Risk of first venous thromboembolism in and around pregnancy: a population-based cohort study

Authors


Dr 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 NG5 1PB, UK.
E-mail: matthew.grainge@nottingham.ac.uk

Summary

Knowledge of the absolute and relative risk of venous thromboembolism (VTE) in and around pregnancy would be crucial in identifying when to commence and cease thromboprophylaxis in women who would benefit from such intervention. We addressed this hypothesis using a large prospective primary care database from the United Kingdom, containing details on 972 683 women aged 15–44 years between 1987 and 2004. Risks of a first VTE during antepartum, postpartum and outside of pregnancy were compared using Poisson regression. The rate of VTE during the third trimester antepartum was six times higher than time outside pregnancy [Incidence Rate Ratio (IRR) = 6·1; 95% confidence interval, 4·7–7·9]. In contrast, both the first (IRR = 1·6) and second (IRR = 2·1) trimesters conferred little increase in risk. The first 6 weeks postpartum was associated with a 22-fold increase in risk, with the peak occurring in the first 3 weeks. Increased age was found to be associated with VTE during postpartum and outside of pregnancy, but not during antepartum. Our findings of a notably raised risk of VTE persisting for 3 weeks postpartum and of a raised antepartum risk constrained to the third trimester have implications for modifying the current recommendations for VTE prophylaxis in pregnancy and the puerperium.

Venous thromboembolism (VTE) remains one of the leading direct causes of maternal deaths in developed countries (Drife, 2003, Centre for Maternal and Child Enquiries 2011). Therefore, knowledge of the absolute and relative risk of VTE in the antepartum and postpartum periods is of crucial importance in identifying pregnant women or those who have recently delivered, who would benefit most from thromboprophylaxis, which is known to be generally safe and highly effective (Greer & Nelson-Piercy, 2005; Roderick et al, 2005). Equally, by determining low risk periods, unnecessary thromboprophylaxis and its potential harm to both the mother and fetus could be avoided. However, clinical guidelines developed by the American College of Chest Physicians (Bates et al, 2008), the UK Royal College of Obstetricians and Gynaecologists (RCOG 2009), and the Society of Obstetricians and Gynaecologists of Canada (Kent et al, 2000) are all limited by both the quality and quantity of the current evidence base and understandably have had to rely upon expert consensus opinion in formulating the recommendations made.

One example of the existing scarcity of evidence is the lack of precise and consistent estimates of the absolute risk of VTE in and around pregnancy. Reported rates vary considerably, from 28 to 86 per 100 000 maternities during the antepartum and 28–128 per 100 000 during the postpartum periods (McColl et al, 1997; Andersen et al, 1998; Gherman et al, 1999; Lindqvist et al, 1999; Chan et al, 2001; Salonen Ros et al, 2001; Simpson et al, 2001; Heit et al, 2005, James et al, 2006; Jacobsen et al, 2008; Liu, et al 2009; Virkus et al, 2011). Inconsistencies also surround the trimester of pregnancy and weeks postpartum during which a woman is at highest risk of VTE and the extent to which age affects these risks (Heit et al, 2005; James et al, 2005; Jacobsen et al, 2008; Pomp et al, 2008). The wide variations in the existing literature may be partly due to differences in study design and study population. For instance, studies solely reliant on International Classification of Disease medical codes (Simpson et al, 2001; James et al, 2006, Liu, et al 2009) without validation of VTE cases could potentially lead to overestimation of the risk. Secondly, the method used in most studies to calculate risk, using number of maternities as denominator (McColl et al, 1997; Andersen et al, 1998; Gherman et al, 1999; Lindqvist et al, 1999; Chan et al, 2001; Simpson et al, 2001; James et al, 2006, Jacobsen et al, 2008; Liu, et al 2009) does not account for differences in the duration of antepartum versus postpartum time, nor does it allow for an effective comparison of VTE risk with a woman’s background risk while not pregnant. Considerably fewer studies have specifically compared rates of VTE around pregnancy either with those outside of pregnancy or in non-pregnant women, and the evidence from these is inconsistent (Salonen Ros et al, 2001; Heit et al, 2005; Virkus et al, 2011).

