The diagnosis of deep vein thrombosis in symptomatic outpatients and the potential for clinical assessment and D‐dimer assays to reduce the need for diagnostic imaging

Abstract

Deep vein thrombosis (DVT) has an annual incidence of between 48 and 182 per 100 000 (Coon et al, 1973; Anderson et al, 1991; Nordstrom et al, 1992; Hansson et al, 1997; Silverstein et al, 1998), so the oft quoted figure of 1 per 1000 is a reasonable estimate. Estimates of the case fatality rate range from 1% to 5% (Anderson et al, 1991). However, the incidence and the case fatality rate are very age dependent (Coon et al, 1973; Anderson et al, 1991; Nordstrom et al, 1992; Silverstein et al, 1998). There is also associated morbidity. Post-thrombotic syndrome, characterized by chronic pain, swelling and occasional ulceration of the skin of the leg occurs in up to one-third of patients who have had a DVT (Prandoni et al, 1996, 1999). The post-thrombotic syndrome can occur early or have a latency of up to 10 years, the cumulative frequency has been estimated as 23% at 2 years and 28% at 5 years (Prandoni et al, 1996). In patients who use elastic compression stockings for at least 2 years, the incidence of post-thrombotic leg can be halved (Brandjes et al, 1997). Increased awareness of DVT and its consequences has resulted in an increased number of patients being referred for assessment. Many of these patients are at low risk and this is reflected by the fall in the percentage of positive diagnoses of DVT in reported series. The objective diagnosis of DVT depends on imaging using compression ultrasound or ascending venography. However, because of the cost of these modalities, the increasing number of negative tests which are being requested and the resultant, frequently incurred delays in access to them, alternative approaches to diagnosis and decision making in suspected cases of DVT have been adopted. These rely on the use of information from clinical history and examination and assays to detect D-dimers. The main emphasis of these methods is on the safe exclusion of a diagnosis of DVT, thus reducing the use of imaging techniques and speeding up the diagnostic process. While clinical examination cannot be relied upon in isolation to make a diagnosis of DVT, in combination with appropriate history taking, it can provide useful information (Wells et al, 1995, 1997). Recently, sensitive D-dimer assays that can help to exclude the diagnosis have been developed. This guideline describes the use of these methods alone and in combination with each other to develop strategies that safely exclude the diagnosis in patients presenting with suspected DVT, without resorting to the use of diagnostic imaging. An understanding of the natural history of calf vein thrombosis, and the risk of extension and embolization is also important in designing a diagnostic strategy. The guideline was drafted by a working party of the Thrombosis and Haemostasis Task Force of the British Committee for Standards in Haematology. Information was gathered from several sources. These include references known to the working party members supported by a search of MEDLINE using the terms DVT, venous thrombosis, venous thromboembolism (VTE) and D-dimer(s). Recommendations are graded according to the level of evidence (Appendix 1). Before the 1970s, the diagnosis of DVT was often made on clinical grounds. The use of venography showed that the clinical diagnosis was often incorrect. However, clinical assessment giving an estimate of the pretest probability of disease does have a role. Wells et al (1997) validated a system combining symptoms, risk factors, signs and possible alternative diagnosis to stratify patients into low, moderate or high pretest probability. Their analysis of multiple variables resulted in the scoring system shown in Table I. In this study 3%, 17% and 75% of the patients with low, moderate and high pretest probability, respectively, had DVT. They did not include patients with previous VTE nor pregnant women and we, therefore, recommend that all these patients should have diagnostic imaging. Fibrinogen structure, and fibrin and D-dimer formation have been reviewed by Gaffney (1987) and Hantgan et al (1994). Fibrinogen is a large (340 kD) glycoprotein consisting of three pairs of disulphide bonded polypeptide chains (α-, β- and γ-chains). It has a central, N-terminal, E-domain, containing numerous disulphide bridges and two outer D-domains. Cleavage of a small peptide (FpA) from the E-domain of the α-chain of adjacent fibrinogen molecules results in fibrin monomer formation. These monomers (desAA fibrin, or fibrin I) are then able to polymerize, eventually forming insoluble fibrin. Cleavage of fibrinopeptide B, from the E-domain β-chain (to form desAABB fibrin, fibrin II), is not essential for fibrin polymerization, but appears to influence the conformation and stability of the fibrin and its degradation products. Polymerization involves interactions between the D- and E-domains of fibrin subunits in parallel strands, causing a half stagger arrangement (Fig 1), and interactions between the D-domains of adjacent subunits. Activated factor XIII catalyses the formation of Glu–Lys cross-links between the γ-chains (γ–γ links) and the α-chains (α–α links) of adjacent fibrin units. These