Postmortem studies indicate that pulmonary embolism is responsible for 10% of deaths in hospitalised patients (Sandler & Martin, 1989; Lindblad et al, 1991; Baglin et al, 1997). In 2001, the United States Agency for Healthcare Research and Quality ranked thromboprophylaxis for appropriate patients in hospital as the number one most important safety practice in the USA (Agency for Healthcare Research and Quality 2001). At an American Public Health Association Leadership Conference in 2003 it was recorded that, in the USA, more people die each year from pulmonary embolism than from motor vehicle accidents, breast cancer and acquired immunodeficiency syndrome (AIDS) combined (American Public Health Association 2003). At this Conference, Samuel Goldhaber stated ‘Venous thromboembolism is the number-one cause of unexpected hospital death…the disconnect between evidence and execution as it relates to DVT prevention amounts to a public health crisis’. Writing in the Journal of Thrombosis and Haemostasis Goldhaber (2007) commented ‘…we need to deliver a more unified, coordinated and stronger message: VTE prophylaxis in high-risk patients is mandatory, not optional’. Prompting this remark were two cited articles, one published in the same edition of that journal showing that, in North America, thromboprophylaxis in medical inpatients was still inadequate. An audit conducted on 200 000 patients in more than 200 hospitals from around the USA showed that only one-third of patients received appropriate thromboprophylaxis (Amin et al, 2007). In a Canadian survey of more than 4000 medical admissions it was only one-sixth (Kahn et al, 2007). The very recent ENDORSE (Epidemiologic International Day for the Evaluation of Patients at Risk of Venous Thrombosis in the Acute Hospital Care Setting) study (Cohen et al, 2008) revealed that 51% of almost 70 000 patients in 358 hospitals across 32 countries were considered to be at risk of venous thromboembolism but only 60% of surgical patients and 40% of medical patients received thromboprophylaxis. In the UK the ‘crisis’ has only recently been recognised at a ‘health policy level’ (House of Commons Health Committee 2005). For patients in England the Chief Medical Officer (CMO) concluded that venous thromboembolism (VTE) risk assessment of every patient on admission needs to become a reality and that potentially at least 25 000 deaths could be prevented annually (Department of Health 2007). The CMO has outlined a National Leadership Venous Thromboembolism Strategy with an Implementation Working Group (IWG) tasked with providing oversight and leadership to the work required by the Healthcare Commission, National Institute for Health and Clinical Excellence (NICE), National Health Service (NHS) Litigation Authority, National Patient Safety Agency, other NHS partners and the wider healthcare sector. The issue is clearly not confined to England and a response is needed throughout the UK. The North American experience illustrates how difficult implementation will be. In the UK specifically, issues relating both to methods of risk assessment and modality of prophylaxis will have to be resolved, for example:
Many studies have been performed to determine the incidence of deep vein thrombosis (DVT) and pulmonary embolus (PE) in various groups of patients and to identify risk factors. The risk of VTE in a hospitalised patient depends not only on the reason for admission (procedural risk) but also on pre-existing patient-related factors (patient-related risk). The decision as to whether a patient requires thromboprophylaxis depends on the absolute thrombosis rate and the choice of prophylaxis depends on the balance between absolute thrombosis and bleeding risks (procedural and patient-related) (Table I). For each patient an assessment of risk should be undertaken on admission and ideally reviewed periodically during hospitalization, particularly in medical patients whose risk may escalate during hospitalisation. Depending on the degree of risk patients should receive advice and treatment to reduce risk. Patients should be mobilised early (Pearse et al, 2007) and prophylaxis should be given to patients at moderate to high risk [Thromboembolism Risk Factors (THRIFT) Consensus Group 1992].