Accurate estimates of rates of VTE that are generalizable to the majority of pregnancies will aid both women and practitioners in their decision making for prevention of VTE. We therefore aimed to determine the absolute and relative risks of VTE in women of child-bearing age around pregnancy with the primary objective of informing revisions to existing clinical guidelines.

Methods

Study population

The Health Improvement Network (THIN) is a computerized database containing longitudinal individual-level health information on approximately 3·9 million anonymized patients from 255 general practices across the UK (Tata et al, 2008). Patients from participating practices are known to be broadly representative of the UK population in terms of age distribution and socioeconomic factors (Lewis et al, 2007). Accuracy of recorded diagnoses of both VTE (Lawrenson et al, 2000) and fertility rates (Tata et al, 2007) in primary care databases have been previously validated. Our cohort consisted of all women aged between 15 and 44 years who were registered at a THIN general practice between April, 1987 and November, 2004. An open cohort design was used, within which women could enter and exit the study at different calendar times and ages. We used prospectively recorded information entered during the computerized registration period to identify live births and incident VTE events during a woman’s potentially fertile period. Women with a VTE recorded prior to the start of the study period were excluded.

Person time at risk

The total person-time for each woman during the study period was divided into antepartum and postpartum periods for live birth pregnancies and other person-time outside of pregnancy. Gestational time periods for pregnancies resulting in non-live births were excluded from the study. We defined our exposure and baseline time periods as follows:

Antepartum and postpartum periods

Antepartum period was defined as the date of conception to the date of delivery. For women without recorded information on the gestational age of the baby at delivery, we defined the antepartum period as the 9 months prior to date of delivery. This was subsequently divided into trimesters and individual weeks. The postpartum period was defined as the 3 months following delivery, which was subsequently divided into early and late postpartum periods (first and second 45 d after delivery respectively) with the early postpartum period further divided into individual weeks.

Baseline period

To calculate baseline rates we used the period women spent outside antepartum and postpartum periods during their registration with the general practice. For women with no recorded pregnancies this included their entire registration during our study period between which they were 15 and 44 years of age. We refer to this time period henceforth as ‘time outside pregnancy’.

Definition of VTE

We considered only the first VTE in the study period. VTE was based on a recorded medical code assigned by a physician supplemented by having either a recorded anticoagulant prescription in the 90 d following the event, a medical diagnosis indicating that prescription of anticoagulants took place in the 90 d following the event, or death within 30 d of the event date. Using this definition, 84% of VTE cases were validated in a previous study using similar electronic primary care data (Lawrenson et al, 2000). One limitation of this approach in pregnancy in particular is that we may inadvertently overestimate the rates of VTE in the first and second trimester where low molecular weight heparin is used both as prophylaxis and therapeutically. We therefore planned a sensitivity analysis restricting all analyses to include only those women with VTE who were prescribed either warfarin or had a medical code indicating anticoagulant therapy in their records.

Statistical analysis

Rates of VTE (expressed per 100 000 person-years) and 95% confidence intervals (CI) were obtained by dividing the number of VTE events by the person-years of follow-up. Incident Rate Ratios (IRRs) were calculated using a Poisson Regression model to compare the rates of VTE in antepartum and postpartum periods with those outside of pregnancy. The absolute and relative rates of VTE in each trimester of pregnancy as well as in the early and late postpartum periods were obtained in a similar fashion.