interactions stabilize the fibrin polymer. Plasmin cleaves the fibrin subunits in a polymer in random order, to yield soluble fragments with a range of molecular weights (X-oligomers). The initial cleavage occurs at the carboxyl terminus of the α-chains, so that the α–α links are lost (Fig 2). Further cleavage by plasmin causes the progressive breakdown of X-oligomers to yield a variety of fragments. In each of these breakdown steps, the γ–γ links are retained, so that D-dimers are present. Thus, D-dimer is not merely a single substance, DD, but is present in the various X-oligomers, and as EDD (E-domain + D-dimer). In addition, there are multimeric structures due to intermolecular γ-chain crosslinks, to form trimers and tetramers of fragment D (DDD, DDDD and intermediate products containing these structures) (Mosesson, 1995). Cross-linked split products can also be generated by neutrophil elastase and matrix metalloproteases, as well as other proteases, and these may cross-react in D-dimer assays. Fibrin polymerization, cross-linking and plasmin digestion. The ovals represent the fragment D domains, consisting of γ and β chains. The boxes represent the E domains, consisting of α, β and γ chains, linked together by multiple disulphide bonds, in the central portion of the fibrin molecule. The solid double lines are the cross-links between gamma chains, and the dotted lines are the polymerization sites. The formation of D-dimer containing products. The polypeptide chains of three cross-linked fibrin monomers in a fibrin polymer and the simplest plasmin-mediated degradation products (D-dimer and fragment E) are depicted. The cross-linked γ chain dimers (crosses), the cross-linked α chain polymers (wavy lines) and the non-cross-linked β chains are shown. The γ chain cross-links cause D fragments from adjacent fibrin monomers to remain associated as D-dimers, following plasmin digestion. Reprinted from Gaffney et al (1976)© with permission from Elsevier Science. A variety of different qualitative and quantifiable assays are available for D-dimer (Nieuwenhuizen & Bos, 1999), and all are based on the use of monoclonal antibodies (Table II). The techniques used include turbidimetry, latex particle agglutination, fluorescence immunoassay, immunofiltration tests and enzyme-linked immunosorbent assay (ELISA). These principles have been incorporated in a variety of automated techniques. There is wide variation in performance, but the ELISA methods are generally more sensitive than the latex agglutination techniques. There are discrepancies in the comparability of the various assays, particularly in terms of normal reference ranges and clinical cut-off values for the exclusion of thrombosis. One explanation for this is the use of a variety of combinations of monoclonal antibodies with differing specificity and affinity. This means that the various forms of D-dimer described above will be detected to differing extents by different assay reagents. The use of different commercial calibrants also influences assay performance. These vary in the method of preparation, solution in which they are dissolved, and the source of primary calibrant, as well as the units of measurement [i.e. μg/ml, fibrinogen equivalent units (FEU)]. The cut-off value for the exclusion of DVT is often set at approximately 500 μg/l or 500 μg/l FEU. Two microgram per litre FEU has an immunoreactivity of 1 μg/l of purified D-dimer, but often it is unclear which units the manufacturer is referring to. Some calibrants are prepared from purified fibrinogen after conversion to fibrin, and cleavage with plasmin; depending on the exact reaction conditions, different spectra of D-dimer containing products are obtained. Some preparations are partially purified by the use of molecular size exclusion techniques, so that only particular fractions of fibrin degradation product are retained. Other calibrants are prepared from citrated plasma instead of purified fibrinogen. As the expression of the D-dimer moiety in clinical samples is heterogeneous, and may differ depending on the nature of the patient, the use of highly purified D-dimer standard preparations has been unsuccessful. The best agreement between assays has been obtained using reference plasma preparations prepared from pools of clinical samples (Nieuwenhuizen, 1997), and further work is under way by the Fibrinolysis Scientific Subcommittee of the International Society for Thrombosis and Haemostasis, to investigate the possible preparation of an International Reference Preparation. It is hoped that this may be utilized by manufacturers and lead to at least partial harmonization of D-dimer results between different methods. Citrated plasma is the sample type of choice for D-dimer assays, as some X-oligomers may be consumed in clot formation, and various fibrin degradation products may be adsorbed to clots. Thus, serum samples could lead to false low or negative values. Thrombus formation is normally followed by an immediate fibrinolytic response. The resultant generation of plasmin causes the release of fibrin degradation products (predominantly containing D-dimer) into the circulation. It follows that absence of a rise in D-dimers implies that thrombosis is not occurring. Thus, negative D-dimer assays may have a role in excluding