Table I. Risk factors for venous thrombosis (procedural and patient risks) and risk factors for bleeding (procedural and patient risk).
|Thrombosis risk factors: procedural|
| Major orthopaedic surgery to lower limb, for example hip or knee replacement|
| Abdominal or pelvic surgery lasting more than 30 min under general anaesthetic|
| Major trauma, hip fracture is associated with a very high risk of deep vein thrombosis|
|Thrombosis risk factors: patient related|
| Age > 40 years and particularly >60 years|
| Obesity, BMI > 30 kg/m2 and particularly >35 kg/m2|
| Previous DVT or PE|
| Known thrombophilia (a predisposing state which may be heritable)|
| Heart failure|
| Respiratory disease|
| Severe infection|
| Oestrogen therapy and high dose progestogens|
| Pregnancy and the postpartum|
|Bleeding risk factors: procedural|
| Eye surgery|
| Other procedures with a high bleeding risk|
|Bleeding risk factors: patient related|
| Haemophilia and other bleeding disorders|
| Thrombocytopenia (platelets < 100 × 109/l)|
| Recent cerebral haemorrhage (in previous month)|
| Severe hypertension|
| Severe liver disease (prolonged PT or oesophageal varices)|
| Peptic ulcer|
To date, risk assessment in surgical patients has been driven primarily by the procedural risk and a simple Risk Assessment Model (RAM) based on procedure and age is frequently used in practice (Kearon & Hirsh, 2001). Nevertheless, despite the availability of simple, cheap and effective prophylaxis appropriate preventive methods are still substantially underused even in surgical patients (Kakkar et al, 2004; Yu et al, 2007). RAMs are not as well established for medical patients and, whilst the majority of hospitalised patients are at moderate or high risk, only a minority currently receive prophylaxis (Amin et al, 2007; Kahn et al, 2007; Tapson et al, 2007; Yu et al, 2007). Patient group-specific RAMs (e.g. inpatient surgery) offer the advantage of relative simplicity, familiarity and understanding for health care staff caring for a defined patient group but a single RAM at the point of entry to hospital for all patients is feasible and may be advantageous (Kucher et al, 2005). A crucial issue is whether risk assessment and prophylaxis are connected by ‘opt-in’ or ‘opt-out’ policies. An ‘opt-in’ policy is one in which patients are only offered prophylaxis if considered to be at sufficient risk. An ‘opt-out’ policy is one in which all patients (or all patients in a group, e.g. medical inpatients) are offered prophylaxis unless they are positively judged to be at low risk of thrombosis. If the default method of prophylaxis is an anticoagulant, such as low dose heparin, then patients at high risk of bleeding would either not ‘opt-in’ or would ‘opt-out’ of the anticoagulant arm of a prophylaxis policy and would be offered a mechanical method.
These issues will have to be considered by the CMO’s IWG, which has been tasked with implementing risk assessment for England and Wales.
NICE Clinical Guideline 46 ‘Reducing the risk of venous thromboembolism (deep vein thrombosis and pulmonary embolus) in patients undergoing surgery’ was published in April 2007 (NICE 2007a). The guideline presents a list of patient-related risk factors and recommends prophylaxis according to surgical specialty. Mechanical methods of prevention are recommended for non-orthopaedic surgery when there are no patient-related risk factors. When there are patient-related risk factors, and for orthopaedic surgery, it is recommended that LMWH or fondaparinux is added. Age over 60 years is included as a patient-related risk factor. The publication of the NICE guideline illustrates the potential problem of different interpretation of evidence by different expert groups. The American College of Chest Physicians (ACCP) 2004 recommendations on prevention of venous thromboembolism (Geerts et al, 2004) were produced by a defined systematic review (Schunemann et al, 2004). The conclusion of the ACCP was that mechanical methods of prophylaxis should be used primarily in patients at high risk of bleeding or as an adjunct to anticoagulant-based prophylaxis. The reasons for this recommendation were that no mechanical method has been shown to reduce the risk of death or pulmonary embolus, most trials of mechanical prophylaxis were not blinded, and mechanical methods may not perform as well in routine clinical practice because of relatively poor compliance (Geerts et al, 2004). In contrast, meta-analysis of studies of unfractionated heparin have shown a reduction in fatal PE and all-cause mortality (Collins et al, 1988) and low molecular weight heparins are at least as effective as unfractionated heparin, and are probably safer (Leizorovicz et al, 1992; Kakkar et al, 1993; Koch et al, 2001). The NICE guideline development group accepted a reduced relative risk of any DVT by any mechanical method as sufficient evidence for their primary recommendation and concluded that there was ‘…no reliable evidence of a difference in the effectiveness between mechanical and pharmacological methods of prophylaxis’ (page 82 of report). In practice, the usual mechanical method