Given that the risk of VTE is known to increase with age (White, 2003); we assessed the potential for both confounding and effect modification by age by treating this as a time-varying covariate (fitted as a 3-level category; 15–24, 25–34 and 35–44 years). IRRs were calculated to assess the effect of age on VTE risk, separately for the antepartum, postpartum and non-pregnant periods, using the 25–34 years age band as the reference. An interaction term between pregnancy status (antepartum, postpartum and time outside pregnancy) and age group was fitted to formally test for an interaction between these terms. All subsequent analyses focusing on the relationship between pregnancy and VTE risk were adjusted for age group (when not stratified by this term) and calendar year (indicating observation time prior to and post 1996). All analyses were carried out using Stata (version 11). This study was reviewed and approved by the THIN Scientific Review Committee (reference number 10-002R).

Results

Patient characteristics

A total of 972 683 women in our cohort contributed 5 361 949 person years of follow up (Fig 1). The median follow-up time for each woman was 6·4 years (interquartile range = 2·2–14·1). There were 207 327 live birth pregnancies, contributing 145 380 person-years of antepartum time and 50 907 person-years of postpartum time, leaving 5 131 554 person-years of time outside of pregnancy.

Figure 1.

 Overview of the study population. 1 Non-live birth includes spontaneous abortions, terminations of pregnancy and stillbirths. 2 Person-years include antepartum and 3 months postpartum. 3 Person-years include time period of ever pregnant women spent outside pregnancy and all time within database for women with no recorded pregnancy.

Risks of VTE in pregnancy overall and by age

For women of all ages, the rate of VTE ranged from 20 per 100 000 person years outside pregnancy (95% CI, 19–21) to 228 per 100 000 person years during the postpartum period (95% CI, 189–273) (Table I). The overall rate of VTE in and around pregnancy (including both antepartum and postpartum periods) was 107 per 100 000 person years (95% CI 93–122 per 100 000 person years). Outside of pregnancy, women in the oldest age band (35–44 years) had a 50% higher rate of VTE than women aged 25–34 years. The rate of VTE did not increase with age in the antepartum period, however, in the postpartum period women aged 35 and over had a 70% increase in risk compared to 25–34 year olds (corresponding to an excess absolute risk of 1·6 per 1000 person-years).

Table I.   Rates of VTE by age and pregnancy status.
VariableVTE eventsPerson yearsRate*95% CIIRR95% CIRate difference*95% CI
  1. IRR, incidence rate ratio; CI, confidence interval; VTE, venous thromboembolism.

  2. *Per 100 000 person years.

  3. †Increase in the risk compared to 15–25 years.

  4. ‡Increase in the risk compared to 25–35 years.

Outside pregnancy
 All ages10245 131 5542019–21    
 15–24 years1481 362 621119–13 0·580·48–0·71 
 25–34 years3431 849 9221917–20 1·001·30–17·17·6†5·0 to 10·3
 35–44 years5331 919 0112825–30 1·509‡6 to 12
Ante partum
 All ages94145 3806552–79    
 15–24 years1835 6925032–80 0·750·44–1·27 
 25–34 years6089 7386752–86 1·0017†−12 to 45
 35–44 years1619 9508049–1301·190·69–2·0813‡−29 to 56
Postpartum
 All ages11650 907228189–273   
 15–24 years1611 61313884–2250·610·35–1·06 
 25–34 years7031 385223176–2811·0085†−0·1 to 170
 35–44 years307909379265–5421·701·10–2·6015611 to 301

Overall, we observed a 3·5-fold (95% CI 2·8–4·3) increase in the rate of VTE in the antepartum and an 11·9-fold (95% CI 9·8–14·5) increase in the postpartum period compared to outside pregnancy (Table II). These IRRs remained similar when stratified according to age. This was confirmed by a non-significant test for interaction between age and pregnancy status (P = 0·40).

Table II.   Overall and age-specific incidence rate ratios of VTE in ante and postpartum compared to outside pregnancy.
VariableIncidence rate ratio unadjusted95% CIIncidence rate ratio adjusted*95% CI
  1. CI, confidence interval.

