Thromboprophylaxis during pregnancy and the puerperium: a systematic review and economic evaluation to estimate the value of future research

Strategies for prophylaxis in women having antepartum risk assessment

The current NICE Guideline on the prevention of VTE in hospitalised women who are pregnant or who are in the puerperium recommends that clinicians use a tool published by a national UK body, professional network or peer-reviewed journal. 6 The NICE Guideline states that the most commonly used tool is the RCOG guideline;6,7 this is considered to represent current practice in the decision analysis. No data were identified to assess the sensitivity and specificity of RCOG in predicting VTE in women having antepartum risk assessment. (The only study assessing the use of RCOG in antepartum women was not suitable for inclusion in the modelling because it was in a small cohort of hospitalised pregnant women with cancer and no sensitivity data were available.) However, two RAMs for antepartum VTE risk assessment [Lyon32,33 and Efficacy of Thromboprophylaxis as an Intervention during Gravidity (EThIG)27] in women at high risk of VTE (prior VTE and/or thrombophilia) were identified in the systematic review (see Chapter 3). Therefore, in the high-risk antepartum population, the Lyon and EThIG RAMs are compared against each other and against strategies of prophylaxis for all and prophylaxis for none. Table 5 summarises current RCOG guidance on antepartum thromboprophylaxis for high-risk women (prior VTE and/or thrombophilia),7 and compares these with the two RAMs for high-risk patients. 27,32

As the RAMs vary in their recommendations regarding the timing of prophylaxis for some groups, the base-case analysis assumes that risk assessment occurs at the time of the antenatal booking appointment and LMWH is offered from booking to women identified as being high risk using the RAM. Scenario analysis is then used to explore whether the conclusions are sensitive to prophylaxis being deferred to 28 weeks. In scenarios where antepartum prophylaxis is offered, it is assumed that prophylaxis is also continued for 6 weeks after delivery.

In the high-risk antepartum scenario, the model estimates outcomes for prophylaxis for the following strategies:

• antepartum prophylaxis followed by postpartum prophylaxis for all [prophylaxis (PPX) from booking]

• antepartum prophylaxis based on a RAM (Lyon/EThIG)27,32 followed by postpartum prophylaxis for all

• postpartum prophylaxis for all but no antepartum prophylaxis [postpartum (PP) PPX only]

• no prophylaxis, either antepartum or postpartum (no PPX).

The exploratory analysis for unselected antepartum women makes similar assumptions regarding the timing of risk assessment (antenatal booking appointment) and the duration of prophylaxis offered to those identified as high risk (from booking until 6 weeks postpartum); however, the comparator strategy of postpartum prophylaxis for all (PP PPX only) is not included as unselected women not receiving antepartum prophylaxis are likely to receive a further risk assessment after delivery.

Strategies for prophylaxis in women having postpartum risk assessment

In the postpartum population model, the strategies compared are:

• postpartum prophylaxis for all (PP PPX for all)

• postpartum prophylaxis based on a RAM

• no postpartum prophylaxis (no PPX).

In each case, postpartum prophylaxis is assumed to be offered for 10 days. This is because for the majority of women receiving postpartum thromboprophylaxis, they would fit the criteria for short-term VTE prevention strategies based on their transient risk factors in line with the RCOG guidance. Extended postnatal prophylaxis lasting 6 weeks is mainly offered to those having antepartum prophylaxis, who are excluded from this analysis, and some women with multiple or persistent risk factors. For the unselected postpartum population, the RAMs compared are RCOG,11,39 SFOG,11,36 Caprini39 and the novel RAM reported by Sultan et al. 11 In the postpartum subgroups selected based on specific risk factors, the RAMs compared are RCOG and the novel RAM reported by Binstock et al. 28 in the post-caesarean section population and the novel RAM reported by Ellis-Kahana et al. 41 in the obese population.

Population characteristics

The average age in the cohort (30 years) is based on the mean age reported by Sultan et al. (2016) from a large UK longitudinal primary care database (CPRD). 11 The cohort included 433,353 women, without a history of VTE, whose pregnancy ended in a live birth or still birth between 1997 and 2014 and who had at least 6 weeks of postpartum follow-up. The average weight, which is required for estimating LMWH dosing, is based on the average BMI of 27.4 kg/m² for 25- to 44-year-olds reported in the 2019 Health Survey for England. 51 For the obese subgroup, we have assumed a BMI of 35.8 kg/m², based on the average BMI in the RAM study in an obese cohort reported by Ellis-Kahana et al. 41

Risk of venous thromboembolism in antepartum women with a prior venous thromboembolism

De Stefano et al. report 19 VTE events in 155 pregnancies where the women had a history of VTE prior to pregnancy but did not receive prophylaxis during pregnancy. 52 This gives an overall probability of 12.3% of having a VTE associated with the current pregnancy. The antepartum VTE risk was 5.8%, and the risk of postpartum VTE was of 6.9% (conditional on not having an antepartum VTE). Pabinger et al. reported similar VTE risks of 4% during pregnancy and 5% postpartum. 53 Brill-Edwards et al. reported a lower risk of VTE during pregnancy (2%), but their cohort excluded women with known thrombophilia and women could be recruited up to 20 weeks gestation, meaning that those having VTE early in pregnancy may have been excluded. 54 The risks from De Stefano et al. have been applied in the model. Higher and lower VTE risks have been explored in scenario analyses.

Risk of venous thromboembolism in unselected antepartum women

The risk of VTE in unselected antepartum women is based on the risk reported by Chauleur et al. in the cohort risk assessed using the STRATHEGE RAM. 31 There were nine VTE events in 2685 women (0.34%), of which four were antepartum and five were postpartum, giving absolute risks of 0.15% and 0.19% for antepartum and postpartum VTE, respectively. These data were only used in the exploratory analysis for unselected antepartum women.

Risk of venous thromboembolism in postpartum women

The risk of VTE in women following caesarean section has been estimated from an earlier analysis of the CPRD database (data from 1997 to 2010) reported by Sultan et al. (2014) in which women with prior VTE were excluded from the analysis. 55 For comparison, the risk of VTE within 6 weeks was 0.071% in this earlier cohort (158 in 222,334 deliveries) compared with 0.072% in the later cohort used to derive the Sultan RAM. 11,55 In this earlier study, the incidence of VTE within 6 weeks of any caesarean delivery was estimated to be 0.137% (74 VTEs occurring within 6 weeks across 31,843 emergency and 22,341 elective caesarean sections). 55 The risk of postpartum VTE within 6 weeks of delivery in women with obesity (BMI ≥ 30 kg/m²) was 0.153% (37 VTEs in 24,141 women). The risk of VTE over 6 weeks in unselected postpartum women was taken to be 0.072% based on the later study by Sultan et al. 11

Ratio of asymptomatic deep-vein thrombosis to symptomatic deep-vein thrombosis in postpartum women

A review by Blondon et al. 56 examining the incidence of VTE following caesarean section or vaginal delivery identified six studies which screened women postnatally to identify asymptomatic DVT. Over the 6 studies, we identified 1 symptomatic and 4 asymptomatic cases in a combined cohort of 717 patients. 57–62 Therefore, a ratio of 4:1 is applied in the base case. All of the asymptomatic cases of DVT identified were distal calf DVTs. A zero rate of asymptomatic DVT is explored in scenario analyses as the clinical significance of asymptomatic distal calf DVTs is unclear.

Proportion of venous thromboembolism that is deep-vein thrombosis without pulmonary embolism

The proportion of symptomatic VTE that is PE compared with DVT without PE has been estimated from studies included in the systematic review by Meng et al., which reported the incidence of PE and DVT without PE (24% of VTE is PE based on ratio of 17,035 DVT without PE to 5401 PE). 5 The review included both antepartum and postpartum VTE and the same ratio is applied to both the antepartum and postpartum incidences of VTE.

Proportion of deep-vein thrombosis that is distal

Data from the Computerized Registry of Patients with VTE (RIETE) were used to determine the proportion of symptomatic DVTs that are distal versus proximal. RIETE is an ongoing prospective registry of patients with objectively confirmed VTE and Elgendy et al. describe clinical characteristics for the subset of women who were pregnant or postpartum (within 2 months of delivery) at the time of VTE presentation. 63 Elgendy et al. report that 71% of postpartum DVTs (215 of 301) were proximal, whereas 78% of antepartum DVTs (342 of 438) were proximal. 63

We assumed that all asymptomatic DVTs are distal as none of the asymptomatic DVTs identified through systematic screening of postnatal women in the six studies described in section Ratio of asymptomatic deep-vein thrombosis to symptomatic deep-vein thrombosis in postpartum women were proximal.

Relative risk of venous thromboembolism in women having antepartum prophylaxis

The relative risk (RR) of symptomatic VTE for antenatal LMWH (with or without postnatal prophylaxis) compared with no prophylaxis is reported as being 0.39 (95% CI 0.08 to 1.98) based on four randomised controlled trials (RCTs) included in the updated Cochrane review by Middleton et al. 8 However, three of the RCTs included in this meta-analysis were considered by our clinical experts to be less applicable to the modelled population of high-risk women with a prior VTE. Two of the papers related to the LMWH (FRagmin^®) in pregnant women with a history of Uteroplacental Insufficiency and Thrombophilia (FRUIT) trial, which aimed to investigate LMWH combined with aspirin to prevent recurrent early-onset pre-eclampsia. This trial specifically excluded women at high risk of VTE due to prior history of VTE. 64,65 The third study was the Thrombophilia in Pregnancy Prophylaxis Study (TIPPS), in which LMWH was not given specifically for the indication of reducing VTE risk and less than half of the cohort had risk factors for VTE. 66 The dose of LMWH used in the TIPPS study was also higher than recommended for prophylaxis by RCOG. 7,66 The remaining RCT by Gates et al. was the only study included in the previous Cochrane review, and this had a RR of 0.33 (95% CI 0.02 to 7.14). 67 It should be noted that this was in fact a pilot study and the numbers recruited were small (n = 8 in each arm), and only one VTE event was observed, hence the wide CIs. The RR from this single pilot RCT was used in the base-case analysis due to the indirectness of the populations recruited in the FRUIT and TIPPS RCTs and also because of concerns regarding the dose used in the TIPPS study and the use of aspirin in the FRUIT study. However, scenario analysis was conducted using the meta-analysed estimate from all four papers reported by Middleton et al. 8

Relative risk of venous thromboembolism in women having postpartum prophylaxis

The updated Cochrane review reports a RR for VTE of 2.97 (95% CI 0.31 to 28.03) for LMWH versus no prophylaxis following caesarean section based on two studies. 67,68 Both of these were pilot studies. The dose used in one was lower than recommended in the RCOG guideline [2500 IU of dalteparin (Fragmin, Roche) daily for 4–5 days]. 68 For symptomatic DVT, an estimate of 1.40 (95% CI 0.17 to 11.55) is reported by Middleton et al. 8 based on two RCTs. 68,69 Two feasibility studies on postnatal prophylaxis in higher-risk postpartum women (low-risk thrombophilia, immobilisation or two or more risk factors) were also included in the updated Cochrane review (Rodger et al. 2015, 2016),18,19 but neither of these reported any VTE outcomes and both struggled to recruit. There is therefore a paucity of data on the efficacy of LMWH when used as postpartum prophylaxis and those studies that do exist estimate a higher risk of VTE compared with no LMWH, which is the opposite of what is expected based on studies in medical (RR = 0.49, 95% CI 0.37 to 0.67) and surgical cohorts [odds ratio (OR) = 0.26, 95% CI 0.09 to 0.87]. 47 To conduct EVPI analysis to estimate the value of future research, it is necessary to have some prior estimate of treatment efficacy even if that is based on indirect sources or expert consensus. In order to capture both our experience from other populations that LMWH is expected to reduce VTE, and the high degree of uncertainty in the efficacy of LMWH when used as postpartum prophylaxis, we decided to use the RR applied for antepartum prophylaxis (0.33, 95% CI 0.02 to 7.14).

It is unclear whether giving 10 days of LMWH provides protection from VTE for 10 days or for a longer period. Studies in general medical and surgical patients usually involve patients being offered LMWH during admission, or for a defined period such as 7 or 10 days and then they report the RR for VTE over a longer period such as 90 days post admission. Therefore, in these studies, the RR attributed to a short period of LMWH has been estimated over a longer time. Therefore, in previous models of thromboprophylaxis in medical and surgical in patients, the RR estimated from the meta-analyses of RCTs has been applied to the whole period at risk. 47 However, in this case, the RR has been taken from a study of antenatal prophylaxis in which LMWH was continued over the whole period at risk. It is therefore unclear whether the RR estimated in this setting should be applied to the whole 6 weeks over which patients are at risk of VTE, or just to the 10 days during which they received treatment. It was considered that giving 10 days of postpartum thromboprophylaxis would provide a risk reduction of VTE for longer than 10 days. This is because the development of clots occurs in the early postpartum period but may present symptomatically after 10 days, but not beyond 6 weeks postpartum. Given this uncertainty, we have assumed in the base-case scenario that risk falling in the first 3 weeks has the full treatment effect and risk falling beyond this has no treatment effect. This gives an average RR of 0.53 across the 6 weeks when applying a RR of 0.33 to risk falling in the first 3 weeks and a RR of 1 to risk falling from then up to 6 weeks. We have also conducted scenarios exploring the two extreme scenarios of having the RR apply to all 6 weeks and only the first 10 days. The proportion of the 6-week VTE risk falling within each time frame was estimated from the data provided by Sultan et al. 55 The RRs applied when assuming that the efficacy applies for 3 weeks (base case), 10 days (pessimistic scenario) and 6 weeks (optimistic scenario) are 0.53, 0.69 and 0.33, respectively.