will be graded pressure stockings and the NICE guideline states that thigh-length graduated stockings are preferred to calf length. However, a recent Health Technology Appraisal concluded that it was not possible to assess whether above-knee compression stockings were more effective than below-knee (Roderick et al, 2005). Furthermore, tolerance and compliance with optimal fitting and continued effective use of stockings is a significant problem and is generally worse with thigh-length compared with calf-length stockings (Byrne, 2001). In trying to assure efficacy, a once daily injection has considerable advantages over the use of compression stockings. With increasing patient and public awareness there will also be an increasing expectation and a request for explanation when venous thrombosis occurs despite prophylaxis. In addressing legitimate patient concerns and responding to potential litigation, clear documentation of prescription and compliance is essential for patient reassurance and satisfactory resolution. Documentation of daily stocking use, particularly the degree of compliance, is very poor. In contrast, compliance with once daily subcutaneous injections of LMWH is easily confirmed by reference to the drug chart.
Age is a continuous variable associated with an increasing risk of venous thrombosis following surgery. The British Committee for Standards in Haematology (BCSH) raised concerns that an age threshold of 60 years, rather than 40, is too high for defining high risk (NICE 2007b). Two studies were cited by NICE in their response to the BCSH comments on age as justification for not adopting an age threshold of 40 years (Anderson & Spencer, 2003; White et al, 2004). In the study of the effect of age on incidence of VTE after major surgery the risk increased steadily between the ages of 40 and 60 years and levelled off over 65 years (White et al, 2004). Therefore, a threshold of 60 years only identifies patients in the very highest risk category by virtue of age. The second citation was a review that concluded ‘patients >40 years of age are at significantly increased risk compared to younger patients and risk approximately doubles with each subsequent decade’ (Anderson & Spencer, 2003).
NICE intend to publish a guideline in 2009 on ‘The prevention of venous thromboembolism in all hospital patients’, for which stakeholder comments have been submitted (NICE 2007c). The ACCP is also preparing a new version of guidance on thromboprophylaxis.
Risk and prophylaxis in elective orthopaedic surgery
Lower limb joint replacement is associated with a risk of VTE but the risk appears to be lower nowadays. Nevertheless, a risk of fatal PE remains but the risk is now considered to be so low by some orthopaedic surgeons that they do not routinely give pharmacological thromboprophylaxis whilst some give only aspirin. The Pulmonary Embolism Prevention (PEP) Trial was published in 2000 and reported that 56% of hip fracture patients and 63% of elective joint replacement patients did not receive any form of heparin prophylaxis [Pulmonary Embolism Prevention (PEP) Trial Collaborative Group 2000].
Some clinicians are understandably confused by studies of thromboprophylaxis in patients undergoing orthopaedic surgery. This is largely because the primary endpoint of these studies is asymptomatic DVT detected by screening and not symptomatic VTE or fatal PE. Whilst asymptomatic DVT is found frequently, only a small proportion of patients suffer symptomatic PE and very few patients die as a consequence. There is strong evidence that asymptomatic VTE is an indicator of relative risk of symptomatic VTE and also fatal PE (Cohen et al, 2001). However, it is necessary to consider the absolute incidence of events as well as the relative risk reduction that may result from prophylaxis. A favourable relative risk reduction is not sufficient evidence for adopting an intervention in routine clinical care. The decision to intervene is dependent on the absolute risk reduction and this must be balanced against the absolute risk of an adverse event due to the intervention. For example if an intervention reduced the risk of an adverse event by 50% it might be considered beneficial. If the incidence of the adverse event was 2 in 1000 then the intervention would reduce the event rate to 1 in 1000, assuming all patients received the intervention. To prevent one adverse event 1000 patients would have to be treated. If the intervention was associated with a 1% risk of a major complication and 10% of patients who suffered this complication died then if 1000 patients were treated 10 patients would have this major complication and one would die. Therefore, the analysis of benefit and risk would indicate that if 1000 patients were treated one death due to the adverse event would be prevented but at the price of one death due to a complication of the intervention. Furthermore, another nine patients would suffer the complication but would not necessarily die. If an alternative example were chosen in which there was an overall benefit with the intervention then a cost-analysis would have to be performed.
