was reviewed and approved by the THIN Scientific Review Committee (reference number 10-002R).

We identified all first VTEs including deep vein thrombosis or pulmonary embolism experienced by women (excluding superficial VTE) in our cohort. VTE cases and person time occurring within 1 month of the study start date were excluded to ensure that only incident cases were retained. As we were assessing only pregnancy-related VTE risk, only women without a prepregnancy history of VTE were included. VTE was based on a recorded medical code assigned by a physician and supplemented by evidence of anticoagulation prescription or a medical diagnosis indicating anticoagulant therapy within 90 days of the event or death within 30 days of the event date, as previously described in detail.^22,24  This definition has been previously validated with reasonable accuracy in a similar primary care database (GPRD), with 84% of all diagnoses confirmed on further investigation.

For each pregnancy, information on maternal factors during or before pregnancy was extracted from the patient’s medical record. Maternal age was considered as a time-varying covariate assessed in 3 equal-sized categories (15-24, 25-34, and 35-44 years). Furthermore, information on BMI (the latest measure recorded by the GP before the date of conception) and smoking status (the latest measure recorded before delivery) was also extracted for each pregnancy.

Pregnancy-related factors considered were mode of delivery, birth outcome (live or stillborn child), length of gestation, multiple gestation, and number of previous births. Pregnancy complications (including eclampsia/preeclampsia, hemorrhage, diabetes, and hypertension) were extracted using Read medical codes if they occurred during the pregnancy/postpartum period. Women were defined as having gestational diabetes if they had a first record of diabetes during pregnancy and no prior prescriptions for oral hypoglycemics or insulin. Gestational hypertension was defined as a medical code indicating the condition during pregnancy or at least 3 readings of high blood pressure after the second trimester (systolic, >140 mm Hg; diastolic, >90 mm Hg), with no antihypertensive treatment before pregnancy. We also separately investigated 2 common acute infections, urinary tract infection and acute respiratory tract infection (including pneumonia, acute bronchitis, chest infection, and influenza), during pregnancy.

Information on important comorbidities was extracted based on previous literature and the current RCOG guideline on thromboprophylaxis.⁷  We defined women as having an existing relevant medical comorbidity if they had ever been diagnosed with cancer, systemic lupus erythematosus (SLE), or nephrotic syndrome or had recorded varicose veins, inflammatory bowel disease (IBD), or cardiac disease (including congestive cardiac disease, coronary artery disease, congenital heart disease, cardiomyopathy, angina, or myocardial infarction) during or before pregnancy. We also defined women as having preexisting diabetes or preexisting hypertension, using combinations of Read and prescription codes for such conditions before conception.

We calculated absolute rates of VTE per 100 000 person-years by dividing the total number of events by the person-years of follow-up separately for antepartum and postpartum periods, which allowed us to directly compare risks per unit of time between antepartum and postpartum intervals.

To assess variations according to women’s risk factors, we calculated absolute rates per 100 000 person-years for each maternal characteristic (eg, maternal age category), each pregnancy-related characteristic and complication (eg, each mode of delivery, obstetric hemorrhage etc), and medical comorbidities (eg, varicose veins, cardiac disease, etc), which were assumed to affect the whole pregnancy, regardless of when during the pregnancy they were recorded. For factors relating to labor or the puerperium (eg, mode of delivery), we only estimated absolute rates in the postpartum.

We then estimated incident rate ratios (IRRs) of VTE associated with each risk factor, using Poisson regression models that were adjusted for maternal age, body mass index (BMI), smoking status, and number of previous births (model 1). These adjustments were selected on the basis of previous literature. We also included risk factors that were associated with increased risk for VTE in our initial antepartum and postpartum analyses (model 2).

As pregnancies are not independent events, a clustering term was fitted in the models to account for women experiencing more than 1 pregnancy during the study period. Our study was sufficiently powered such that for a relatively rare risk factor that affects 1% of all pregnancies, we had greater than 90% power to detect a 2-fold increase (IRR, 2) in risk for VTE in both antenatal and postpartum periods separately.

The current UK RCOG guidelines for thromboprophylaxis during pregnancy and postpartum use a set of recognized clinical factors to categorize pregnant women as low, intermediate, or high risk during the antepartum and postpartum periods to guide which women are offered hydration/mobilization and which women are offered low–molecular weight heparin (LMWH) as thromboprophylaxis.⁷  We estimated absolute rates of VTE, using the number of maternities as the denominator in this instance. This was done for each factor in isolation where possible.