Risk of major bleeding in women having antepartum and postpartum prophylaxis

A paper by Nelson-Piercy et al. reports the incidence of serious antepartum bleeds within their reporting of adverse events in a cohort of women having antenatal tinzaparin (Innohep^®, LEO Pharma). 70 Serious was defined as ‘clinical events that: resulted in death; were life-threatening; required inpatient hospitalisation or prolongation of existing hospitalisation; resulted in persistent or significant disability/incapacity; were congenital anomalies/birth defects; were other medically important conditions’. The incidence was 3 in 1267 (0.24%), but this included 1013 women having LMWH as prophylaxis and 254 having LMWH as treatment for VTE. 70 Therefore, this may overestimate the risk of major antepartum bleeding for prophylaxis doses of LMWH as some women were having higher doses of LMWH for VTE treatment. All three of the serious bleeds were recorded as possibly, but not probably, related to LMWH. In a UK cohort, reported by Schoenbeck et al., one of the 91 women who received both antepartum and postpartum prophylaxis experienced major obstetric haemorrhage (placental abruption requiring caesarean section at full term) giving a major antepartum bleeding risk of 1%. 71 In contrast, Cox et al. reported four severe antepartum bleeds requiring urgent delivery in 98 pregnancies (4.08%) exposed to LMWH in a New Zealand cohort. 72 Therefore, the risks of antepartum major bleeding appear to vary greatly (0.24–4.08%). Some of this variation is likely to be due to inconsistent definitions of what constitutes major bleeding and some due to differences in the cohorts of women described.

Tardy et al. conducted a review of RCTs of pregnant women having heparin to identify how RCTs have reported bleeding complications from heparin in the past. 48 Tardy et al. conclude, ‘at present it is impossible to estimate the rates and severities of either antepartum bleeding or primary PPH occurring during prophylactic treatment with heparin’. 48 They propose a definition for major bleeding in antepartum women that includes both the standard definition for major bleeding applied in medical inpatients and risks specific to pregnancy. 48,73 Although the definition proposed by the ISTH includes outcomes such as placenta previa requiring delivery and placenta abruption, it is unclear if these are likely to be causally related to the use of LMWH. Tardy et al. also state that antepartum bleeding due to placenta previa and placenta abruption is observed in 2–5% of all pregnancies in the absence of thromboprophylaxis. 48 We have applied the risk of serious antepartum bleeding from Nelson-Piercy et al. in the base case (0.24%), because it has been estimated in a large cohort of women receiving antenatal LMWH (N = 1267) and the definition of serious antepartum bleeding is specified. 70 However, given the uncertainty associated with this parameter, a range of major antepartum bleeding risk up to 4.08% is explored in sensitivity analysis to determine the significance of this parameter for decision-making.

The rate of postpartum major bleeding in women having both antepartum and postpartum prophylaxis has been estimated from a cohort reported by Schoenbeck et al. 71 In this cohort, that 5 of 91 women (5.5%) having antepartum prophylaxis followed by postpartum prophylaxis required a postpartum transfusion due to bleeding. This is higher than the risk of serious postpartum haemorrhage (PPH) (3.2%) reported by Nelson-Piercy et al., which was estimated across both treatment and prophylaxis doses of LMWH. 70 Nelson-Piercy et al. also report that 2.9% of women having prophylaxis doses of LMWH had bleeding requiring intervention, although this is not divided into antepartum and postpartum bleeding. 70 Given that postpartum bleeding requiring transfusion clearly meets the ISTH criteria for major PPH, we have applied the risk of 5.5% from Schoenbeck et al. in the base case, but higher and lower bleeding risks have been explored in sensitivity analysis.

Relative risk of major bleeding in women having antepartum prophylaxis compared to no antepartum prophylaxis

The only RCT reporting the RR of major bleeding in women having antepartum prophylaxis was the TIPPS study (RR = 1.48, 95% CI 0.25 to 8.72). 66 However, this population was considered somewhat indirect because not all women were included due to their VTE risk, and the dose used in this study was higher than the dose usually given to prevent VTE (women had the RCOG recommended dose of 5000 IU of dalteparin daily from randomisation until 20 weeks gestation but the dose was then doubled to 5000 IU twice daily until at least 37 weeks). 66 Furthermore, the timing of the major bleeding was not reported in the TIPPS study, so it is unclear if the bleeding occurred during the antepartum prophylaxis or during the postpartum period when both study arms received LMWH for 6 weeks. The RR of major bleeding in medical patients has been previously reported as 1.53 (95% CI 0.80 to 2.92), based on a meta-analysis of three RCTs. 6 Although it is expected that the absolute risk of bleeding is likely to differ between pregnant women and general medical inpatients, it was decided that the RR of bleeding from the medical cohort could be applied to women having antepartum prophylaxis, as the dose of LMWH used in medical in patients is consistent with that recommended by RCOG. Use of the alternative estimate provided by the TIPPS study is explored in scenario analysis.

Risk of wound haematoma in women having antepartum and postpartum prophylaxis

Lindqvist et al. recorded a risk of 2.5% for haematoma in women who had a prior VTE who were having LMWH during and after pregnancy and a risk of 0.4% in controls not having LMWH (p < 0.001). 74 We have therefore assumed an increased risk of wound haematoma attributable to LMWH of 2.1% in women having antepartum and postpartum prophylaxis.

Risk of major bleeding in women having postpartum prophylaxis

Although the severity of PPH is often defined according to volume of blood loss, with major PPH usually defined as blood loss > 1000 ml,75 the definition of major bleeding for postpartum women proposed by Tardy et al. includes blood loss < 1000 ml leading to transfusion. 48 In addition, blood loss of > 1000 ml is not defined as major bleeding by Tardy et al. unless it is combined with the need for transfusion or other intervention (e.g. second-line uterotonics or surgical intervention). Therefore, we tried to identify a study which reported the incidence of major postpartum bleeding that is consistent with the definition proposed by Tardy et al. rather than one based purely on volume of blood loss.

The risk of major bleeding in women having postpartum prophylaxis following a caesarean section is taken from a study by Gizzo et al. 76 In this study, the incidence of bleeding requiring transfusion after starting prophylaxis (12 hours after caesarean section) was reported as 4.6% (16/349). In this study, haemoglobin levels both post caesarean section and pre transfusion are reported, along with units of blood transfused, and transfusion was considered necessary if the haemoglobin fell below 8 g/dl. Based on these data, it was considered that the outcome of requiring transfusion in this study was reasonably consistent with one of the definitions of major bleeding proposed by ISTH, which is transfusion of two or more units of whole blood or red cells to maintain a haemoglobin level > 7–9 g/dl. 48,76 However, it is acknowledged that there is significant uncertainty regarding the incidence of PPH meeting the ISTH proposed definition and therefore both higher and lower incidence of major bleeding are explored in sensitivity analysis.

Relative risk of major bleeding for postpartum prophylaxis compared to no postpartum prophylaxis

There were no useful data on major bleeding in women having postpartum prophylaxis following a caesarean section from studies included in the Cochrane review. 8 Although some data were reported for bleeding-related adverse events, none of the studies reported any cases of major bleeding using a definition consistent with that proposed by the ISTH. The open-label pilot study by Rodger et al. (2016) reported one major bleeding episode in the LMWH arm and none in the prophylaxis arm giving a RR for major bleeding of 3.53 (95% CI 0.15 to 81.11). 19 However, the wide CIs produced by this single-pilot RCT were considered to be unrepresentative of the broader evidence on the safety of LMWH in indirect populations (i.e. medical and surgical patients who are not pregnant or in the puerperium). Therefore, the RR of major bleeding for medical inpatients receiving LMWH for VTE prophylaxis was used in the base-case analysis and the data from Rodger et al. were applied in a sensitivity analysis.

Risk of wound haematoma in women having postpartum prophylaxis

Ferres et al. reported an incidence of wound haematoma of 1.7% (11/653) in women at high risk of VTE after caesarean section who had received enoxaparin (Clexane^®, Sanofi) and an incidence of 1.1% in those who were eligible but who were not offered enoxaparin (11/1042). 77 Although this difference was not statistically significant, we were keen to capture the potential for increased wound haematoma as this was considered an important side effect of LMWH for women. We therefore assumed in the base case that postpartum prophylaxis increases the risk of wound haematoma by 0.6%.

Risk of fatal bleeding and non-fatal intracerebral haemorrhage

From the report published by MBRRACE-UK, we estimated that the rate of fatal bleeds, including either obstetric haemorrhages or ICH, was 0.53 per 100,000 maternities in the antepartum period and 0.57 per 100,000 maternities in the postpartum period. 78

The incidence of ICH has been estimated from a study by Ban et al. which used routine hospital (HES) and GP records (CPRD) to estimate the incidence of strokes in the antepartum, peripartum (1 day prior to 1 day after delivery) and early postpartum (within 6 weeks of delivery) periods. 79 This study reported the incidences separately for ischaemic stroke, ICH, subarachnoid haemorrhage and unspecified strokes. We used the data reported to estimate an incidence of 0.9 per 100,000 pregnancies and 1.1 per 100,000 pregnancies for non-fatal ICH in the antepartum and postpartum groups, respectively. As it is unclear whether the risk of fatal bleeding and non-fatal ICH is causally linked to women receiving LMWH as prophylaxis for VTE, we have also conducted a scenario analysis in which these outcomes are excluded from the model.

Risk of major bleeding during treatment for venous thromboembolism

The Global Anticoagulant Registry in the FIELD (GARFIELD) and RIETE registries report subgroup analyses for women who are pregnant or postpartum at the time of their acute VTE diagnosis. 63,80 They report the incidence of major bleeding during treatment for pregnancy-associated VTE. The RIETE registry provides a larger sample than the GARFIELD registry. The incidence of major bleeding reported in the RIETE registry is 0.8% for women having pregnancy-related VTE and this has been applied in the model. 63

No fatal bleeds on VTE treatment doses are reported for the postpartum VTE subgroup in the RIETE cohort and only one was reported in the pregnancy-associated VTE subgroup of the RIETE cohort. 63 However, data are also presented for a subgroup of younger women (aged under 50) with non-pregnancy-associated VTE who were used as a comparative cohort for the pregnant/postpartum subgroups. As this provided a larger sample size (N = 8084), the data from the non-pregnant cohort were used to estimate the proportion of major bleeds that are fatal (6.3% = 4/63). 63 When combined with the 1.1% absolute risk of major bleeding, this gives an absolute risk of fatal bleeds during VTE treatment of 0.07% (7 in 10,000), which is similar to the lowest risk category for fatal bleeding identified from a separate analysis of the whole RIETE cohort. 81 There are two limitations with this: the non-pregnant cohort diagnosed with VTE may have other risk factors for VTE that place them at increased risk of death such as cancer; and there is a higher rate of direct oral anticoagulant (DOAC) use for long-term anticoagulation in the non-pregnant cohort (9.4% for non-pregnancy-related VTE vs. 4.7% for postpartum VTE). 63 Given the uncertainty regarding the risk of fatal bleeding during treatment of pregnancy-related VTE, we have conducted a scenario analysis exploring a zero rate of fatal bleeding to see how important this parameter is to the decision analysis.

Data on the incidence of non-fatal ICH for VTE treatment doses of LMWH for women having pregnancy-associated VTE are sparse. The RIETE registry does not report the site of major bleeding. 63 The GARFIELD registry does report the site of any major or minor bleed and none of these are reported to be intracranial, although 11 bleeds in the non-pregnancy-associated VTE group and 1 in the pregnancy-associated VTE group are reported to as being ‘other’. 80 The GARFIELD registry does report 2 events as stroke/transient ischaemic attack in 29 women having major bleeding during treatment of non-pregnancy-related VTE, but these could be ischaemic events. 80 If we assume that half of these strokes were haemorrhagic events,79 then the proportion of major bleeding events in GARFIELD which were ICH would be 3.4% (= 1/29) in the non-pregnancy-associated VTE subgroup. 80 This is slightly lower than the proportion of non-fatal major bleeding events which are ICH in the RIETE registry as a whole (9%). 81 This gives an absolute risk of non-fatal ICH on treatment doses of LMWH of 0.04% (4 in 10,000), which is applied in the base-case analysis. A scenario analysis exploring a zero rate of non-fatal ICH is also explored to determine how important this parameter is to the decision analysis.

Risk of chronic complications

The risk of CTEPH in patients having PE was taken from a systematic review by Ende-Verhaar et al. 82 A cumulative incidence of 3.2% (95% CI 2.0 to 4.4) over 2 years was reported in those who survived the initial 3-month period after PE and this was applied to the model giving a risk of 1.6% per annum. We assumed no risk of new CTEPH beyond 2 years based on a study with a median follow-up of 94 months, which reported no new cases of CTEPH after 2 years. 83

A study describing the risk of PTS following VTE related to pregnancy was identified from a review by Kourlaba et al. 84 This paper by Wik et al. reported that the risk of PTS, defined as a Villalta score ≥ 5 was 42% following DVT. 85 Wik et al. also report that age, smoking, timing of VTE (postnatal rather than antenatal) and location (proximal rather than distal) were significant predictors of the risk of PTS in pregnancy-related VTE. Wik et al. also found a significant interaction between timing and location of DVT such that proximal DVT occurring postpartum gave the highest risk of PTS. However, this association between timing and location did not occur in women having antenatal DVT. We, therefore, used the data from Wik et al. to estimate separate risks of PTS for antenatal DVT (34%), distal postpartum DVT (31%) and proximal postpartum DVT (66%). In the base-case scenario, asymptomatic DVTs are assumed to carry the same risk of PTS as symptomatic DVTs, but in a scenario analysis we have assumed no risk of PTS from asymptomatic DVTs. The proportion of cumulative PTS risk falling in years 1–5 after DVT was based on a study by van Dongen et al. which followed up patients with a DVT (not specifically pregnancy-related DVT) every 6 months for a maximum of 5 years (median follow-up 4.9 years) to assess them for signs and symptoms of PTS. 86

To explore how sensitive the model is to the risk of PTS, we have also conducted a scenario analysis in which the risk of PTS (15.6% and 32.4% in distal and proximal DVT, respectively) were estimated from a non-pregnant population. 46

Mortality risks

Patients who survive the first 28 days after having an ICH have an increased risk of mortality in the long term. Fogelholm et al. report a standardised mortality ratio (SMR) of 4.5 in the first year and 2.2 in years two to six for patients who survived 28 days after ICH, compared to age- and sex-matched controls. 87 These SMRs have been applied as multipliers to the risks of all-cause mortality in women having non-fatal ICH. This means that women having ICH have an increased risk of death in the long term compared to women not having an ICH, but the absolute risk of death in women having ICH is much lower than in the study reported by Fogelholm, where the average age was > 65 years in both men and women recruited to the study. Given that the SMRs have been calculated in a predominantly older cohort, and the relationship between ICH and increased mortality may not translate to younger patients, we have conducted a scenario analysis in which women who have a non-fatal ICH (i.e. survive 28 days after an ICH) do not have any increased mortality in the long term.