  2. *Adjusted for age and calendar year.

All ages
 Outside pregnancy1·01·0
 Antepartum3·22·6–4·03·52·8–4·3
 Postpartum11·49·4–13·811·99·8–14·5
15–24 years
 Outside pregnancy1·01·0
 Antepartum4·62·8–7·54·62·8–7·6
 Postpartum12·67·5–21·212·47·4–20·8
25–34 years
 Outside pregnancy1·01·0
 Antepartum3·62·7–4·73·52·6–4·6
 Postpartum12·09·3–15·511·69·0–15·0
35–44 years
 Outside pregnancy1·01·0
 Antepartum2·81·7–4·72·61·6–4·4
 Postpartum13·69·4–19·712·58·6–18·0

Risks by trimester and weeks postpartum

Absolute rates of VTE were roughly similar in the first and second trimester and the late postpartum period with incidence rates of 30, 40 and 35 per 100 000 person years respectively (Table III; Fig 2). The rate of VTE was higher in the third trimester (114 per 100 000 person years; 6·1-fold increase in risk) and peaked in the early postpartum period (421 per 100 000 person years; 22-fold increase in risk). Figure 3 shows the rate of VTE per 100 000 person-years by weeks of antepartum and postpartum period showing that the rate of VTE is much higher in the first 3 weeks of postpartum compared to late third trimester of the antepartum period. However, the rate of VTE sharply declined from the fourth week onward in the postpartum period.

Table III.   Rate of VTE per 100 000 person years during different time periods of pregnancy and postpartum compared to time outside pregnancy.
Time periodRate*95% CIIncidence rate ratio unadjusted95% CIIncidence rate ratio adjusted†95% CI
  1. CI, confidence interval.

  2. *Per 100 000 person years.

  3. †Adjusted for age and calendar year.

Outside pregnancy2019–211·001·00
Antepartum
 First trimester3017–501·40·8–2·51·60·9–2·8
 Second trimester4025–632·01·2–3·22·11·3–3·4
 Third trimester11489–1465·74·4–7·46·14·7–7·9
Postpartum
 Early postpartum421349–50921·117·3–25·722·118·1–27·1
 Late Postpartum3518–671·70·9–3·41·80·9–3·5
Figure 2.

 Log10 of the rate of venous thromboembolism (VTE) and 95% confidence intervals during different time periods during and outside pregnancy. Outside pregnancy: Includes time for ever pregnant women spent outside antepartum and postpartum period and all time for women with no recorded pregnancy during study period. Early postpartum: First six weeks from date of delivery. Late postpartum: Subsequent six weeks postpartum.

Figure 3.

 Rate of venous thromboembolism (VTE) per 100 000 person-years in weeks of antepartum and postpartum period.

Sensitivity analysis

Overall, over 90% of our VTE cases had warfarin prescribed in primary care within 90 d of the event occurring. For those VTE events that occurred in the first and second trimester this proportion was far lower (eight out of 26, 31%) increasing to 78% in the third trimester and 99% in the postpartum period. Results remained broadly unchanged when using this restricted definition (Table IV). Absolute risks were therefore somewhat lower in the first and second trimesters but the rate ratios did not differ significantly from the null as per when the inclusive VTE definition was used. Rates and rate ratios for the third trimester and postpartum periods therefore changed little.

Table IV.   Rate of VTE during different time periods of pregnancy and postpartum compared to time outside pregnancy using a restricted case definition.
Time periodNumber of VTE eventsRate*95% CIIncident rate ratio†95% CI
  1. CI, confidence interval.

  2. *Per 100 000 person years.

  3. †Adjusted for age and calendar year.