a diagnosis of VTE. The strategy for using D-dimer assays in the diagnosis of DVT, is to employ a sensitive test with a high negative predictive value (NPV). The specificity of the assays for DVT is variable and depends on the patient population. False positive results are common in hospital inpatients, particularly in patients with infection and cancer, and in postsurgery patients. False positive results are so common in elderly patients that some investigators have suggested raising the cut-off value of D-dimer assays for use in this group. As the cut-off value for defining a negative test is lowered, the sensitivity will rise (and the specificity will fall) but fewer patients will have the diagnosis excluded based on the result of the test. Initially the only tests sensitive enough to be used in this way were ELISAs. However, these assays are time consuming and were designed to batch samples in a research setting. The slide-based latex bead agglutination assays used to detect raised D-dimers in conditions such as disseminated intravascular coagulation are not suitable (Bounameaux et al, 1997). Now many rapid ELISAs for single use are available. For example, the VIDAS D-dimer has been the subject of several publications (D'Angelo et al, 1996; Borg et al, 1997; Janssen et al, 1997a) and a review combined these studies to give an estimated sensitivity of 98% (94–100%) and specificity 54% (47–62%) (Bounameaux et al, 1997). More recently modified methods in which latex bead agglutination is measured using the photo-optical detection system of an automated coagulometer have been developed. The rate of agglutination is proportional to the D-dimer concentration and can be interpolated from a reference curve, giving a sensitive method and a quantifiable result. A study has compared 13 methods in the same 99 patients with suspected DVT (van der Graaf et al, 2000); results are shown in Table In a E different quantifiable assays were used and a variation in was A of a sample with raised D-dimers as The analysis from the of D-dimer et al, has shown that some assays are for the exclusion of DVT. commercial from used in the were on various using samples from patients presenting in with a suspected diagnosis of pretest probability scoring and objective imaging were Some methods had very all manufacturers cut-off values for the exclusion of DVT and analysis of that the suggested cut-off values were not cut-off values giving a sensitivity level by analysis were between and of patients without VTE were excluded on the of D-dimer assays. of the results from the et al, with of der Graaf et al the in that occurs are used on different There are several other that to the and of D-dimer assays. of the studies vary in the time between the of and the of D-dimer et al that D-dimer to normal after a DVT. While der Graaf et al patients up to after the of of DVT, have with a of and in measured D-dimer with time (D'Angelo et al, 1996). In this the results of after of had a on the value in patients with DVT. of the of the at the time of calf vein thrombosis of all DVT. the of D-dimer assays in and DVT The reported of qualitative assays from 17% to for calf DVT compared with for all DVT (van der Graaf et al, of a variety of methods has also been for calf thrombosis et al, 1997; der Graaf et al, that the of in the it is that these to the age of the it is that the detection of these by D-dimer assays to of In DVT is but access to a diagnostic is to be it is common to the diagnosis is While this is highly to the of imaging it may the of D-dimer assays. studies have shown a fall in D-dimer following with et al, Janssen et al, et al, 1999). et al a fall in D-dimer values in patients that were after with or low molecular for VTE. the D-dimer level on 1 in patients on on the time of with detected at of et al, 1999). In a study of patients with for at least 5 in D-dimer were that have the of the assay as a test for excluding the diagnosis of VTE et al, There are on the of a single of on the of D-dimer assays for exclusion of a diagnosis of DVT. recommend an assay with high sensitivity in studies a patient to that in which it will be The assay should be validated in each by an of its The suitable cut-off value for negative of DVT be to differ for the various cut-off samples should be from patients presenting with suspected DVT. In this should be made based on a standard diagnostic of the D-dimer result. The D-dimer test should be with the to the clinical and The size of the study should be by but be to include at least patients. The use of analysis techniques to sensitivity and specificity values at various D-dimer may then be used to the cut-off value for the in that clinical setting. It is not to this at and may often have to rely on the which should be as described above D-dimers be used with the patient has had for 2 for D-dimer are by and this be in using these assays to exclude DVT. D-dimer assays should be on that is to with excluded the diagnosis in some patients suspected of DVT, using clinical scoring and D-dimer a of patients further The in the diagnostic of DVT is some form of imaging. with suspected venous thrombosis not have the diagnosis by objective in of patients the who have venous thrombosis, have vein thrombosis, and the have thrombosis to the calf