A review by Kourlaba et al. reports a case fatality rate for PE of 2% (95% CI 1.44 to 2.56), which was based on a meta-analysis of four studies. 84 This estimate was largely dependent on the rates from two large studies based on discharge records which reported case-fatality rates of 1.73% (Liu et al.) and 2.43% (James et al. ), respectively. 3,88 Case-fatality rates from registry studies such as RIETE were lower (< 1%),63 but these may be biased because women who die shortly after PE would not necessarily be recruited into registry studies of anticoagulant treatment. A case-fatality rate of 2% has therefore been applied in the model.

Mortality risks in patients with CTEPH having either medical or surgical management were based on survival curves reported by Goodacre et al., which were estimated from an international prospective registry of patients with CTEPH. 89 Deaths related to PE occurring within 1 year of PE are already accounted for in the model. 90 For this reason, the hazard of death for patients with CTEPH are only applied from 1 year onwards. To ensure that the risk of death in the CTEPH group was not artificially low compared with the risk of death in the general population, general population mortality risks were applied whenever these were higher than the risk in the CTEPH population, based on the survival curves.

All-cause mortality is estimated from age- and sex-specific general population mortality estimates and is applied to all women not experiencing CTEPH, ICH, fatal bleeds or fatal PE. 91 These have not been adjusted to account for any increased risk of mortality associated with pregnancy.

Cost of prophylaxis

We have assumed that the pharmacological prophylaxis used is LMWH as the use of oral anticoagulants including vitamin K antagonists, such as warfarin, and DOACs should be avoided in women who are pregnant. 7 Although warfarin can be used postpartum in women who are breastfeeding, clinical expert advice was that most women preferred to use LMWH postnatally due to the monitoring requirements associated with warfarin.

Women having antepartum prophylaxis are assumed to receive prophylaxis from 10 weeks as this is the typical time of the antenatal booking appointment. It is acknowledged that some women with a prior VTE may be on anticoagulant treatment prior to pregnancy and may immediately switch to LMWH as soon as pregnancy is confirmed, and some women may take longer than 10 weeks to have their risk assessed and start prophylaxis, but 10 weeks was considered a reasonable average starting time. Prophylaxis started antenatally is assumed to continue until 6 weeks after delivery.

As some women are recommended to have antepartum prophylaxis only from 28 weeks under the RCOG guidance, a scenario analysis exploring the impact of delaying antepartum prophylaxis until 28 weeks in this population is described in the section Exploratory deterministic analysis for antepartum women with three risk factors. In addition, the Lyon score recommends that some women having antepartum prophylaxis from booking and some from 28 weeks. Therefore, a scenario analysis exploring the impact of delaying prophylaxis to 28 weeks is described in the section Deterministic scenario analyses for antepartum women with a prior venous thromboembolism.

Women starting prophylaxis postnatally are assumed to be offered 10 days of postpartum prophylaxis in line with current RCOG guidance for those at intermediate risk. Six weeks of postpartum prophylaxis is only offered to those at high risk of VTE, many of whom will already be identified as requiring antepartum prophylaxis.

Patients having a major bleed while taking antenatal prophylaxis would be likely to stop antepartum prophylaxis and restart postnatally. We assumed a 66% reduction in their antepartum prophylaxis cost in such cases. The cost of postpartum prophylaxis is not assumed to reduce if a major bleed occurs postnatally, because even if treatment is discontinued, it is likely that the rest of the 10 days or 6 weeks course would be wasted.

It is assumed that the lowest cost preparation of LMWH is prescribed. Therefore, drug costs for LMWH were based on the cost of dalteparin (5000 units given every 24 hours by subcutaneous injection, based on a weight of 73 kg) as this had a lower cost per day (£2.82) than either enoxaparin or tinzaparin (based on NHS drug tariff prices for lowest cost preparation). 92 The cost was increased to £4.23 per day in the obese cohort (7500 units per day for a 95 kg woman). 92 The cost for administration of LMWH has been estimated by adjusting the costs estimated by Menakaya et al. for outpatient LMWH. 93 This includes the cost of counselling women to self-inject when they first start prophylaxis with LMWH and the cost of administration by a district nurse in the small minority of women (4%) who are unable to self-inject. The cost of administration is £74.94 for 10 days of postnatal LMWH and £321.64 for those starting LMWH antenatally and continuing until 6 weeks after delivery (see Appendix 4, Table 23). In addition, for women having antepartum prophylaxis, it is assumed that they will receive one additional outpatient appointment (£205 for a multiprofessional face-to-face follow-up appointment)94 in late pregnancy to discuss the need to discontinue LMWH at the onset of labour or prior to any planned delivery.

In the strategies examining the use of prophylaxis based on RAMs, we have assumed that the risk assessment will require 5 minutes of time spent by a hospital consultant, which is a cost of £9.92 when applying a unit cost per hour of £119. 45 This cost is applied to all strategies involving the use of a RAM, but not to the comparator strategies of prophylaxis for all and prophylaxis for none. A sensitivity analysis has also been conducted to explore whether the optimal strategy would be different if it could be assumed that the risk assessment would result in no additional cost, with the aim of exploring whether the cost of the risk assessment itself is a significant source of decision uncertainty.

Costs of treating venous thromboembolism

Treatment for VTE (PE or DVT) in the model is assumed to consist of 3 months of LMWH, because warfarin and DOACs are not safe during pregnancy. Although warfarin is a possible alternative treatment for postnatal women, it is less commonly used because of the need for regular blood tests and many mothers would prefer subcutaneous injections without monitoring. Equally, although DOACs can be offered postnatally, these are limited to women who are not breastfeeding. The recommended doses for each of the three available LMWHs (enoxaparin, dalteparin and tinzaparin) have been based on Table 1a–c of the RCOG guideline on the acute management of thromboembolic disease (Green-top guideline no. 37b). 75 A weight of 73 kg has been assumed based on the average weight of women aged 25–44 years in the 2019 Health Survey for England. 51 Where the dose required does not match exactly one of syringe sizes provided by manufacturers, it is assumed that the next size up is used and any excess is discarded so that a new syringe is used each time. The proportion of patients receiving each of the three LMWH preparations (enoxaparin 65.2%, dalteparin 21.6%, tinzaparin 13.2%) was based on data from a recent survey of clinicians by McFarlane et al.,95 which also reported that 56.5% used once daily rather than twice daily dosing. This was used to estimate the proportion using once daily dosing (51%) for dalteparin and enoxaparin after accounting for the fact that tinzaparin is only recommended for once daily dosing. The total cost for drug acquisition for 91 days of treatment is £887.21. This was increased to £1155.32 when using the higher weight of 95 kg for the obese postpartum population.

Women having VTE treatment with LMWH who experience major bleeding are assumed to stop LMWH while actively bleeding, but it is assumed that treatment dose LMWH will be started as soon as it is safe to do so. Therefore, no reduction in VTE drug treatment cost is assumed in the base-case analysis. However, a scenario analysis is conducted assuming that LMWH is stopped for 4 weeks to see whether this factor is an important driver of cost-effectiveness. This reduces the costs of VTE treatment by 11% for those having antepartum VTE and by 18% for those having postpartum VTE.

The costs for administering LMWH, including the cost of training patients to self-administer and the cost of administration by a district nurse for a small minority of patients (4%), are based on the costs reported by Menakaya et al.,93 but these were adjusted to reflect the longer duration of treatment (91 days compared with 42 days in Menakaya), giving a cost of £157.51 for administration. In the base-case analysis, it is assumed that women having treatment dose LMWH for either antepartum or postpartum VTE will have monthly joint outpatient clinic appointments with a haematologist and obstetrician while receiving treatment dose LMWH [£205; Healthcare resource group (HRG) WF02A Service code 303]. 94 However, it was noted that many women experiencing antepartum VTE will already be having regular clinical appointments to manage the comorbidities that put them at increased risk of VTE and not all of the monthly appointments will be required solely due to the LMWH treatment. To explore this, a scenario analysis was conducted where it was assumed that women having antepartum LMWH have three additional clinic appointments and those having postpartum LMWH have one additional clinic appointment. The total cost of drug treatment for VTE is £1659.79 for those having postpartum VTE and £2748.29 for those having antepartum VTE (see Appendix 4, Table 24). This is because women having antepartum VTE are assumed to experience VTE on average at 24 weeks gestation (see section Timing and duration of utility decrements applied in the decision tree), resulting in 154 days of VTE treatment.

Resource use for management of acute VTE is provided in a summary table (see Appendix 4, Table 25). The costs of diagnosing DVT and PE in patients having these outcomes postnatally have been based on the costs used previously in non-pregnant populations having DVT or PE in an outpatient setting. 46 These previous analyses assumed that 10% of patients having proximal DVTs and 60% of patients having PEs would be admitted but none having distal DVTs would be admitted. In addition, it was assumed that 10% of those having PE would be admitted to critical care. For women having an antenatal DVT or PE, we would expect a greater likelihood of admission compared to a non-pregnant population. We have assumed that the likelihood of admission for proximal DVT is double that assumed in the non-pregnant population (20% vs. 10%) but remains zero for distal DVT. We have assumed that the likelihood of admission increases from 60% to 90% in women having antenatal PE. We have also assumed that the risk of admission to critical care is double for pregnant women having a PE compared to non-pregnant people having a PE (20% vs. 10%). We assumed maximum resource use for people having fatal PE (i.e. ambulance transfer to the ED leading to a short-stay admission including a critical care unit stay) but have excluded the long-term cost of VTE drug treatment.

For patients having antenatal PE, we have taken the diagnostic tests used from the Diagnosis of Pulmonary Embolism in Pregnancy (DiPEP) study. 96 We have assumed a 50 : 50 split between ventilation/perfusion (V/Q) single photon emission computed tomography (SPECT) and V/Q planar as Goodacre et al. only report the frequency of ventilation-perfusion scanning and not the specific type. As the DiPEP study did not report on the use of echocardiogram, we have taken the proportion from a previous UK Obstetric Surveillance System (UKOSS) data set reported by Knight et al. 97 (Appendix 4, Table 25). The total cost for acute management of VTE (i.e. excluding long-term drug treatment) ranged from £311.32 for symptomatic distal DVT to £3261.24 for fatal PE (details provided in Appendix 4, Table 25). Given that many assumptions have been employed to estimate resource use associated with diagnosis and management of VTE, the importance of these costs is explored in a scenario analysis. In this scenario, we assume that all antenatal VTE results in admission and 50% of antenatal PE also results in a critical care stay.

Costs of treating major bleeding

The cost over 90 days of fatal haemorrhagic stroke provided by Luengo-Fernandez et al. was uplifted to current prices and applied as the cost of fatal bleeds in the decision-tree phase of the model. 98 This paper also provided costs over 90 days for non-fatal haemorrhagic stroke stratified by the level of disability. A weighed average cost was calculated across non-disabling, moderately disabling and totally disabling haemorrhagic strokes. 98 This was then uplifted to current prices and applied as the cost of non-fatal ICH in the decision-tree phase of the model. Luengo-Fernandez et al. also report the average costs per annum from 90 days to 5 years post stroke, but these are not reported separately for haemorrhagic stroke. The costs of GP care and emergency care are reported to be statistically significantly higher post stroke compared to the year before stroke. In addition, they report the cost of residential care in patients not living in residential care prior to their stroke. The total post-acute (beyond 90 days) costs for primary care, emergency care and residential care were calculated and uplifted to current prices and applied in the state-transition phase of the model to those having non-fatal ICH. A pro rata cost is also applied to those having stroke more than 90 days before the end of the decision-tree phase of the model.

There are limited data available to determine the cost of managing non-fatal, non-ICH bleeds in pregnant women. The cost of managing non-fatal non-ICH major bleeds was assumed to be similar to the cost of managing a gastrointestinal (GI) bleed, despite these being different pathologies. This is consistent with the approach that has been taken in previous published models of VTE prevention covering both patients having outpatient lower limb immobilisation and patients at risk of hospital acquired VTE. 46,47 It was also the approach taken in a US economic evaluation of heparin use in pregnant women reported by Johnston et al. (2005). 99 The cost of GI bleeding was estimated based on a weighted average cost for non-elective inpatient and non-elective short-stay management of GI bleeds using NHS reference costs for bleeds requiring single, multiple or no interventions (HRG codes FD03A to FD03H). 94 The average length of stay was 3 days for the most common HRG code for GI bleeding which comprised 46% of the spells for GI bleeding, suggesting that the median length of stay is around 3 days. Costs estimated using GI bleeding as a proxy are probably more applicable to women having a major non-obstetric bleed, where the bleed itself would be the reason for admission. This is because it is assumed that many women having major bleeding at the time of delivery (i.e. a PPH) are likely to already be receiving inpatient care. A cost-effectiveness model examining the use of uterotonics to prevent PPH estimated that the average length of stay would be 1.5 days in women without PPH (< 500 ml blood loss), increasing to 3 days in women who have > 1500 ml of blood loss, who are also assumed to require transfusion of two units of blood,100 in which case the additional length of stay attributable to major PPH would be closer to 1.5 days. An international survey of midwives (N = 100) estimated that women having a major PPH would have an increased length of stay of 1 day compared to women not having a PPH. However, when UK-specific midwives were questioned (N = 25), it was suggested the estimated additional length of stay attributable to a major PPH would be longer at 2.3 days. 101 Based on these data, it is possible that the cost of an admission for GI bleeding is higher than the additional cost of managing a major PPH in a woman already admitted for delivery. Given the likely heterogeneity associated with the costs of major bleeding in this population, we have conducted a scenario analysis in which we assume no additional cost, and a scenario analysis in which we assume the cost is twice that expected for GI bleeding, to determine how sensitive the results are to this parameter.