The efficacy of thromboprophylaxis has been assessed in lower limb orthopaedic surgery by measurement of asymptomatic VTE identified by contrast venography or ultrasound examination. Experts consider asymptomatic VTE as a useful surrogate marker of risk of fatal PE (Cohen et al, 2001). They also point out that the aim of thromboprophylaxis is not only to reduce the incidence of fatal PE but also non-fatal VTE, as this may result in significant morbidity, will require treatment with anticoagulant drugs at doses that result in significant bleeding risk, delay hospital discharge, result in hospital re-admission and cause the post-thrombotic syndrome, which directly produces morbidity as a consequence of venous insufficiency and also increases the risk of subsequent venous thrombosis. It is difficult to accurately measure PE as a cause of death in clinical studies. Even in autopsy-based studies many patients do not have an autopsy. In the absence of an autopsy cardiac failure is likely to be overdiagnosed and PE underdiagnosed. This is suggested by the two-fold lower rate of fatal PE in observational studies compared to autopsy studies (Dahl et al, 2005). Whilst accepting that asymptomatic VTE may be an indicator of relative risk for symptomatic VTE, including fatal PE, the absolute rates of asymptomatic VTE, symptomatic DVT, symptomatic PE and fatal PE are considerably different. Large studies are required to demonstrate an effect on symptomatic PE, and even more for an effect on fatal-PE. More than 14 000 patients were included in the meta-analysis of effect of low dose heparin which demonstrated a significant reduction in fatal PE in combined general, urology and orthopaedic surgery patients (Collins et al, 1988). In orthopaedic patients specifically (elective and trauma combined) the relative reduction in fatal PE was similar to that observed in general surgical patients. In orthopaedic surgery patients there were five deaths attributed to PE in 826 patients receiving heparin and 15 deaths in 801 control/placebo patients equating to rates of fatal PE of 0·6% and 1·9% respectively [odds ratio 0·32, 95% confidence interval (CI) 0·12–0·89]. This is statistically significant. If only patients undergoing elective orthopaedic surgery are included, there were zero deaths attributed to PE in 506 patients receiving heparin and eight deaths in 498 control/placebo patients giving rates of fatal PE of 0% and 1·6% (odds ratio if one death assigned to heparin group 0·12, 95%CI 0·02–0·98). This is statistically significant. This large meta-analysis indicated a significant reduction in fatal PE in patients undergoing elective hip and knee surgery who received low dose heparin (relative reduction >50% with absolute reduction of about 10–15 per 1000 patients treated).