In addition, we incorporated factors from our analysis that are not currently in the RCOG guideline but were highlighted in either the existing literature or this study (including preexisting diabetes, stillbirth, and preterm birth). These were placed in either the intermediate- or high-risk groups based on their absolute rate of VTE in this study. To ensure reasonable precision, we only calculated absolute rates for maternities in which 5 or more VTE events occurred among women with the risk factor. For this reason, SLE, cancer, nephrotic syndrome, cardiac disease, and IBD were grouped together, as current guidelines indicate that women with any of these risk factors should be considered for prophylaxis.

Pregnancies complicated by medical comorbidities in which there were no VTE events (eg, nephrotic syndrome) during the antepartum and/or postpartum periods were still included in the analysis so as to contribute to the denominator for calculating the incidence rate for VTE, but they were not investigated individually. In addition, high BMI (≥30 kg/m²) was separated into obese (BMI ≥ 30 Kg/m² and BMI < 40 kg/m²) and class 3 obese (BMI ≥ 40 kg/m²) categories, whereas acute respiratory tract and urinary tract infections were collectively assessed as stated in the current thromboprophylaxis guideline.⁷  All analyses were carried out using Stata SE11 (Statacorp, College Station, TX).

Among 280 451 women who had 1 or more pregnancies in our cohort, there were 376 154 pregnancies resulting in live births or stillbirths. Women’s basic characteristics for each pregnancy are summarized in Table 1. The overall incidence of VTE during the antepartum and postpartum periods was calculated to be 84 and 338 per 100 000 person-years, respectively.

The relative risk for VTE during antepartum was only marginally higher for women aged 35 years or older, women with a BMI of 30 kg/m² or higher, and women who were cigarette smokers, corresponding to a 42%, 50%, and 15% increased risk, respectively, compared with baseline (Table 2). Of the pregnancy-related characteristics and complications, only urinary tract infection was found to be significantly associated with an increased risk for VTE (88% increase in risk). However, medical comorbidities, including preexisting diabetes, recording of varicose veins, and IBD (but not preexisting hypertension or cancer), were all associated with higher rates of VTE, with absolute rates ranging from 216 (varicose veins) to 288 (IBD) per 100 000 person-years. Our IRRs for risk factors remained unchanged when mutually adjusted for other risk factors associated with an increased risk for antepartum VTE (model 2).

In the postpartum period, we found a 4-fold increased risk for VTE in women with a BMI of 30 kg/m² or higher (IRR, 3.75; 95% confidence interval, 2.76-5.08) compared with those with normal BMI (Table 3). However, the rate of VTE was only moderately higher for other maternal characteristics, including age 35 years or older and smoking status, when compared with baseline.

For pregnancy-related characteristics and complications, we found a 2-fold or greater increase in the rate of VTE compared with baseline for those with caesarean delivery, 3 or more previous births, obstetric hemorrhage, and preterm (<37 weeks) delivery (with absolute rates ranging between 637 and 963/100 000 person-years). Pregnancy ending in stillbirth was associated with a 6-fold increase in the rate of VTE compared with a live-birth outcome (absolute rate, 2444 VTEs/100 000 person-years), with rates associated with medical comorbidities (including varicose veins, IBD, and cardiac disease) ranging from 1330 to 2258 VTEs/100 000 person-years. When including all risk factors associated with an increased risk for VTE in our regression model (model 2), our estimates remained broadly similar.

Figures 1 and 2 contain the risk factors currently listed in the RCOG guideline⁷  for antepartum and postpartum thromboprophylaxis, with the addition of preexisting diabetes (antepartum), stillbirth, and preterm birth (both postpartum only). We found that among women who were classed as being at intermediate risk for antepartum VTE on the basis of a single risk factor, the highest risk for VTE was 180/100 000 pregnancies for women with preexisting diabetes (Figure 1). Postpartum women with a BMI of 40 kg/m² or higher had a higher risk for VTE (absolute rate, 221/100 000 pregnancies) than those women with any other single intermediate risk factor (preterm birth, prior medical comorbidities, caesarean delivery, and hemorrhage; Figure 2). Of factors currently used to classify women as low risk for postpartum VTE if occurring in isolation (ie, those whom are offered nonpharmacological thromboprophylaxis, including early mobilization and hydration), high risks for VTE were observed for women with a BMI between 30 and 40 kg/m² (absolute rate, 143/100 000 pregnancies) and varicose veins (absolute rate, 188/100 000 pregnancies) in the absence of other risk factors. The risk for VTE, however, in the postpartum after stillbirth was calculated to be 0.5%, which was higher than any other risk factor.