Costs of wound haematoma

Wound haematomas can lead to a delay in discharge for women after delivery. Therefore, we assumed that a wound haematoma would result in a long-stay admission instead of a short-stay admission using the reference costs for normal delivery (cost difference of £1372 between non-elective and short-stay admission NZ30C). 94 We explored a more conservative scenario in which a wound haematoma only leads to one ED attendance in a scenario analysis.

Costs of managing post-thrombotic syndrome and chronic thromboembolic pulmonary hypertension

The management of PTS is assumed to involve one first and one follow-up vascular surgery outpatient appointment in the first year after diagnosis and two follow-up GP appointments every year thereafter. This is consistent with the assumption applied in a previous analysis. 46 An alternative cost based on the burden estimated in a US cohort is considered in a sensitivity analysis. 102

Drug costs for medical management of CTEPH were based on the costs used in CG92, which were uplifted to give a cost of £18,980 per year. The costs for medically managed patients are applied each year to those surviving with CTEPH. The proportion of patients having surgical management for CTEPH (59%) is based on data from Delcroix et al. 90 The cost for surgical management of CTEPH is based on a weighted average of the reference costs for complex thoracic procedures (DZ02H/J/K) giving an average cost of £8175. 94 A proportion of patients having surgical management (29%) are assumed to require medical treatment as a bridging therapy (average of 4.6 months). Including these costs brings the total cost in the first year to £10,282 for surgical management. No costs are applied beyond the first year for those managed surgically.

Utility values

A recent systematic review of utility values by Etxeandia-Ikobaltzeta et al. was identified which included utilities values in studies published up to April 2018. 103 This review identified only one study of relevance to pregnant/postpartum women, but this study used the EuroQol-5 Dimensions (EQ-5D) visual analogue scale (VAS) which is not a preference-based measure of utility. Estimates of utility following DVT and PE in the general population from the Prevention of Thromboembolic Events – European Registry in VTE (PREFER-VTE) study (not specific to pregnancy or puerperium)104,105 have been used in a previous model of thromboprophylaxis in hospitalised patients. 46,47 As no additional utility values were identified from the review by Etxeandia-Ikobaltzeta et al., the utility values from the PREFER-VTE study applied in previous VTE prevention models were maintained in this model. These gave utility multipliers of 0.962 and 0.960 when averaging the reported utility values over the first 6 months after DVT and PE, respectively (see Appendix 4, Table 26), and long-term utility multipliers of 1.00 and 0.99, respectively (see Appendix 4, Table 27).

One of the key assumptions in previous models was that the utility decrement of PE and non-ICH major bleeding is similar in the month following these events. 46,47 This assumption is somewhat supported by Etxeandia-Ikobaltzeta et al., who found that the EQ-5D VAS score was 30 for both PE and major obstetric bleeding. 103 Therefore, it seemed reasonable to maintain this assumption in the model for pregnant women. However, it should be noted that the utility decrement for major non-ICH bleeding is applied for only 1 month, whereas an ongoing utility decrement is applied for PE, so PE has a larger impact on QALYs than major bleeding. Given the lack of utility values measured directly in women following major bleeding, a scenario analysis in which no utility decrement is applied for those having major non-ICH bleeding has been conducted to explore the importance of this parameter. For wound haematoma, we have assumed a utility decrement equivalent to major non-ICH bleeding for 1 week to capture any adverse impact on HRQoL. A scenario analysis removing this utility decrement for wound haematoma was also explored.

No pregnancy-specific estimates of utility following PTS or CTEPH were identified in the review by Etxeandia-Ikobaltzeta et al.,103 so the sources applied in previous models were maintained. 46,47 In our previous analysis of prophylaxis during lower limb injury, we used utility data from the Catheter-Directed Venous Thrombolysis in Acute Iliofemoral Vein Thrombosis (CaVenT) study to estimate the utility decrement in patients with PTS. 106 An estimate of 10% was applied from diagnosis onwards which was obtained from the CaVenT study by comparing the EQ-5D scores in those with and without PTS at 2 years. The CaVenT study did not stratify the utility estimates by severity of PTS, so this estimate was applied to all patients with PTS in the model regardless of severity. This may overestimate the utility decrement if the proportion of patients having severe PTS is lower in the modelled population than in the CaVenT study which recruited patients with acute iliofemoral DVT. A study by Lenert and Soetikno reported utility estimates for mild and severe PTS (0.98 and 0.93, respectively) obtained by using health state descriptions and a standard gamble valuation technique in a sample of volunteers. 107 These were not used in the base case as utility measured using the EQ-5D in patients with the condition is preferable to utility measured using standard gamble in volunteers based on descriptions of the condition. However, a scenario analysis was conducted in which the data from Lenert and Soetikno107 were combined with data on the proportion of PTS that is severe (6%) from a registry study in outpatients having VTE,108 to estimate a utility decrement of 2% across all patients with PTS. In the current analysis, we have taken the same approach and have applied a 10% decrement in the base case and a 2% decrement in a scenario analysis.

The utility decrement in patients with CTEPH was estimated from a study by Meads et al. by comparing the utility in patients having CTEPH (0.56) and the utility in patients with disease categorised as New York Heart Association (NYHA) class 1 (0.89) in which the HRQoL impact of symptoms would be expected to be minimal. 109 This gave a utility multiplier of 0.63 or a 37% decrement. This decrement is applied lifelong in the model to those having medical management of CTEPH, but only for 1 year in those having surgical management who have the utility multiplier for PE applied thereafter.

The values applied following non-fatal ICH were also taken from those used in previous VTE prevention models. Utility values following ICH were based on data from 5-year follow-up of the Oxford Vascular Study (OXVASC) study as these data were applied in a previous analysis of thromboprophylaxis after lower limb injury. 110 An absolute decrement of 0.22 was assumed in the decision-tree part of the model where time since stroke was < 1 year and a decrement of 0.09 was assumed in the long-term part of the model. This study was chosen as the source of utility values previously as the duration of follow-up allowed time since stroke to be accounted for, and a comparison was made against general population norms.

In the previous analysis of thromboprophylaxis after lower limb injury, we identified several sources which estimated the utility decrement associated with VTE prophylaxis or VTE treatment. 46 The study selected for use in the previous model was a study by Marchetti et al. which reported that patients would be willing to trade 2.7 of 365 days to avoid treatment with LMWH. 111 These data were previously used to estimate a utility decrement of 0.007 for LMWH. These same decrements have been applied in this model. However, it is noted that the utility decrement may differ for women during pregnancy or the puerperium. Therefore, to determine how sensitive the model results are to this parameter, we conducted scenario analyses in which we assumed that the utility decrement is either double the value assumed in the base case or zero.

Utility values for patients not experiencing any utility decrement due to prophylaxis, treatment, symptomatic VTE events, bleeding events (ICH or other major bleeds), long-term sequelae following VTE (PTS or CTEPH) or death are based on general population norms for a cohort of the same age and this is allowed to vary as the cohort ages during the model. 112

Timing and duration of utility decrements applied in the decision tree

To calculate the QALYs gained by patients having different paths through the decision tree, it is necessary to make some assumptions regarding the timing of events as these are not explicitly modelled in a decision tree. The average timing of postpartum VTE is 21 days based on the timings reported by Sultan et al. 55 However, we are interested in the timing of the VTE that is prevented by prophylaxis. If you assume that the VTE events prevented by prophylaxis occur on average halfway through the prophylaxis, then that would mean assuming they occur at 5 days. However, it is also possible that early prophylaxis for 10 days prevents VTEs that would have been diagnosed later in the puerperium after prophylaxis has ended. Therefore, in our base-case analysis, we have assumed that the VTEs being prevented occur on average at 21 days, and we explore the impact of varying this from 5 to 42 days in scenario analysis.

The average timing of VTE occurring during pregnancy was estimated using data from Voke et al. which provides a scatter plot of timing of VTE events. 113 From this it was estimated that the average timing of VTE was 24 weeks.

Timing of postpartum bleeds during prophylaxis (3 days) was based on the average timing reported by Gizzo et al. 76 Bleeds occurring during antepartum prophylaxis are assumed to occur 28 days before the timing of VTE. Bleeds occurring during treatment for VTE are assumed to occur at 13, 32 and 12 days post diagnosis of VTE for fatal, ICH and other major bleeds, respectively (based on data from the RIETE registry reported by Nieto et al. ). 81

We made the following assumptions when estimating QALYs in the decision tree:

• Baseline utilities using general population utility values for the starting age are applied to those not having treatment and not having any clinical events (e.g. VTE, bleeds).

• A disutility for ICH is applied lifelong, but separate values are applied in decision-tree and state-transition phases of the model.

• Disutility of non-fatal non-ICH major bleeding is assumed to last a maximum of 28 days.

• Disutility of prophylaxis applies for the duration of prophylaxis.

• Disutility of treatment for VTE applies for the duration of treatment.

We made the following assumptions when estimating QALYs in the state-transition model:

• Utility values for patents without any long-term sequelae (ICH, CTEPH, PTS) are taken from general population values and decrease as patients age in the model.

• All other utility values are applied as multipliers such that the absolute utility value decreases due to ageing in all patients.

• Utility decrements continue in the state-transition model for the remainder of the patients’ lifetime for PE but not for DVT where patients are assumed to return to general population utility values at 1 year.

• Patients with CTEPH who are treated medically have a lifelong utility decrement, whereas those treated surgically return after 1 year to the same utility as those surviving PE without CTEPH.

• Patients with PTS have the same utility decrement from diagnosis to death.

• Patients with ICH have the same utility decrement from the start of state-transition model to death.

Sensitivity and specificity of risk assessment models

Estimates of sensitivity and specificity are based on the data presented in the systematic review (see Chapter 3). Only those studies that reported both sensitivity and specificity could be included in the modelling and any study reporting a sensitivity of 0% was excluded from the modelling. In the antepartum model for high-risk women, only the EThIG and Lyon RAMs had data suitable for inclusion. The modelling for high-risk women assumes that patients classified as low risk by the EThIG and Lyon RAMs will receive postpartum prophylaxis and those categorised as high risk based on the RAMs will receive antepartum prophylaxis in addition to postpartum prophylaxis. Therefore, it is the sensitivity and specificity of the RAMs in predicting antepartum VTE that are relevant to the economic analysis. The data available for the high-risk antepartum women for the EThIG and Lyon RAMs are summarised in Figure 5 as a receiver operating characteristics (ROC) curve.

FIGURE 5.

Receiver operating characteristics curve for RAMs to predict antepartum VTE in high-risk women. (References for figure: Bauersachs 2007,27 Dargaud 2017. 32)

In the unselected postpartum population, data were available for the RCOG,11,39 SFOG11,36 and Caprini39 RAMs and the novel RAM reported by Sultan et al.,11 for which performance data were reported for multiple cut-offs (defined according to those falling in the top 1%, 5%, 10%, 20% and 25% of absolute risk). In the postpartum population with obesity, only the novel RAM reported by Ellis-Kahana et al.,41 is available, but performance data are provided for two versions of this RAM, in which thrombophilia was either included or excluded from the risk algorithm. In the post-caesarean section population, a novel RAM is reported by Binstock et al. along with data for the RCOG RAM. 28 The data for postpartum RAMs across the three populations are summarised in Figure 6.

FIGURE 6.

Receiver operating characteristics curve for postpartum VTE in unselected postpartum women and women with specific risk factors (obesity, caesarean section). (References for figure: Binstock 2019,28 Ellis-Kahan 2020,41 Lindqvist 2008,36 Tran 2019,39 Sultan 2016. 11)

In the unselected antepartum population, the only RAM identified with available performance data was the STRATHEGE RAM,31 but the data suggested that it had poor performance (sensitivity of 0% and specificity of 98%). Therefore, the analysis in the unselected antepartum population was limited to exploratory analysis. In this analysis, various theoretical combinations of sensitivity and specificity values were tested to determine the range of sensitivity and specificity values that would be required for a RAM to be cost-effective in this population.

Model inputs for secondary scenarios where antepartum prophylaxis is offered at 28 weeks

In the main analysis for antepartum prophylaxis in women at high risk of VTE, we have assumed that women are offered prophylaxis from booking if they are identified as being at high risk by either the EThIG or Lyon RAMs. However, for the Lyon RAM, some women are offered antepartum prophylaxis from 28 weeks only. 32,33 Therefore, we have conducted a scenario analysis where prophylaxis is only offered from 28 weeks in the high-risk subgroup to explore whether this results in a different strategy being most cost-effective. For this analysis, the efficacy of prophylaxis is adjusted so that it only applies to the 40% of antepartum VTE risk occurring after 28 weeks gestation and the average timing of VTE is moved from 24 to 34 weeks, with consequent impacts on the cost of VTE treatment. The average timing of VTE occurring after 28 weeks was based on the scatterplot provided by Voke et al. 113 Other risks, such as the risks of major antepartum bleeding, are not adjusted.

In the main analysis for women being offered antepartum prophylaxis, we have focused on women at high risk being offered prophylaxis from booking. However, within the RCOG guidance, antepartum prophylaxis is also recommended from 28 weeks in women having three risk factors. 7 A secondary scenario has been conducted to explore whether this is cost-effective. Some data on the risk of antepartum VTE for specific risk factors and for some combinations of risk factors were available from an analysis of a GP database by Sultan et al. (2013). 114 This used a different GP database (The Health Improvement Network) from that used to generate the postpartum RAM (CPRD),11 but again it excluded women with a prior history of VTE. This provided the risk of VTE for women with any 2 or more risk factors as being 95 antepartum events per 100,000 pregnancies and 111 postpartum events per 100,000 risk factors (0.20% VTE risk overall). 114 The risk for women with three or more risk factors is not provided; however, if we applied the RR for the strongest individual risk factor, which was varicose veins (RR of 2.21 for antepartum VTE and 3.90 for postpartum VTE), to the absolute risks for two or more risk factors, then this would suggest an upper limit for the absolute risk of 217 antepartum VTEs per 100,000 pregnancies and 433 postpartum VTEs per 100,000 pregnancies. This suggests an upper limit for women with three risk factors of around 0.6%. As an exact risk cannot be identified for women with three or more risk factors, we have used the model to identify the level of risk that would be required in this group for prophylaxis at 28 weeks to be cost-effective.