The findings of Collins et al (1988) relate to PE rates experienced in the 1970s (12/13 studies were reported before 1980). Therefore, whilst there is no evidence to indicate that the relative risk reduction with low dose heparin has changed, it is likely that with faster surgical techniques (Sharrock et al, 1992), regional anaesthesia (Prins & Hirsh, 1990) and early mobilisation (Pearse et al, 2007) the absolute risk reduction is now lower. Sheppeard et al (1981) reported the results of an autopsy study to determine the death rate attributable to PE in patients who did not receive prophylaxis for elective hip replacement in a single institute. Nineteen of 3016 patients died, with 11 due to PE, giving a fatal PE incidence of 0·36% (Sheppeard et al, 1981). In 1995 another single institute reported a fatal PE rate of 0·34% (95%CI 0·09–0·88) in 1162 hip replacement patients who had not received prophylaxis (Warwick et al, 1995). The total symptomatic VTE rate was 3·4% (95%CI 2·5–4·7%). A meta-analysis published in 1996 determined mortality after hip replacement in 93 000 patients in studies in which data on total mortality and fatal PE were reported but no attempt was made to restrict the search to controlled trials or even prospective studies (Murray et al, 1996). Cause of death was determined from certification and not post mortem examination. Given the methodology, the fatal PE rate was predictably lower (0·2% or less) than that reported from the single institute studies and risk was probably underestimated. The absolute and relative rates of VTE can be estimated using the fatal PE rates from the single centre studies with modern lower limb joint replacement in patients not receiving prophylaxis. The absolute rate of symptomatic VTE is about 3·5%, symptomatic PE 1% and fatal PE about 0·35%. Therefore, the relative risk of symptomatic VTE is about 10% of the asymptomatic rate and the fatal PE rate is about 10% of the symptomatic VTE rate, and 1% of the asymptomatic VTE rate.
A systematic review of autopsy studies reporting fatal vascular events after orthopaedic surgery was published by Dahl et al (2005). Autopsy rates were more than 60% in the studies identified. Total mortality was lower at 0·57% (95%CI 0·51–0·62%) in elective joint replacement patients receiving prophylaxis compared to those having no prophylaxis, in whom it was 0·93% (95%CI 0·57–1·29%). In elective surgery patients PE accounted for as much as 85% of the mortality. In a recent study of outcomes in more than 1000 patients, 99% of whom received an injectable antithrombotic drug and 95% received this for more than 21 d, the symptomatic VTE rate was 1·8% (95%CI 1·0–2·6%) and there were no fatal emboli (Samama et al, 2007). A UK hip registry recently reported a rate of VTE of 4% with a fatal PE rate of 0·1% after hip replacement with the majority (>90%) having received prophylaxis (Williams et al, 2002). However death rates were according to certification and not post mortem examination.
Thromboprophylaxis with low dose LMWH reduces the risk of symptomatic VTE in patients undergoing lower limb joint replacement by at least 50%, and possibly by as much as 65% (Freedman et al, 2000). The number needed to treat (NNT) to prevent one event can be calculated based on estimates of VTE incidence without prophylaxis (Table II). Working on a 65% risk reduction treatment of 1000 patients for 10 d would prevent 23 symptomatic VTE, including 11 PE, and would save two to three lives. The issue for orthopaedic surgeons and patients is whether the absolute benefit in terms of reduced thrombosis is offset by potential complications of prophylaxis, particularly the need to replace the prosthesis as a result of either loosening or infection. When considering this it is important to distinguish the benefit of reduced mortality and reduced morbidity. For example a complication of prophylaxis might be considered to be of lower relative importance when the analysis is in terms of reduced mortality due to PE. When the benefit is a reduced incidence of non-fatal VTE then a complication of prophylaxis might be considered to be of relatively higher importance. A study of only 2500–3000 patients would be needed to determine the effect of low dose LMWH on symptomatic VTE and would allow measurement of relevant adverse outcomes including infection and prosthesis loosening due to haematoma and a direct comparison of hospital discharge and readmission rates, speed of recovery and incidence of post thrombotic syndrome at 6 months. Such a study would indicate if there is a net benefit on morbidity from prophylaxis in patients undergoing elective lower limb joint replacement. There are approximately 60 000 joint replacements in England annually (http://www.dh.gov.uk/en/Publicationsandstatistics/Pressreleases/DH_4007248) and so recruitment to such a study could be completed within a few months.