In this large, nationally representative cohort of almost 400 000 pregnancies, we have provided population-based estimates of the AR and relative risk for VTE in women during and immediately after pregnancy, combining their sociodemographic, lifestyle, and clinical risk factors to better inform targeting of thromboprophylaxis. We have demonstrated that except for preexisting diabetes, risk factors have a greater effect in the postpartum period in terms of influencing the AR for VTE than in the antepartum period. We also found that women with stillbirths, obstetric hemorrhage, high BMI, preterm birth, prior comorbidities (IBD and cardiac disease), and caesarean delivery have substantially higher risk for VTE postpartum. In contrast, cigarette smoking, maternal age 35 years or older, acute infection, and number of previous births were only moderately associated with VTE in both the antepartum and postpartum periods.

The use of nationally representative data not only makes our study findings generalizable to the majority of the pregnant women in the United Kingdom (ie, those who have not had a prior VTE) but also provides information on risk factors for VTE in a contemporary and population-based manner. Most previous studies have relied on cross-sectional analyses of hospital discharge records collected around the time of delivery,^2,10,15  were unable to separate first from subsequent VTE,^2,3,15  and relied on retrospective recall of risk factor information.^8,12,13  In contrast, our open-cohort approach to analysis and prospective nature of data recording enabled us to look at the effect of existing and pregnancy-related risk factors on the incidence and relative risk for VTE separately for the antepartum and postpartum periods.

A potential limitation of our study is the lack of validation studies of obstetric complications in our data set and the scarcity of national UK or international data estimating the true prevalence of obstetric complications in the general population. Although all major medical events in secondary care should be recorded in the general practice notes, minor complications that are well managed in the hospital setting may not be recorded in primary care data. For instance, the estimated prevalence of obstetric hemorrhage ranges from 2% to 8%, with the prevalence of severe hemorrhage estimated to be between 0.01 and 1.86%.^25,-27  We found a prevalence of 1% in our study, which indicates our cases are more likely to represent severe hemorrhage (the same may be the case for gestational hypertension and preeclampsia).

We also acknowledge that certain risk factors including mild to moderate preeclampsia might not always get reported in the hospital discharge summaries that are sent to GPs, thus underestimating their prevalence.^28,,-31  In particular, residual confounding caused by the underascertainment of preeclampsia could have accounted for some of the postpartum effect we observed for stillbirths and preterm births. However, around 50% of stillbirths are unexplained.³² 

In our study, none of the 6 women with stillbirth who developed VTE had recorded coexisting medical risk factors or pregnancy-associated complications. We do acknowledge that we were not able to assess certain parameters, including mother’s ethnicity and fertility treatments that may often be associated with increased risk for VTE. However, we believe that the following should be considered: 91%³³  of the UK population is white, which we believe would have limited the effect of confounding by race on our estimates, and in addition, a case–control study from the UK showed no association between antenatal pulmonary embolism and ethnicity.¹³  Also, the overall reporting of fertility problems in the United Kingdom, using primary care data,³⁴  is reported to be 0.5% per annum, with an infertility treatment rate of 0.1% per annum. Moreover, the actual number of women conceiving after treatment may be even lower, although we were not able to evaluate this in our data. This depicts the limited scope of this variable to modify our conclusions from these data.

A few other aspects of our data are worthy of note. Our data relied on BMI measured before pregnancy, in line with most existing research. However, a previous study found that weight gain during pregnancy was a more important predictor of VTE risk,¹²  something we were unable to assess. Our finding of a high risk for VTE in those with a prior diagnosis of varicose veins should also be interpreted with caution, as it may potentially be a consequence of a past unrecorded or concurrent deep vein thrombosis. Finally, we were not able to assess thrombophilia as a potential risk factor in our study. We believe that a diagnosis of thrombophilia cannot be used to predict VTE outcome, as routine thrombophilia screening is not recommended for pregnant women.

Our definition of VTE had a positive predictive value of 84% when validated among women of childbearing age in the GPRD,²²  a UK primary care database to which a large proportion of the practices in THIN also contribute. Such validation, however, does not give an indication of the negative predictive value (or sensitivity), and we cannot ignore the potential for our absolute rates of VTE to be underestimated if some anticoagulant prescriptions emanated from secondary, rather than primary, care. Despite this, Huerta et al³⁵  reported an age- and sex-standardized incidence rate of VTE using the GPRD that was similar to that observed in other Western studies when using a VTE definition identical to that of the present analysis.

It may be argued that our estimates do not take into account the number of pregnant women already receiving thromboprophylaxis before a VTE event. Although we excluded women with prior history of VTE from our study, we found that 0.4% of pregnant women without VTE received heparin/LMWH prescribed by a GP during the antepartum/postpartum period (also included in the analysis), which may be a result of certain clinical risk factors, as suggested in the current RCOG guideline. This, however, is certainly an underestimate, as we were unable to look at thromboprophylaxis prescriptions emanating from secondary care. We believe though that the effect of this would be small, as the first national RCOG guidelines for antenatal thromboprophylaxis were only published in 2004, with an updated version published in 2009. In the light of this fact, we calculated the risk for VTE during pregnancy and postpartum pre- and post-2004, which showed no statistical difference in the postpartum and a 46% (statistically significant) increase in the risk for VTE post-2004 in the antepartum period, suggesting a minimal effect of the national guidelines on the incidence of VTE.