Approach to quantifying decision uncertainty

A probabilistic sensitivity analysis (PSA) has been conducted to incorporate uncertainty regarding the model inputs and determine how this uncertainty propagates through the model to translate into uncertainty in the incremental costs and QALYs and therefore decision uncertainty regarding the optimal prophylaxis strategy. The PSA is based on 10,000 parameter samples (probability distributions are provided, see Appendix 4, Tables 22 and 28). In the PSA, the OR for VTE was sampled using the event rates from the study by Gates et al. 67 (OR 0.29, 95% CI 0.01 to 8.37), and this was used to calculate the expected RR given the sampled absolute risk of VTE in the model for people not receiving prophylaxis.

In addition, the decision uncertainty associated with not having perfect information on all model parameters is estimated by the EVPI analysis. The overall EVPI provides an estimate of the increase in net monetary benefit that could be achieved by having perfect information on all model parameters simultaneously. The increase in net monetary benefit that can be achieved by obtaining perfect information on individual parameters or groups of parameters is known as the expected value of perfect parameter information (EVPPI). We have estimated EVPPI using the online Sheffield Accelerated Value of Information (SAVI) tool which uses a regression-based approach to obtain estimates of EVPPI directly from the outcomes of the PSA and the set of parameter inputs that generated those PSA outputs. 115,116 We provide EVPI and EVPPI estimates per patient and we also estimate the EVPI and EVPPI over 5 years of births, assuming 640,370 live births per annum in England and Wales,117 and discounting of future costs and benefits at 3.5%.

Aspects of structural uncertainty such as the choice of one data source over another to inform the parameter distribution, or the impact of various model assumptions, are explored within scenario analyses using the mean estimates for the parameter inputs (referred to as the deterministic model).

Deterministic base-case results for antepartum women with a prior venous thromboembolism

The deterministic base-case results obtained when applying the midpoint parameters estimates to the base-case scenario for antepartum women with a prior VTE are shown in Figure 7. It can be seen that all of the strategies have an incremental cost-effectiveness ratio (ICER) under £30,000 per QALY when compared to a strategy of offering no antepartum or postpartum prophylaxis (no PPX). The strategy of offering only postpartum prophylaxis (PP PPX only) is cost saving compared to no PPX and generates additional QALYs and therefore dominates no PPX. The EThIG RAM has the highest QALY gains and it has lower costs that offer prophylaxis to all from booking (PPX from booking). Therefore, PPX from booking is said to be dominated by the EThIG RAM. The ICER for the Lyon RAM compared to PP PPX only is £53,757 per QALY, whereas the ICER for the EThIG RAM compared to the Lyon RAM is less at £1468 per QALY. Therefore, the Lyon RAM is extendedly dominated because it would never be preferable when the EThIG RAM and PP PPX only strategies are available. The ICER for the EThIG RAM compared to PP PPX for all is £24,982. Therefore, based on the deterministic analysis, PP PPX only would be most cost-effective when applying a cost per QALY threshold of £20,000 and the EThIG RAM would be most cost-effective when applying a cost per QALY threshold or £30,000.

FIGURE 7.

Deterministic base-case results for RAMs validated in antepartum women at high risk of VTE. (References for figure: EThiG,27 Lyon. 32)

Probabilistic base-case results for antepartum women with a prior venous thromboembolism

The results based on the mean outcomes from 10,000 probabilistic model runs are summarised in Table 8. The broad conclusions are the same, in that all of the strategies are cost-effective compared to no prophylaxis, and the optimal strategy when valuing a QALY at £20,000 is PP PPX only. However, the ICER for the EThIG RAM compared to PP PPX only is £56,761 per QALY in the probabilistic analysis, whereas the ICER for this comparison was under £30,000 in the deterministic analysis. This means that the optimal strategy when valuing a QALY at £30,000 is the PP PPX only strategy based on the probabilistic analysis.

However, there is significant uncertainty in the incremental costs and QALYs as demonstrated by Figure 8, which shows the spread of incremental costs and QALYs for antepartum prophylaxis based on the EThIG RAM compared to a strategy of PP PPX only, with 42% of the PSA samples providing an ICER of under £30,000 per QALY for the EThIG RAM compared to PP PPX only. The cost-effectiveness acceptability curve (CEAC) is presented in Figure 9. This shows that the PP PPX only strategy has the highest probability of being most cost-effective (36%), when valuing a QALY at £30,000. However, no PPX, and PPX according to the EThIG RAM both have a > 20% probability of being optimal.

FIGURE 8.

Probabilistic base-case results for antepartum prophylaxis according to the EThIG RAM vs. postpartum prophylaxis only (PP PPX only).

FIGURE 9.

Cost-effectiveness acceptability curve for high-risk antepartum women. (References for figure: EThiG,27 Lyon. 32)

Due to this high degree of uncertainty regarding the optimal strategy, the overall EVPI associated with all parameters included in the PSA, when valuing a QALY at £30,000, was £1454 per patient. Therefore, the population EVPI over 5 years of births would be £21.8 million when taking into account that there are 640,000 births per year, and 0.5% of these are in women with a prior history of VTE. 15,117

Expected value of perfect parameter information was used to determine which individual parameters and groups of parameters were the greatest drivers of uncertainty regarding the optimal strategy. Full results for individual parameters are provided (see Appendix 5, Table 29), but the single most important parameter was the RR of VTE which accounted for 94% of the overall EVPI. Given that any study which provides additional evidence on the RR of VTE could also be used to capture additional information on the RR of bleeding, we estimated the EVPI for these two parameters, which was £1363 per patient, or £20.4 million over 5 years of births (see Appendix 5, Table 30). The remaining groups of parameters examined all had an EVPPI that was less than 1% of the total EVPI.

Deterministic scenario analyses for antepartum women with a prior venous thromboembolism

The deterministic scenario analyses were conducted to explore which model assumptions and inputs were key drivers of decision uncertainty. To do this, deterministic results were generated using midpoint parameter inputs when varying individual model inputs or assumptions. In the base-case deterministic analysis, the optimal strategy, when valuing a QALY at £30,000, was using the EThIG RAM to determine antepartum prophylaxis. The sensitivity of the model results to the various alternative assumptions and data inputs are expressed using INMB benefit for EThIG RAM compared to no prophylaxis. A negative INMB would mean that antepartum prophylaxis using the EThIG RAM has an ICER over £30,000 per QALY compared to a strategy of no prophylaxis.

Figure 10 shows the results for the scenario analyses that had the greatest impact on the INMB when comparing antepartum prophylaxis according to the EThIG RAM against a strategy of no PPX. (Full results for the deterministic scenario analyses are provided, see Appendix 6, Table 40.) It can be seen that none of the scenario analyses result in a negative INMB. Also, it can be seen that the factors that had the greatest impact were those related to PTS, patient characteristics, efficacy and safety of LMWH and the utility decrement associated with daily LMWH injections.

FIGURE 10.

Deterministic scenario analyses for antepartum prophylaxis in high-risk women using EThIG RAM27 compared to offering no prophylaxis (no PPX). (References for figure: Cochrane review,8 Caprini,102 Lenert. 107)

Figure 11 shows the deterministic scenario analyses when comparing antepartum prophylaxis according to the EThIG RAM against a strategy of postpartum prophylaxis only (PP PPX only). In this comparison, many of the same factors are important, but for three scenario analyses, the INMB was negative meaning the optimal strategy switched from antepartum prophylaxis according to the EThIG RAM to PP PPX only. This was true when assuming a lower utility decrement for PTS and in the two scenarios which assumed higher BMI (36 kg/m²), which affects the dosage and therefore the costs of prophylaxis.

FIGURE 11.

Deterministic scenario analyses for antepartum prophylaxis in high-risk women using EThIG RAM27 compared to offering postpartum prophylaxis only (PP PPX only). (References for figure: Cochrane review,8 Caprini,102 Lenert. 107)

In the base-case analysis for high-risk antepartum women, we have assumed that those identified as high risk according to the RAM receive antepartum prophylaxis from their booking appointment. However, the Lyon RAM actually recommends antepartum prophylaxis from 28 weeks for those with a score 3–6 and only recommends prophylaxis from booking in those with a score of 6 or more. As this will affect both the costs and efficacy of offering antepartum prophylaxis using the Lyon RAM, we have therefore conducted a scenario analysis to determine the impact of assuming that antepartum prophylaxis is deferred until 28 weeks. The results for this scenario analysis are provided in Figure 12. In this scenario, the Lyon RAM has an ICER of £83,144 per QALY compared to PP PPX only, which is less favourable than the deterministic ICER in the base case (£53,757 per QALY). This is because, although delaying prophylaxis to 28 weeks reduces the costs of prophylaxis, it also reduces the period of effective prophylaxis and therefore lowers the QALY gains and the cost savings. This scenario analysis suggests that the benefits of the Lyon RAM may be overestimated in the base-case scenario because it recommends a mixture of prophylaxis from booking and prophylaxis from 28 weeks gestation, but the base-case analysis assumes prophylaxis is given from booking in any patient with a Lyon score ≥ 3 as this is the cut-off for offering any antepartum prophylaxis.

FIGURE 12.

Deterministic scenario analysis for the Lyon RAM in women at high risk of VTE when assuming that antepartum prophylaxis is delayed until 28 weeks. (Reference for figure: Lyon. 32)

Given that there were uncertainties in the evidence used to determine the risk of VTE and major bleeding in the cohort of high-risk antepartum patients, a two-way scenario analysis was conducted to explore whether the optimal prophylaxis strategy would vary if the average risks of VTE and major bleeding were higher or lower. The results (see Appendix 6, Table 35) show that the optimal strategy would be to offer only postpartum prophylaxis if the VTE risk was under 10%. The results are not particularly sensitive to the risks of major bleeding, although the optimal strategy does change for higher bleeding risks (> 7%) when the VTE risk is in the range of 11–12%.

Deterministic base-case results for unselected postpartum women

The deterministic base-case results obtained when applying the midpoint parameter estimates to the base-case scenario for unselected postpartum women are presented in Figure 13. For the Sultan RAM, results are presented for various cut-offs defined according to the proportion of women defined as being high risk (e.g. top 1%) when using the Sultan calculator to determine absolute risk. The ICER for the SFOG RAM when using the sensitivity and specificity data reported by Lindqvist et al. 36 is £29,777 compared to no prophylaxis. However, this was a low-quality study, and this RAM did not perform as well in the higher-quality study conducted by Sultan et al. and using this performance data resulted in a deterministic ICER of £86,142 compared to no prophylaxis.

FIGURE 13.

Deterministic base-case results for unselected postpartum women being offered prophylaxis according to various RAMs (RCOG, SFOG, Caprini, Sultan) compared to no prophylaxis (no PPX). (References for figure: Lindqvist 2008,36 Tran 2019,39 Sultan 2016. 11)

Probabilistic base-case results for unselected postpartum women

The PSA was run to compare the RCOG, SFOG and Sultan RAMs using performance data from the Sultan paper. This was chosen as it was the highest-quality study and was considered to provide the most robust estimates of performance for these three RAMs. Also, it had the benefit of estimating the performance of all three RAMs in the same cohort which minimises the risk of bias due to difference in the cohort characteristics or differences in the methods employed.

The results based on the mean outcomes from 10,000 probabilistic model runs are summarised in Table 9. It can be seen that the average QALY gains for all RAM-based strategies are now negative. This means that no PPX is the dominant strategy when incorporating uncertainty regarding the parameter inputs. Figure 14 shows the incremental costs and QALYs for using the Sultan RAM to offer prophylaxis to the top 1% of VTE risk versus no PPX. It can be seen that there is significant uncertainty regarding the incremental QALYs, with a 95% CI of −0.0011 to 0.0003 and 24% of PSA samples resulting in a negative incremental QALY gain.

FIGURE 14.

Probabilistic base-case results for using the Sultan RAM (top 1%) vs. no prophylaxis (no PPX) in unselected postpartum women.

The CEAC for postpartum prophylaxis in unselected women is presented in Figure 15. When valuing a QALY at £30,000, a strategy of offering no prophylaxis has an 89% probability of being optimal, whereas all of the remaining strategies have less than a 10% chance of being optimal.

FIGURE 15.

Cost-effectiveness acceptability curve for prophylaxis strategies in unselected postpartum women.

The overall EVPI associated with all parameters included in the PSA when valuing a QALY at £30,000 was £0.68 per person. Although this is a small amount of EVPI per person, the amount across 5 years of births, assuming 640,000 births per annum,117 would be £2.0 million. No individual parameter had an EVPPI of more than £0.01 per person.

The broad spread of incremental QALY estimates appears to be driven by the uncertainty in the RR of VTE. However, in this case, there is not a large EVPI associated with this parameter as only 11% of the PSA samples resulted in a strategy other than no PPX being optimal (defined as having the maximum INMB when valuing a QALY at £30,000).

Deterministic scenario analyses for unselected postpartum women

Deterministic scenario analyses were conducted to explore which model assumptions and inputs were key drivers of decision uncertainty. To do this, deterministic results were generated using midpoint parameter inputs when varying individual model inputs or assumptions. In the base-case deterministic analysis, the optimal strategy, when valuing a QALY at £30,000, was no prophylaxis, but the strategy with the second highest INMB was offering prophylaxis to patients in the top 5% of VTE risk using the Sultan RAM. Therefore, the sensitivity of the model results to the various alternative assumptions and data inputs are expressed using INMB benefit for Sultan (top 5%) compared to no prophylaxis. A positive INMB would mean Sultan (top 5%) has an ICER of under £30,000 per QALY compared to a strategy of no prophylaxis, whereas a negative INMB would mean that no prophylaxis remains the optimal strategy as in the base case.