Table II. Calculation of number of patients needed to be treated to prevent one episode of venous thrombosis.
|Event||Event rate in absence of prophylaxis*(%)||NNT calculation 50% reduction with prophylaxis||NNT 50% reduction with prophylaxis||NNT calculation 70% reduction with prophylaxis||NNT 70% reduction with prophylaxis|
Aspirin and the PEP study
The use of aspirin as thromboprophylaxis in orthopaedic patients was promoted by the report of the Pulmonary Embolism Prevention (PEP) Trial (2000). The interpretation of the results of the study by the PEP Group was that aspirin reduced the risk of PE and DVT by at least a third. However, rates of VTE were the same in elective surgery patients receiving aspirin or placebo. In hip fracture patients fatal PE was reduced by 68% (odds ratio 0·43, 95%CI 0·24–0·73). The study was criticised in relation to design, analysis of primary endpoints and under-reporting of bleeding events in patients receiving aspirin (Cohen & Quinlan, 2000) with the authors additionally concluding ‘the main outcomes of the trial are that aspirin did not reduce vascular deaths, had no significant effect on major non-fatal vascular events other than deep vein thrombosis, but did result in an excess of 6 per 1000 postoperative transfused bleeds’ (Quinlan & Cohen, 2000). Any beneficial effect of aspirin on the rate of venous thrombosis may have been related to the duration of treatment, as patients were randomised to either aspirin for 35 days or no treatment after discharge from hospital and so the aspirin group effectively received extended thromboprophylaxis. It is likely that aspirin does reduce the incidence of all venous thrombosis by as much as 25% (Hovens et al, 2006) but this efficacy is inferior to low dose heparin. Furthermore, the bleeding risk with aspirin appears greater. In the PEP hip fracture patients given aspirin there was an excess of fatal bleeds and haematemesis and melaena (odds ratio 1·5, 95%CI 1·2–1·8) with an excess of 10 bleeds per 1000 patients treated. In summary, aspirin probably has a beneficial effect with a reduction in VTE but this effect is inferior to low dose heparin and LMWH in particular is not associated with a high risk of major bleeding.
Fondaparinux is a synthetic indirect factor Xa inhibitor with a similar mode of action to LMWH. Fondaparinux may be more effective than LMWH in terms of VTE risk reduction (Lassen et al, 2002; NICE 2007a) but is associated with a greater risk of major bleeding (NICE 2007a, particularly if given within 6 h of surgery. In a costing analysis, NICE (2007a) concluded that fondaparinux was cost-effective or cost-saving compared to LMWH. A major advantage of fondaparinux is the absence of heparin-induced thrombocytopenia with thrombosis (HITT). This complication of heparin necessitates regular monitoring of the platelet count (Keeling et al, 2006), which will be a major disincentive to extended duration thromboprophylaxis. A disadvantage is the relatively long duration of action and hence bleeding risk associated with spinal anaesthesia preoperatively. A pragmatic solution that might promote compliance with NICE guideline 46 would be to initiate fondaparinux postoperatively in elective joint replacement surgery and continue to day 28. In hip fracture patients where surgery may be delayed for 24–48 h LMWH could be started on admission with a switch to fondaparinux on the third postoperative day and continued to day 28. This would avoid the early risk of spinal haematoma in all patients and intra-operative bleeding in elective surgery, as well as reduce the risk of early postoperative bleeding and the later complication of HITT in all patients. According to the NICE analysis it would also be cost-saving due to a reduction in venous thrombosis as compared with no prophylaxis or LMWH alone.
The role of the haematologist
There is now recognition by Government and the Department of Health of the need to reduce the incidence of venous thrombosis in hospitalised patients. This will be achieved by individual patient risk assessment on admission and targeting of prophylaxis at moderate and high risk patients. Whilst this may appear to be a seemingly simple process, a reduction in the burden of disease will require a concerted effort. How risk assessment will be performed, by whom and at what point(s) during hospitalisation will have to be considered. The most appropriate form of prophylaxis for different patient groups, based on clinical effectiveness and cost will have to be defined (NICE 2007a,d). Whilst haematologists will not be personally responsible for standards of care for individual patients (in relation to prevention of venous thrombosis) it is likely that, in most hospitals, they will engage in the process of developing local policies and procedures. In many hospitals haematologists will be identified as the appropriate lead clinician for implementing guidance and recommendations. As new antithrombotic drugs become available, such as dabigatran and rivaroxaban, there will be an opportunity to influence standards of care through consultation with haematologists at local level and the British Committee for Standards in Haematology at national level.