We should also emphasize that assessing the effectiveness of the current national guidelines for obstetric thromboprophylaxis is beyond the scope of this study. One way of assessing this could be calculating the number needed to treat (NNT) and number needed to harm from our data. For instance, if we were to assume that prophylaxis reduces the risk for VTE by 50%, as has been reported in trials outside of pregnancy,³⁶  then based on our estimate of the AR, 89 VTE events could be prevented per year in women whose pregnancies end in caesarean delivery (NNT, 1980). These values should be interpreted with caution, as they are based on speculative data regarding the reduction in risk resulting from LMWH, from which there is an absence of RCT data in pregnant women.

Our relative increases in the risk (more than 2-fold) of VTE in those with preterm birth, obstetric hemorrhage, caesarean delivery, stillbirth, and varicose veins compared with women without those respective risk factors are of roughly similar magnitude to those reported in other studies.^{2,3,8,10,12,15,37}  We believe that the strong association between VTE events and stillbirths we observed is a finding of real importance that has received only limited attention to date.³⁷  Our finding of low relative increases in the risk for VTE among women aged 35 years or older, current smokers, and those with high BMI during the antepartum period are also in concordance with previous studies.^3,10,15 

Our results may have important implications for the way obstetric thromboprophylaxis is delivered in the healthcare settings of developed countries, and we hope that they will aid the targeting of such prophylaxis in 3 ways.

First, we found an increased risk for VTE among pregnancies of women with preterm birth or stillbirth, factors that have received limited consideration to date and are not currently incorporated in the guidelines for risk assessment of VTE. If they were incorporated, then thrombotic events associated with those risk factors could potentially be prevented.

Second, our data support many of the existing national RCOG guideline recommendations (in terms of high absolute rates), especially that postpartum thromboprophylaxis may be indicated in women with very high BMI (≥40 kg/m²), those who have prior comorbidities, those who have obstetric hemorrhage, or those who have a caesarean delivery.

Third, our results showed a high risk for VTE in women with a BMI between 30 and 40 kg/m² or varicose veins, even if these risk factors occur in isolation, which may require careful consideration. The recommendation on whether thromboprophylaxis with LMWH may be effective in pregnant and postpartum women with the above highlighted risk factors will of course highly dependent on the risk reduction from prophylaxis, for which there is a noticeable void of data from pregnant women.

Another important consideration is the costs involved in prophylaxis, both financial and in terms of the tolerability surrounding a daily heparin prescription, not to mention the well-recognized adverse effects of allergy and bleeding. For instance, the benefits would need to be weighed against a risk for major hemorrhage, which is believed to occur in 1% of pregnant women.¹⁶  Such a risk–benefit analysis clearly goes beyond the scope of the present work; however, we believe our presentation of population-based risks for VTE, based on a number of established risk factors, goes some way to helping clinicians involved in making decisions in this area.

In summary, our study provides new and interesting observations on the absolute rate of VTE in a range of categories of pregnancy and postpartum period. It provides valuable information to clinicians for better decision making in terms of identifying high-risk pregnant and postpartum women who may require some form of thromboprophylaxis.

Contribution: A.A.S., J.W., and M.J.G. conceived the idea for the study, with L.J.T. and K.M.F. also making important contributions to the design of the study, which used a data set created by L.J.T. and L.F. linking maternal records from the THIN database to those of their offspring; A.A.S. carried out the data management and analysis and wrote the first draft of the manuscript; C.N.-P. provided clinical input and interpretation at all stages of the project; all authors were involved in the interpretation of the data, contributed toward critical revision of the manuscript, and approved the final draft; and A.A.S. had full access to all of the data and had final responsibility for the decision to submit for publication.

Conflict-of-interest disclosure: C.N.-P. was codeveloper of the currently available guidelines on VTE prophylaxis in pregnancy issued by the RCOG (green top guideline 37a) and has received honoraria for giving lectures from Leo Pharma and Sanofi-Aventis (makers of tinzaparin and enoxaparin LMWHs used in obstetric thromboprophylaxis) and has received payment from Leo Pharma for development of an educational slide kit about obstetric thromboprophylaxis. The remaining authors declare no competing financial interests.

Correspondence: Matthew J. Grainge, Division of Epidemiology and Public Health, School of Community Health Sciences, University of Nottingham, Clinical Sciences Building Phase 2, City Hospital, Nottingham, United Kingdom, NG5 1PB; e-mail: matthew.grainge@nottingham.ac.uk.