The 10 scenario analyses that had the greatest impact on the INMB are presented in Figure 16. (Full results for the deterministic scenario analyses are provided, see Appendix 6, Table 41.) It can be seen that the only one that resulted in Sultan (top 5%) having a positive INMB, and therefore an ICER under £30,000 per QALY, was when we assumed no cost for conducting the risk assessment. It should be noted that this also resulted in a positive INMB for Sultan (top 1%) but the INMB for Sultan (top 5%) was higher meaning that the latter would be the optimal strategy in this scenario. Other factors that appear to be important based on the deterministic scenario analyses were those related to PTS, patient characteristics, safety and efficacy of LMWH and finally the cost and utility impact of non-fatal non-ICH bleeding. The assumption regarding the duration of efficacy applied for 10 days of LMWH was fairly influential, but even assuming a full 6 weeks of efficacy, instead of the 3 weeks assumed in the base case, did not result in a positive INMB for Sultan (top 5%) compared to no prophylaxis.

FIGURE 16.

Deterministic scenario analysis for Sultan (top 5%) vs. no prophylaxis (no PPX) in unselected postpartum women. [References for figure: Sultan (top 5%),11 Caprini,102 Lenert,107 Rodger. 19]

A two-way scenario analysis was also conducted to determine how sensitive the conclusions are to the absolute risks of VTE and major bleeding. The results (see Appendix 6, Table 36) show that using a RAM to select patient for postpartum prophylaxis would be cost-effective (when valuing a QALY at £30,000) if the risks of VTE were higher than assumed in the base-case analysis. For example, an increase in VTE risk from 0.07% to 0.14% would mean that offering prophylaxis using the Sultan (top 5%) would be most cost-effective, but only if the risks of bleeding were 2–7%. Offering prophylaxis to a broader group, using Sultan (top 20%), would be optimal if the risks of VTE were 0.14% and the bleeding risks were under 2%. However, at the level of VTE risk assumed in the base-case scenario, the optimal strategy is not sensitive to the risks of major bleeding.

Deterministic base-case analysis for obese postpartum women

Figure 17 shows the incremental costs and QALYs versus no prophylaxis (no PPX) for the subgroup of postpartum women with obesity as a specific risk factor. It can be seen that prophylaxis for all (PPX for all) would not be the optimal strategy in this population as the ICER is above £30,000 per QALY. In comparison, both versions of the RAM developed by Ellis-Kahana et al. have ICERs under £30,000 per QALY. The two RAMs presented by Ellis-Kahana differ in that one included thromboembolic disorder within the risk score (referred to as the full RAM) and the other excluded this specific risk factor.

FIGURE 17.

Deterministic base-case results for Ellis-Kahana RAM vs. no prophylaxis (no PPX) in obese postpartum women. (References for figure: Ellis-Kahan 2020. 41)

Probabilistic base-case analysis for obese postpartum women

In the PSA, we have compared the Ellis-Kahana RAM (full RAM) with a strategy of no prophylaxis. Results based on mean costs and QALYs are summarised in Table 10. Figure 18 shows the spread of incremental costs and QALYs on the cost-effectiveness plane for the Ellis-Kahana RAM (full RAM) compared to no prophylaxis. The mean QALY gain is negative (−0.0001), but there is a wide spread of incremental QALY estimates, with a 95% CI of −0.013 to 0.004. Therefore, although using the Ellis-Kahana RAM (full RAM) has an ICER under £30,000 versus no prophylaxis in the deterministic analysis, this strategy is dominated by no prophylaxis when using the mean outputs of the PSA as on average it has lower QALYs and higher costs (see Table 10). This is despite the fact that the ICER falls under £30,000 for 64% of the PSA samples. The CEAC for the alternative prophylaxis strategies in obese postpartum women is shown in Figure 19. This shows that the strategy of using the Ellis-Kahana RAM (full RAM) to determine postpartum prophylaxis in obese women has the highest probability (64%) of being the optimal strategy, when valuing a QALY at £30,000.

FIGURE 18.

Probabilistic results for Ellis-Kahana RAM (full RAM including thromboembolic disorder) vs. no prophylaxis (no PPX) in obese postpartum women.

FIGURE 19.

Cost-effectiveness acceptability curve for prophylaxis strategies in obese postpartum women. (References for figure: Ellis-Kahana. 41)

The overall EVPI for this population when comparing these three prophylaxis strategies is £22.35 per patient. This would mean an overall EVPI of £13.4 million over 5 years of births117 when assuming that around 20% of pregnant women are obese. 118 The single most important individual parameter in the EVPPI analysis was the RR of VTE which had an EVPPI that was 99% of the overall EVPI, meaning that obtaining perfect information on this individual parameter would lead to an expected gain of £13.4 million over 5 years of births (see Appendix 5, Table 31). The EVPPI for both the RR of VTE and the RR of bleeding combined was similar (see Appendix 5, Table 32).

The broad spread of incremental QALY estimates appears to be driven by the uncertainty in the RR of VTE. In the population of obese postpartum women, there is a large EVPI associated with this parameter as the wide spread of incremental QALY gains, which is driven by uncertainty in the efficacy of LMWH to prevent VTE, results in the optimal prophylaxis strategy (defined as having the maximum INMB when valuing a QALY at £30,000) being uncertain.

Deterministic scenario analysis for obese postpartum women

The deterministic scenario analyses were conducted to explore which model assumptions and inputs were key drivers of decision uncertainty. To do this, deterministic results were generated using midpoint parameter inputs when varying individual model inputs or assumptions. In the base-case deterministic analysis, the optimal strategy when valuing a QALY at £30,000, was using the Ellis-Kahana RAM (full RAM). Therefore, the sensitivity of the model results to the various alternative assumptions and data inputs are expressed using INMB for Ellis-Kahana RAM (full RAM) compared to no prophylaxis. A negative INMB would mean that the Ellis-Kahana RAM has an ICER of over £30,000 per QALY compared to a strategy of no prophylaxis in that scenario, whereas a positive INMB would mean that using the RAM remains optimal as in the base case.

It can be seen from Figure 20 that the factors that have the largest impact on the INMB were related to PTS and the efficacy of 10 days of LMWH to prevent VTE over 6 weeks. However, only two of these scenarios result in a negative INMB and these were assuming a lower utility decrement for PTS and assuming no risk of PTS in asymptomatic DVT. Other factors that are moderately important are patient characteristics, the RR of major bleeding and cost and QALY implications of non-fatal, non-ICH bleeds, although the impact of these is smaller. Results for the deterministic scenario analyses are provided in full (see Appendix 6, Table 42).

FIGURE 20.

Deterministic scenario analyses for Ellis-Kahana (full RAM) vs. no prophylaxis (no PPX) in obese postpartum women. (References for figure: Cochrane review,8 Caprini,102 Ellis-Kahana,41 Lenert,107 Rodger. 19)

The two-way scenario analysis (see Appendix 6, Table 37) demonstrates that the choice of optimal strategy is not particularly sensitive to the risk of major bleeding when the VTE risk is below 0.6%. However, a strategy of no PPX would be optimal if the VTE risk was 0.07%, similar to that in unselected postpartum women. This suggests that the difference in optimal strategy between this specific at-risk subgroup and the general postpartum population is the level of VTE risk.

Deterministic base-case results for postpartum women after caesarean section

It can be seen from Figure 21 that neither the RCOG RAM nor the novel RAM reported by Binstock et al. 28 is cost-effective when used in postpartum women after caesarean section. This is because both these RAMs had poor specificity in the cohort reported by Binstock et al. and therefore result in over 90% of women being offered prophylaxis after caesarean section in the model. However, given that the risks of VTE are similar in the post-caesarean section group and the obese postpartum group, we considered it likely that a RAM with a better performance would be cost-effective in this subgroup. Therefore, we decided to include the Sultan RAM in the analysis for post-caesarean section women to explore whether a RAM with similar performance to the Sultan RAM would be cost-effective. The results should be interpreted with caution as the novel RAM reported by Sultan et al. 11 has been validated in a broad population of postpartum women and the sensitivity and specificity may be different in the specific subgroup of postpartum women with obesity. However, it can be seen that a RAM with performance similar to the Sultan RAM would need to select the 5% of patients with the highest risk of VTE for prophylaxis to have an ICER under £30,000.

FIGURE 21.

Deterministic base-case results for RCOG, novel RAM (Binstock et al.) and Sultan RAM compared to no prophylaxis (no PPX) in women following caesarean section (NB: Sultan RAM not validated in this specific group). (References for figure: Binstock,28 Sultan. 11)

Probabilistic base-case results for postpartum women after caesarean section

Based on the results of the deterministic analysis, we decided to include the Sultan RAM in the probabilistic analysis along with the RCOG and novel RAMs reported by Binstock et al. 28 This was to explore whether a RAM with performance similar to the Sultan RAM would be cost-effective in the post-caesarean section population. The mean outputs of the PSA are shown in Table 11, where it can be seen that no PPX dominates (i.e., has lower costs and higher QALYs than) all alternative strategies. This is due to the wide spread of incremental QALYs on the cost-effectiveness plane, which is shown in Figure 22 for the Binstock novel RAM and for the Sultan RAM (top 5%). For the Binstock RAM, 41% of PSA samples resulted in a negative incremental QALY gain compared to no prophylaxis, whereas for the Sultan RAM (top 5%) this occurred in only 24% of PSA samples. The CEAC in Figure 23 shows that no PPX had the highest probability of being optimal (57%) in the post-caesarean section population when valuing a QALY at £30,000.

FIGURE 22.

Probabilistic results for the Binstock novel RAM and the Sultan RAM (top 5%) compared to no prophylaxis (no PPX) in women following caesarean section. (References for figure: Binstock,28 Sultan 2016. 11)

FIGURE 23.

Cost-effectiveness acceptability curve for prophylaxis strategies in women following caesarean section. (References for figure: Binstock,28 Sultan 2016. 11)

The overall EVPI was £7.74 per patient, which is equivalent to £5.6 million over 5 years of births117 taking into account that 24% of births are by elective or emergency caesarean section. 55 The parameter with the highest EVPPI was the RR of VTE, which has an EVPPI equivalent to 68% of the overall EVPPI (see Appendix 5, Table 33). This is equivalent to £3.8 million over 5 years of births. No other individual parameter had significant EVPPI (i.e. none >£1 per person). An analysis exploring the EVPPI for various groups of parameters is reported in full (see Appendix 5, Table 34). The EVPPI for both the RR of VTE and the RR of bleeding is estimated to be £5.47 per person or £4.0 million over 5 years of births. All other groups of parameters had an EVPPI that was much lower (≤£1 per person).

If the Sultan RAM is excluded from the analysis, the overall EVPI is lower at £2.06 per person, but none of the individual parameters have significant EVPPI (i.e. none >£1 per person). This is because the optimal strategy is less uncertain with no PPX having a 93% probability of being optimal (i.e. maximising INMB when valuing a QALY at £30,000) when compared to PPX for all and prophylaxis using either the novel Binstock RAM or the RCOG RAM.

In the population of postpartum women who have had a caesarean section, there is significant EVPPI associated with the RR of VTE but only when assuming that a RAM that performs similarly to the Sultan RAM is available. When assuming that only the RCOG or Binstock novel RAMs are available in the post-caesarean section population, the uncertainty regarding the optimal strategy (defined as having the maximum INMB when valuing a QALY at £30,000) is much lower and choice of optimal strategy is less sensitive to the uncertainty regarding the efficacy of LMWH in preventing VTE.

Deterministic scenario analyses for postpartum women after caesarean section

We decided to use the Sultan RAM (top 5%) to explore the sensitivity of the model to the various assumptions and data sources as this strategy was the only strategy with an ICER under £30,000 per QALY in the deterministic analysis. Therefore, in the scenario analyses, if the INMB becomes negative, this means that the Sultan RAM (top 5%) has an ICER over £30,000 per QALY compared to no prophylaxis and the optimal strategy has changed to become no prophylaxis.

The optimal strategy is sensitive to many assumptions because the ICER for using the Sultan RAM (top 5%) compared to no prophylaxis is £29,281 per QALY meaning that factors that have a small impact on the costs and benefits have the potential to change the optimal strategy (see Appendix 6, Table 43 for full deterministic scenario analysis results). However, it can be seen in Figure 24 that factors related to PTS are again important drivers of the INMB with a lower utility decrement for PTS and a lower incidence of PTS resulting in no prophylaxis becoming the optimal strategy. The results are also particularly sensitive to the assumptions regarding whether the efficacy of LMWH is applied for 10 days or 6 weeks rather than the 3 weeks assumed in the base case.

FIGURE 24.

Deterministic scenario analyses for Sultan RAM (top 5%) vs. no prophylaxis (no PPX) in the post-caesarean section population (NB: Sultan RAM not validated in this specific subgroup). (References for figure: Cochrane review,8 Caprini,102 Lenert,107 Rodger. 19)

We also conducted the deterministic scenario analyses for the Binstock Novel RAM compared to no prophylaxis, but none of the scenarios explored resulted in the Binstock novel RAM having an ICER under £30,000 per QALY. The same was true when we used the data from Binstock for the RCOG RAM in the post-caesarean section population.

The two-way scenario analysis (see Appendix 6, Table 38) demonstrates that the VTE risk would need to be much higher for the Binstock novel RAM to be more cost-effective (when valuing a QALY at £30,000), than using a RAM with performance similar to the Sultan RAM. In addition, the risk of VTE would need to be similar to that observed in the unselected postpartum population (0.07%) before no prophylaxis became the optimal strategy. Also, if the risk of VTE was above 0.5%, then the optimal strategy would depend on the risk of major bleeding. For example, prophylaxis for all would be optimal if the risk of bleeding was lower than assumed in the base case and prophylaxis using the Binstock RAM would be optimal if the bleeding risk was similar to that assumed in the base case.