Outside pregnancy9891918–201·0
Antepartum overall503426–451·941·4–2·5
First trimester372–210·30·1–1·2
Second trimester5114·4–250·50·2–1·4
Third trimester427857–1054·33·1–5·9
Postpartum overall115225188–27112·310·1–15·0
Early postpartum106417345–50522·818·6–28
Late postpartum93518–671·90·9–3·6

Discussion

Using data from a large population-based cohort, we report a relatively low rate of first VTE in pregnancy; however, this rate was still much higher than that seen in the time outside pregnancy, with a noticeably raised risk in the first 3 weeks postpartum. During the antepartum period, women in their third trimester were at a 6·1-fold increased risk of first VTE compared to their time outside pregnancy whereas in the first and second trimester this rate was only marginally higher. We found that age was an important risk factor in the development of VTE in the postpartum period, but not during the antepartum period, with women aged 35 years and over having an absolute excess rate of VTE of about 1·6 per 1000 compared to women aged 25–34 years (a 70% relative increase). Our findings are generalizable to the majority of pregnant women who have not had a prior VTE providing valuable information to accurately assess the risk of VTE and the potential need for prophylaxis.

Strengths and weaknesses

Our study used information from approximately 1 million women aged between 15 and 44 to accurately determine the absolute and relative risks of first VTE within specific antepartum and postpartum periods in a contemporary, population-based manner. We used an open cohort study approach to our analysis that calculated the rate of VTE in pregnancy taking account of time-varying risk periods using person-years as the denominator. This method adjusted for differences in the duration in antepartum and postpartum periods unlike traditional approaches (using number of maternities as the denominator), which assumes that the time women spend in antepartum and postpartum periods is constant. Furthermore it also enabled us to capture the time period outside pregnancy and make appropriate unbiased comparisons of these rates with those in the antepartum and postpartum periods.

One potential limitation with respect to the generalizability of the study is that we excluded pregnancies not resulting in live births due to lack of data available to differentiate between spontaneous miscarriage, termination of pregnancy and stillbirth. We believe that the effect of those exclusions on our estimates is likely to be small with only 20 VTE events occurring in women with these non-live outcomes excluded. Secondly, our comparisons to time outside pregnancy need to be interpreted with the understanding that some of the non-pregnant women will be taking a combined oral contraceptive which is known to increase VTE risk (Farmer et al, 1997), however we believe that trying to characterize periods of oral contraceptive use within the non-pregnant time was beyond the scope of the present work. Finally, the absolute rates of VTE were lower among our cohort in the period outside pregnancy than in those from Denmark and the USA (Heit et al, 2005; Virkus et al, 2011). Our definition of VTE was found to have a positive predictive value of 84% when validated among women of childbearing age in the General Practice Research Database (GPRD), a UK primary care database to which a large proportion of the practices in the THIN database also contribute (Lawrenson et al, 2000). 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 al (2007) reported an age- and sex-standardized incidence rate of VTE using the GPRD, which was similar to that observed in other Western studies, when using an identical VTE definition to that of the present analysis.

Our ability to identify the use of either heparin or warfarin in primary care allows us a greater level of detail than previous studies. When we applied a restricted case definition including only cases with evidence of therapeutic anticoagulant use (i.e. excluding any potential use of prophylactic heparin) we found that absolute rates of VTE were reduced in the first and second trimester but not in the third trimester or in the postpartum period. Given that heparin is the treatment of choice in early pregnancy this is not surprising but means that our estimated rates may actually be too low in the sensitivity analysis, as we could have potentially removed VTE cases that were receiving heparin as treatment rather than prophylaxis. Nonetheless, the overall interpretation of our results is not altered.