Exploratory deterministic analysis for unselected antepartum women

The deterministic results for the STRATHEGE RAM31 are shown in Figure 25 alongside various theoretical combinations of sensitivity and specificity. These theoretical combinations are provided to explore the trade-off between sensitivity and specificity that would be required for a RAM to achieve a cost per QALY under £30,000, when being used to determine antepartum prophylaxis in an unselected cohort. It can be seen from Figure 25 that the poor sensitivity (0%) of the STRATHEGE RAM,31 which results in 2% of women having antepartum prophylaxis, results in negative QALYs compared with no prophylaxis but at additional cost and it is therefore dominated by a strategy of offering no prophylaxis. From the theoretical combinations of sensitivity and specificity explored, it can be seen that a high degree of specificity would be required for a RAM used in this population, with a specificity of 90–95% being required for a RAM whose sensitivity is between 100% and 53%, respectively.

FIGURE 25.

Deterministic results for using the STRATHEGE RAM or a theoretical RAM (shown for various combinations of sensitivity and specificity) compared to no prophylaxis (no PPX) in an unselected antepartum cohort. (References for figure: STATHEGE. 31)

Exploratory deterministic analysis for antepartum women with three risk factors

Antepartum women with three risk factors are currently offered antepartum prophylaxis from 28 weeks gestation according to the RCOG guidance (provided none of the risk factors is prior VTE or another risk factor that warrants earlier prophylaxis). 7 Any woman offered antepartum prophylaxis within RCOG is then eligible for 6 weeks of postpartum prophylaxis. As we were unable to obtain an exact estimate for absolute VTE risk in the group with three antepartum risk factors (see Model inputs for secondary scenarios where antepartum prophylaxis is offered at 28 weeks), we have conducted an exploratory analysis to determine the optimal strategy in this group across differing levels of VTE and bleeding risk. The strategy of prophylaxis from 28 weeks gestation (followed by postpartum prophylaxis for 6 weeks) is compared against a strategy of offering no prophylaxis at all and a strategy of offering no antepartum prophylaxis but assuming that all women will receive 6 weeks of postpartum prophylaxis. The results (see Appendix 6, Table 39) show that for the level of bleeding risk assumed in the base case (4.58%), an absolute risk of VTE of > 0.5% would be required for 6 weeks of postpartum prophylaxis for all to be optimal. A more precise threshold analysis identified that 6 weeks of postpartum prophylaxis had an ICER under £30,000 compared to thromboprophylaxis for none only when the risk of VTE was > 0.57%. We estimated in section Model inputs for secondary scenarios where antepartum prophylaxis is offered at 28 weeks that the upper limit of VTE risk in antepartum women with three risk factors (excluding a prior VTE) was likely to be around 0.6%. This exploratory analysis suggests that offering 6 weeks of postpartum prophylaxis to women with three antepartum risk factors would only be cost-effective if the absolute risk in this group is at the higher end of the expected range. In addition, offering antepartum prophylaxis from 28 weeks is unlikely to be cost-effective in this group. These findings only apply to women where none of the three risk factors are a prior VTE or another risk factor that currently qualifies the woman for prophylaxis from booking under the RCOG guideline as these women were excluded when calculating the absolute risks.

Workshop findings

We identified six themes that may impact on future recruitment to clinical trials.

1. Pregnant women receive limited information about VTE or risks and benefits of thromboprophylaxis during pregnancy or postpartum.

Participants described receiving little information about VTE or thromboprophylaxis in terms of either risk of VTE during pregnancy, understanding why they had been given anticoagulants (particularly for low-risk participants) or the risks and benefits of treatment. For participants who had pre-existing conditions, some had investigated their treatment and identified the need for thromboprophylaxis prior to being asked to take them by a healthcare professional. However, although the majority of participants had some general awareness about DVT and PE, it was not considered to be something that was spoken about or discussed as part of their maternity care and participants without pre-existing conditions that increased their risk of VTE could recall little or no discussion of the increased risk of VTE during or after pregnancy. This lack of information was perceived to be a potential barrier to participation in a trial.

W1P6: I had no idea, I didn’t really understand any of it, it was ticked off in your book, it wasn’t really spoken about openly and I think if more people knew more about it and the risks involved, they would much more likely take part in a trial.

W1P1: I mean there is some talk of VTE, but in pregnancy I don’t think, it didn’t come across as a real risk to me and I definitely didn’t know that I’d be almost permanently disabled because of the clot that wasn’t treated properly.

W2P3: No-one has the time to talk to you in pregnancy and I hardly had no more than 10 minutes with the midwife during pregnancy before the DVT. [ … ] I didn’t even know blood clots could happen during pregnancy at that point.

Participants in the low-risk groups in particular recalled little to no discussion of risks and benefits of anticoagulants. Participants took anticoagulants because they had been told to, but often did not understand why they had been prescribed them and made their own assumptions about the rationale behind their treatment. Notably, few participants recalled any discussion of risks associated with thromboprophylaxis.

W2P2: After surgery, they just said ‘Here’s a Paracetamol and here’s an injection’. And sent me home with a bag of injections and told to do it.

W3P3: I’m not aware of the risks, so no one’s explained it to me, and I didn’t feel that there was much point searching for it after being on it for a while, so that was it.

W3P4: I was on the blood-thinning injections for ten days after giving birth, and I, to me it was because I lost, I lost 1.5 litres of blood and I had pre-eclampsia, so they’re the reasons I believe that I was put on it, but I wasn’t actually told why I needed them, I was just discharged with them, and I didn’t really question it.

This lack of understanding of risk factors or rationale behind treatment may affect compliance with treatment. Some participants described stopping treatment early because they did not understand why they had been prescribed thromboprophylaxis, or because they did not understand why their treatment duration differed from previous pregnancies or from peers.

W2P2: The first time round I just took the injections because I was told to, but if I missed I wasn’t really bothered.

W3P4: […] I probably did about 7 injections in total, including the two in hospital. So I missed three basically. (Int: Right. And did you understand why you were doing it?) W3P4: No. I think that’s probably why I didn’t really continue.

W4P6: But, for this time, it seemed much, much longer and it seemed like it was they’d said almost like a month for the blood thinners and the injections which I was really upset out because you know ten days is bad enough but to keep injecting yourself for a month without any kind of explanation as to why the time had increased. On reflection, I had lost a lot of blood during the birth, during the surgery so I was put two together myself there. However, I stopped after ten days because just the stress of having to it with two little ones running round was just too much and I didn’t have any blood pressure issues.

Women described information seeking and doing their own independent research to understand why they had been given thromboprophylaxis. Others described how they sought information on the internet or using forums such as Thrombosis UK.

W2P3: I was sent home with injections and after I went home and it all sunk in, I realized what the impact of that was, not understanding what that meant for the pregnancy. It took a lot of researching you know by myself.

1. Pregnant women who had previously received thromboprophylaxis accepted current prescribing practice and perceived potential future trials to be withholding treatment.

Some high-risk participants, such as those who were on long-term thromboprophylaxis, felt that taking part in a trial with a placebo option was not an option for them as a placebo was not a feasible option; ‘I couldn’t choose not to take the blood thinners’ (W1P7). Although reporting limited knowledge about risks and benefits associated with thromboprophylaxis, they perceived the risk factors to be a reason why they needed treatment and would not welcome taking part in a trial.

W1P1: I had a risk factor in that Dad had DVT but if I didn’t have enough risk factors to qualify then I would have been happy to partake in a trial because I wouldn’t have had it anyway so I’d have been happy to take part in a trial to reduce my risk. If I have placebo I’m no worse off than I would have been. But now, knowing what I know now there’s no way that I wouldn’t be taking clexane.

W1P2: How ethically can you deny someone a treatment if they have an identified risk factor?

W3P5: So from my perspective because of what happened because of how traumatic it was, I wouldn’t have said I’d take a trial, and maybe have it, I think just accepted that what they were giving me was what I needed to have.

Thromboprophylaxis was perceived as potentially life-saving and the prescription of thromboprophylaxis accepted as best practice. Participants from all groups struggled to understand the concept of a poor evidence base underpinning current guidelines and perceived the introduction of a trial as removing current best practice rather than offering a choice of treatments where the current evidence is unclear. They perceived receiving the placebo as a risk and would take part in a trial when they saw it as an opportunity to obtain a treatment that would otherwise be withheld.

W4P1: Of course I took that [thromboprophylaxis], but if somebody mentioned the word trial to me, I would have said no, cos I would not have put anything at risk for me or my children. And I say, that feels quite uncomfortable for me to say a flat out no cos I’m not usually a no person, but I feel in this situation I would have said no.

W3P2: I would definitely take it (LMWH rather than be in a trial), 100%. I think if it can’t harm you what’s the harm in doing it?

W3P6: I’ve been really anxious. If it had been a choice of you know, ‘you’re not going to get them, but if we put you on a trial there is a fifty-fifty chance you’ll get them, or placebo’, then I’d have gone for it, but if it was a case of ‘you can have them or you can go into a trial’, I would definitely have wanted them, because of my anxiety around being in a you know, over coagulated state not having the anticoagulants.

Even when not fully understanding reasons for needing the treatment, most participants complied with the treatment and did not question whether it had been prescribed appropriately. They appeared to be passive recipients of the treatment and even when they did not fully understand the rationale behind why they had been prescribed thromboprophylaxis, they complied despite being unhappy about it.

W4P3: I don’t really understand the mechanisms other than, you know, ok a blood clot can be very serious so, you know, there wasn’t, I didn’t really feel around that point that I knew enough to challenge or to refuse. But, I also I didn’t really feel like I wanted to, I just sort of was resigned to have to do this, you know, unpleasant thing for a while.

W4P6: [ … ] the resentment and resignation are the two sort of words that really spring out to me in my experience, the kind of resentment of feeling sort of done to and the just being resigned to just having to do it, so they just resonated with me.

W3P5: I prefer not to have done it, I didn’t really like having to have to do injections, I didn’t enjoy it, but you know, I just did it. Got on with it.

1. Negative experiences associated with injections were minimised by healthcare practitioners but may increase likelihood of attrition.

Participants spent a significant amount of time discussing the side effects of the injections which they felt would prevent adherence to active treatment for people who had not had previous experience of VTE. They felt that healthcare professionals greatly underestimate the negative impact of undertaking the injections and minimise the problems associated with the pain and discomfort of the injections themselves, as well as the significant bruising or lumps on injection sites.

W1P3: My experience has been that clinical staff involved don’t necessarily understand what the injections are so the midwives have been ‘oh yes, it’s just a bit stingy’.

W1P4: I found a lot of people who administer these injections have got no idea. They just go (mimics giving injection) and then you get that burn. It takes a while to know how to do it and you don’t need to have that burn at all if you know how to do it.

For some high-risk participants who valued the LMWH injections as a more acceptable alternative to warfarin due to the difficulties in moderating international normalised ratio (INR), the injections were difficult but welcomed as an opportunity to ‘keep them safe’. For others, although a minority did not struggle with the injections (notably those who had to inject for other reasons), many described feelings of resentment and hating the experience of doing the injections, finding the process of injecting to be difficult both physically and psychologically. One high-risk participant said she would welcome participating in a trial as an opportunity not to have the injections and other high-risk patients reported choosing not to have another child due to the impact of the injections.

W4P2: I actually ended up with some physical lumps on my stomach from it, but I in the end had to inject for 18 weeks which was so, so painful and sore in my stomach and I just resented it, I really hated it.

W2P2: I think my bruising hurt more than my caesarean section [ … ] I certainly would never consider having another baby now because of the thought of injecting myself.

W1P2: I decided not to have a second baby because I couldn’t face the idea of taking those injections twice a day.

Again, a lack of information was felt to be a contributor to the anxiety surrounding the injections due to ‘not knowing what was normal’ (W2P2) and participants would have valued information from health-care professionals about what to expect with regard to the injections, particularly the potential for lumps and advice about how to inject less painfully. Without being shown how to do the injections, they were unclear about the most appropriate place to place the needle (particularly during the latter stages of pregnancy) and were unsure whether the pain meant they were not doing the injections correctly. In the absence of information from clinicians, they obtained information about how to undertake the injections from Thrombosis UK, social media, internet sites or watching others do the injection.

W2P1: I absolutely freaked [at the pain], I thought something was wrong.

W3P1: I was so happy that I’d actually seen somebody do it, it gave me a lot more confidence, to know what I’m doing, like how much fat do I need to grab on my stomach.

Negative experiences associated with the injections also included difficulty in undertaking the injections while looking after a newborn (potentially alongside other young children), and practical issues such as being able to dispose of sharps bins safely. These factors were all felt to potentially impact on attrition rates within future trials but may be addressed by improved information and understanding of the rationale behind the trial.

W1PM. I don’t know how you could recruit normal people who didn’t have that trauma and that personal history to actually inject themselves because it’s such a massive thing to do. For a normal person that doesn’t have diabetes or other reasons to inject themselves when they haven’t had to inject themselves that I just don’t know how you could recruit thousands of people to get enough evidence.

W1P4: I think you’d have to make it really clear what the benefit to the person is and their case, because for the greater good to have an injection every day that stings and really awkward to do, and this may be not necessary, I think that would be hard to sell.

W3P3: I think if you understand fully why you’re doing it, and also how to do it and how to do it safely, and all the rest of it, I think it then becomes easier and actually you know why you’re doing it, so even if it hurts you’re more likely to carry on and do it, for a period of time.

1. Participants saw RCTs as an opportunity to access improved care and information, as well as improving future care for others.

A number of participants had previously taken part in clinical trials either while pregnant or at other times and spoke favourably about participating in clinical trials as a way of helping future pregnant women. They understood the potential benefits to future patients, and even specified that they would have preferred to take treatment as part of a clinical trial as it would provide wider benefit. While accepting some level of risk to themselves, they were clear that they would be unwilling to take part in a clinical trial that may cause any level of risk to the baby.

W3P6: So I feel like if I could help people in the future so that their post-recovery, was better, I would like to take part. [ … ] I’d like to feel that I would take part in a trial for that reason. Not specifically just to benefit myself, but to help with the research as well going forward for other women in the future.