Comparison with other studies

Only three previous studies have compared rates of VTE in pregnancy with those outside of pregnancy (Salonen Ros et al, 2001; Heit et al, 2005; Virkus et al, 2011). Recently, Virkus et al (2011) reported absolute and relative risks of VTE within individual trimesters of antepartum and weeks of early postpartum among Danish women, which were similar to those in the present study. However, their rates were higher in the late third trimester, suggesting that the risk of VTE in that period is similar to that of weeks immediately following childbirth, in contrast to our results. This contradictory finding could potentially be due to lack of precise date of diagnosis of VTE. The use of date of admission as the date of diagnosis by Virkus et al (2011) might have shifted the distribution of VTE event times to the left, so that events which occurred in the 1–2 weeks following delivery were recorded as taking place just before delivery. This could explain why, in our study, the incidence of VTE in the weekly intervals immediately following delivery were noticeably higher than the weeks preceding delivery, unlike in the Danish study (Virkus et al, 2011). Our findings were also broadly similar to those reported in a population-based cohort in Olmsted County in the United States (Heit et al, 2005). However, their considerably smaller sample size (50 080 live births and 105 VTE events occurring around pregnancy) meant that there was insufficient power to rigorously compare risks in each trimester or postpartum period with time outside pregnancy or by age strata (Lee, 2005).

Implications for clinicians and policymakers

Whilst guidelines issued in 2004 on the prevention of VTE around pregnancy could possibly have contributed to a reduction in the number of maternal deaths from VTE recently reported in the United Kingdom (Centre for Maternal and Child Enquiries 2011), there is still considerable morbidity associated with VTE, which further targeted prophylaxis could effectively reduce. In our analysis we demonstrated high absolute risks of VTE in the third trimester and first 3 weeks postpartum, whereas during the first two trimesters and late postpartum periods the risks are similar to time outside pregnancy. Furthermore, when comparing postpartum women of age 35 years and over to the majority of pregnant women we found only a small excess absolute risk and no difference in risk by age during antepartum periods.

We believe our results have important implications for the way in which thromboprophylaxis is delivered in the health care settings of developed nations and would hope that they will aid the targeting of such prophylaxis in three ways. Firstly, our observation that the highest rates of VTE occur mainly within 3 weeks postpartum suggests that giving thromboprophylaxis in those considered at high risk for this length of time may be appropriate. Secondly, we have found that the risk in the first and second trimester is not greatly increased for those women without previous VTE compared with time outside pregnancy. Therefore, initiation of prophylaxis where it is needed could potentially be delayed until the start of the third trimester, thus avoiding any adverse effects of low molecular weight heparin and prolonged antenatal use in some women. Finally, current clinical guidelines often include age (RCOG 2009), i.e. being >35 years old, as an independent risk factor for VTE. Our results imply that the absolute excess risk in postpartum women 35 years or older compared to the majority of postpartum women is of the order of 1·6 per 1000 person years. If we assume that thromboprophylaxis is 100% effective at preventing VTE then our results imply that to prevent one excess VTE occurring in this age group we would need to give prophylaxis to approximately 640 women. Overall we believe our study provides new evidence that will aid the ability of clinicians to judge which women are and are not at high risk of VTE related to pregnancy and therefore these findings should influence the development of updated thromboprophylaxis guidelines and impact the delivery of care to pregnant women across the developed world.

Acknowledgements

CNP has received honoraria for giving lectures from Leo Pharma and Sanofi Aventis (makers of tinzaparinand enoxaparin LMWHs used in obstetric thromboprophylaxis) and has received payment from Leo Pharma for development of an educational ‘slide kit’ about obstetric thromboprophylaxis. No other authors have competing interests to declare. AAS is a recipient of a studentship from the Aga Khan Foundation and JW is in receipt of a NIHR/University of Nottingham Senior Clinical Research Fellowship.

Author contributions

JW and MJG conceived the idea for the study, with AAS, LJT and KMF also making important contributions to the design of the study, which used a dataset created by LJT linking maternal records from the THIN database to those of their offspring. AAS carried out the data management and analysis and wrote the first draft of the manuscript. CNP provided clinical input and interpretation at all stages of the project. All authors were involved in the interpretation of the data, contributed towards critical revision of the manuscript and approved the final draft.

Conflict of interest

The authors have no competing interests.

Ancillary