W1P1: Yes, just going to say that I would view the risks to the baby would be different to me, so I would probably be happy to take some risks to me, but have a very low tolerance to having risks to the baby.

Overall concerns about being dismissed, not listened to and offered little information about their care meant that participants saw a benefit to enrolling in a trial as a way to receive a better standard of care and discussion of the risks and benefits of thromboprophylaxis. Participants felt that taking part in a clinical trial would offer additional monitoring and access to health care, and the offer of additional scans or appointments may encourage people to participate in a clinical trial.

W1P2: I think if you’re doing it as a trial, then you get a lot more contact with health professionals, specifically about the injections. Whereas if it was standard care [ … ] you probably don’t have so much contact with somebody, specifically about the anticoagulation, so you probably get better yeah, better adherence if you were in a trial than if you just had it in bog standard of care.

W3P3: I think for me it’s, it would be about clarity of information and actually, almost providing ‘okay this what we’re trying to research, but if you join this trial, we’ll give you xyz’. So you get extra check-ups, extra scans, extra, if you were doing it obviously pre-, during your pregnancy so that you were confident that regardless of whether you were or you weren’t, your standard of care is almost raised up another level. so you weren’t just being looked after, you were being like gold standard, you know you were getting check-ups, you know, once a month.

W3P6: I think if you had [ … ] perhaps you had a midwife or somebody from the trial team, who are checking in on you. If you’re finding the injections okay, or if you’re on a no treatment arm, just checking in that you’re psychologically okay with that, I think again having somebody checking in on you in those early weeks particularly if you might not have other support at home, that could be a benefit for some people.

The provision of additional care may also provide reassurance for people who perceived the trial as introducing additional risk. Again, concerns related mainly to the potential risks associated with not receiving LMWH, rather than potential risks of LMWH and participants felt that clear information about potential benefits of the trial would be needed.

W2P1: I think for me it would be how closely you’re going to monitor it, you know, like what sort of tests I suppose, would you be doing to monitor if I’m getting a blood clot anywhere? You know, how closely are you going to be looking after me kind of thing? For me personally. I’d want to know are you going to see me quite often?

W4P3: If someone had said to me (at planned caesarean section) you could be prescribed this drug because of factors age and planned caesarean section, but you know if you had the opportunity not to and we would monitor your situation, I think I’d be more inclined.

W4P6: But, I think one thing that could have helped me [ … ], if you were going to have to help me get the risk of an unexpected bleed or something like that we might have you in more regularly for blood pressures checks, or that we might have you go to a particular clinic just to keep an eye on you, that kind of thing.

There were mixed views about whether participation in a trial would increase or decrease their likelihood of continuing with injections. While some felt that they would feel a moral obligation to continue with uncomfortable or painful treatment due to the wider contribution to research, others perceived being offered treatment as part of a trial as evidence that they did not really need the treatment.

W4P1: But, yeah if it was part of the trial I probably would have been much more likely to continue because I would have had the rationalised reasons why it needed to be for that period of time.

W4P5: I would feel like I was obliged, that I would be letting people down or affecting their research and outcome if I weren’t to, if I was to give up.

W3P5: I think your mindset changes if you’re doing it for a trial, you don’t necessarily need it, so you kind of feel like you can opt in or out, if you’re finding it too difficult, if you don’t feel like you’re getting anything out of doing this.

1. Consent for future trials should be undertaken antenatally rather than postnatally. Information provision and understanding are key.

Within these workshops, participants who had received their thromboprophylaxis antenatally or had their risk factors explained antenatally [e.g. in vitro fertilisation (IVF) births] were more likely to have had the risks and benefits of treatment explained to them. Participants strongly supported antenatal recruitment to trials, with information provision and provisional consent provided at a time when they had ‘headspace’ (W4P1) and time to understand the information given, and to discuss with their partners.

W2P1: I think like [W2P2] said, it would be good if they were given information before giving birth, just so they’ve got the capacity to understand it, and process it and everything.

W3P1: obviously it’s not something you would want to have pounced on you, just as you’ve given birth, ‘do you want to take part in this trial?’, I think it would definitely have to be, you know, something that you talk about at least in the last maybe three months of your pregnancy, with your midwife.

W4P6: It was only just to sort of say that I agree with the sort of consensus there, that part of information, like I say high quality of information with a trusted person during pregnancy so you had the chance to have that some form of better clarity of decision making and then I say the opportunity to probably opt-out just depending on kind of how the birth went, how it felt.

Postnatal recruitment and provision of full informed consent were considered unfeasible and impractical, particularly following emergency caesarean section or a difficult labour. Participants described the confusion and feelings of being overwhelmed after giving birth, which would make them unwilling or unable to consider participating in a trial if they had not already undertaken to do so.

W3P3: I think if someone had sat down, if the option was a trial or nothing, I would have been up for doing a trial, but I wouldn’t have wanted to make that decision after birth. I think once you’ve just given birth it’s that kind of, there’s a lot going on, and you’ve got a lot to process, and for me I was a first-time Mum, so it was a lot to take in, sort of first baby, I’m like ‘oh my god, what do I need to do?’,

W3P2: No, not at all, if someone said to me about a trial I’d say ‘sorry I’m not interested at the moment, I’m trying to find my feet at the moment being a parent’, and I think that would even be the same with a second child because you’re adding another child into the mix, kind of balance, having another one at home with you I feel like I wouldn’t be interested at all, personally.

W3P6: And I think what you were saying about communication and understanding is going to be key, particularly given the difficulties and how overwhelmed you can be in that immediate post birth period, both kind of physically and emotionally. A slow burn in terms of awareness of the trial and in sort of mid to late pregnancy would be the way to do it, I think.

1. Cluster randomisation was felt to provide greater buy-in from clinicians, and lead to quicker identification of any problems than individual randomisation.

When describing management of prior VTE, participants described a lack of consistency of advice between different departments of the hospital and felt that, for example, midwives and ED staff had different understanding of how to manage patients with VTE during pregnancy. Participants generally favoured the concept of cluster randomisation as they perceived that the treatment arm provided would be more acceptable to all clinicians, which would lead to improved consistency of management throughout the hospital. They felt that support would be improved and that any problems arising would be more visible and picked up quicker than within individual randomisation.

W4P3: Then, yeah the kind of security of knowing that there are other people, other women in that situation [ … ] you’d feel like you’re monitored as group, not just as an individual, and individuals, you know, sometimes you don’t want to feel like you’re the one that gets overlooked or falls through the cracks.

W1P4: I think it would be easier to provide support because the whole hospital would be going the same direction, I think I would actually be reassured as a patient, because if something was going really wrong I think the doctor would pick it up faster,

W1P6: I think I’d be much more likely to take something that was done as a hospital, you know that everyone else is doing the same thing, and it’s not sort of just a one person thing, if that makes sense? I think I’d feel more comfortable. [ … ] I think just because that’s what they’re doing, they’re all doing the same thing, rather than it just being you know it’s just a certain amount of people. It would be everybody that’s in that same care setting as you.

Participants saw benefits to randomising as a unit (hospital) in terms of having a higher likelihood of meeting other people who were on the same treatment, and one participant who had previously taking part in clinical research felt that randomising as a unit (hospital) would prevent discussions or concerns about which was the ‘best’ treatment between patients in hospital (e.g. if on a ward together). However, there were some concerns that cluster randomisation may lead to a ‘postcode lottery’ that would result in you being offered treatment depending on where you lived, and participants expressed the need to ensure that hospitals were well matched in terms of populations. (Again, concerns centred around hospitals delivering the placebo as offering a higher-risk option.)

W1P4: I would want to be reassured that it wasn’t like all the hospitals in the North are doing it one way, and all the hospitals in the South [waved gesture 34m 56s] so that they actually do take into account populations and make it you know, really sort of correct [Int: Yeah], just age, and socioeconomic, and race, all that.

W2P3: I would definitely, no it depends on the occasion, you say some hospitals offer and some don’t, not all of them might be you know convenient for you to access, so if the nearest hospital to me didn’t offer it, I’d probably not want to be there, so, yeah...

W3P6: I think it kind of the postcode lottery pops kind of into your head then [ … ] I do think that if you were told if you were giving birth in Plymouth you’re going to have a different post-birth care than your giving birth in Exeter or wherever. I think if you could opt out of that and just be guided by your individual consultant or midwife lead, and they would decide based on current evidence what is best for you, I think that would be okay. But if you were kind of put in the situation where that was what was going to happen just because of where you are, I think that would probably be less acceptable.

Who would clinicians be prepared to randomise in a future trial of low-molecular-weight heparin or no low-molecular-weight heparin?

The questions and results for the patient scenarios for patients who were not eligible for antepartum prophylaxis, and patients who were eligible for antepartum prophylaxis are reported in Tables 13 and 14 respectively. We asked participants to state whether they would randomise the patient, not randomise and prescribe LMWH, not randomise and not prescribe LMWH for each scenario, assuming the patient has no other risk factors. For the patients who were eligible for antepartum prophylaxis, we asked them to specify each option (1) from booking, (2) from 28 weeks or (3) postnatally.

Concerns about recruiting patients in the scenarios listed into randomised controlled trials

We also asked clinicians to explain any concerns they may have about recruiting any of the patients listed above into a RCT (36 responses). Free-text comments indicated that clinicians were reluctant to randomise for women with high BMI (some said > 30 kg/m², others > 40 kg/m²) or previous VTE but more support for the groups who were perceived to be lower risk; age 35–40 years, BMI 30–35 kg/m².

This was reflected in the scenario results, which suggested lower support for randomisation with higher BMI [e.g. 30% (34/113) willing to prescribe for BMI > 40 kg/m² vs. 75% (85/113) willing to prescribe for BMI 32 kg/m²and age 36 years]. Similarly, willingness to randomise for patients with previous VTE was low antenatally, although 68% said they would randomise from booking for prior VTE associated with major abdominal surgery and no other risk factors. Four people commented that emergency caesarean section has a higher risk than elective caesarean section, which was reflected in the scenarios. While 69% (38/113) would be willing to randomise patients with BMI 32 kg/m² and elective caesarean section, only 34% (38/111) would be willing to do so for a patient with BMI 32 kg/m² and emergency caesarean section.

Some free-text comments indicated that the lack of detail within the scenarios made it difficult to provide a response (e.g. ‘depends on the mode of delivery’ or ‘I would want to consider other risk factors, how long they were in labour, other confounding factors before I was happy for them to be randomised’). While some explained the rationale that underpinned their understanding of risk (e.g. ‘emergency caesarean section patients are usually less mobile and take much longer to recover’, ‘really difficult with BMI over 40 kg/m² as often also very sedentary’), most comments related to ‘risk’ as currently perceived within the existing RCOG guidelines. Some recognised the lack of evidence base and the need for further interventional trials, whereas others suggested a high level of trust in existing guidelines and reluctance to deviate from guidelines or understanding of the need for further evidence. Free-text comments as well as question responses suggest that clinicians would be unwilling to randomise patients who are currently assessed as high risk within the RCOG guidance and that there is a low level of clinical equipoise in this population.

In your preamble you suggest that I should not use clinical guidelines to influence my choice however my practice is entirely formed by the clinical guidelines! I trust the professionals and processes behind the guidelines. All the patients listed meet criteria for postnatal LMWH: they all have at least 2 risk factors.

As far as I can make out there is virtually no intervention-based study is, regional and national guidelines are based on observational and uncertain population studies and clinical opinion. Much better interventional data required at all levels for thrombosis treatment and prevention.

For all of the women listed I would consider that they have sufficient risk factors to warrant prophylaxis which is my current practice and is well tolerated and I would not deny them this by entering into a trial.

Clinician perspectives of acceptability of recruitment to cluster randomised controlled trials

We asked clinicians whether they felt that it would be acceptable to randomly allocate hospitals or NHS Trusts to provide LMWH or no LMWH for the specified patient groups, rather than the traditional approach of randomly allocating each individual person to either LMWH or no LMWH.

Table 15 shows that two-thirds of the clinicians felt that it was only acceptable to allocate treatment at an individual level, but few provided further details explaining their answer. Some participants questioned how treatment groups could be matched when using cluster randomisation (i.e. ensuring similar populations at different hospitals), and one participant was concerned that hospitals/NHS Trusts who were randomised to no LMWH would have to report higher rates of VTE. One clinician who felt that it was acceptable to allocate treatment at hospital/NHS Trust level commented that cohort randomisation would simplify the trial.

If it was felt that the treatment groups could be appropriately matched i.e. similar district generals given treatment and not. Otherwise would have to be same centre with the 2 options.

In my view I 100% agree on cohort randomisation to simplify it.

Not blinded and may change other management within trust if all patients managed the same way therefore better to randomise each individual patient for more meaningful results.

The trust allocated no LMWH would presumably have higher rates of VTE to be reported which would be unfair.

What guidelines do clinicians use to support decision-making?

We asked what guidelines clinicians use to support decision-making, to understand what influenced their perspectives. RCOG were most commonly referenced, but clinicians also referenced American Society of Hematology (ASH) guidelines (n = 2), ACOG (n = 2) other local regional or hospital-based guidelines (n = 7). Their responses are summarised in Table 16.

Which groups of patients would benefit from improved evidence from clinical trials?

We asked clinicians whether there were any particular groups of patients who they felt would benefit from improved evidence from clinical trials. Although 24 clinicians responded, the groups were disparate and there was no single group who was highlighted more than others, suggesting a wide range of conditions for which clinicians felt further evidence would be valued. Respondents indicated a range of combinations of risk factors, as well as the following: family history of VTE/thrombophilia, low-risk thrombophilias, IVF-assisted reproductive technology (ART) patients, advanced maternal age, elective caesarean section, emergency caesarean section, blood loss 500–1000 ml, first trimester pregnancy loss, BMI < 35 kg/m², e-cigarette use, non-Caucasian patients, pre-term birth, hyperemesis severe dehydration (temporary), hypertriglyceridaemia, hypothyroid, previous history of transient ischaemic attack or stroke of unknown aetiology while on birth control pill.

Demographic information (n = 82)

The clinician role and demographic information for survey respondents are detailed in Table 17.