et al, tests are available the in Table should be into The diagnostic test for DVT of the is ascending venography. can detect the calf and the and However, venography is of a vein and of of is also in terms of and time and is a choice for of the to a can be high in patients with or in in the system et al, et al, is also high with values of et al, et al, 1992; et al, and between and for DVT et al, in the may be as the of from the of the to of normal studies the or may not the of in with DVT Other include the which are not in up to 75% of an is used & Many have shown the value of ultrasound in the diagnosis of DVT. et al, et al, et al, The to the vein is the diagnostic and other not the sensitivity of the method et al, 1995). A a sensitivity of for DVT with a sensitivity for DVT of to only for DVT et al, 1998). The of compression ultrasound is and by on patients with a high probability of DVT, the agreement was high et al, However, agreement calf DVT et al, 1996). only is the diagnosis of DVT between two but the of and extension of is also well et al, The rate for compression ultrasound is et al, et al, 1996). An venography is the of ultrasound to alternative diagnoses of leg and Deep vein thrombosis in the calf et al, & et al, but by the time develop of patients have in the or more et al, et al, It has been that in patients presenting with calf DVT approximately of into the et al, & 1 However, a more study an extension rate of only et al, DVT that not lead to et al, et al, but in that the risk of is et al, et al, This has the for using The test will detect thrombosis, a calf vein may remain but a 1 will up the important that have It is safe to from patients with suspected DVT who have normal results on compression at the time of and at 1 et al, The studies that validated clinical scoring excluded patients who had had a previous DVT. these patients are at high risk of and should all have diagnostic imaging The diagnosis of DVT can be as the normal relied upon to the or absence of may be on compression ultrasound may be detected in up to of patients evidence of disease in the following a DVT (Prandoni et al, venography of patients with previous DVT may a to but can also be detected or partial and which may the diagnosis of disease et al, However, thrombosis is a clinical with the cumulative risk of being 17% at 2 years, to at 5 years and at years (Prandoni 1997). While venography is generally as the diagnostic test in some studies have suggested that ultrasound may be or more of will often rely on a combination of these two tests & As with the for time DVT ultrasound and D-dimer are negative the patient can be safely However, this is to to the of patients. A further is to a examination with previous available (Prandoni et al, there has been and D-dimer is could be to In the patient has a positive D-dimer or previous are to be to for that are of thrombosis. The can clinical assessment and D-dimer be used the for diagnostic imaging in a number of with suspected DVT, ultrasound is can the results of D-dimer assays be used to the for It has been suggested that an initial negative D-dimer test can be used to the diagnosis of DVT of the pretest clinical probability assessment et al, 1999). for example, the D-dimer test has a sensitivity of and a specificity of and to the pretest probability the in Table The for the moderate probability is in More information on is available at and this has a that can be used to In this example, the D-dimer the diagnosis in with a low pretest probability but it does not in with a high pretest probability. a negative test is enough to the diagnosis in with a moderate pretest probability is not from this Some studies that the tests with the sensitivity can be used to in patients that have a low or moderate pretest probability (van der Graaf et al, et al, However, each should for the assay they in It has been shown that it is safe to with the ultrasound after a negative ultrasound in patients with a low pretest probability (Wells et al, 1995, 1997). It has been shown that it is safe to with the ultrasound after a negative ultrasound the D-dimers are also negative et al, 1998). It is that all strategies for the diagnosis of DVT will result in cases being The being be in to best imaging techniques are to all suspected cases of DVT and to the of and from In a a combined clinical scoring system and D-dimer assay results in a risk of DVT, it is appropriate to that this is equivalent to compression that has a of for thrombosis. the case from DVT was a strategy that has an of result in for suspected cases in which the diagnosis was to have been In patients suspected of a DVT, clinical pretest probability assessment and D-dimers can be used to the for diagnostic imaging A low pretest probability and negative D-dimers the diagnosis without for diagnostic imaging The of negative D-dimer results to exclude the diagnosis in patients with moderate pretest probability is on the sensitivity of the test This assessment and D-dimers should not be used alone to exclude the diagnosis in patients with a high pretest probability A low pretest probability and negative initial ultrasound the diagnosis without for ultrasound or venography D-dimers and a negative initial ultrasound the diagnosis without for ultrasound or venography Many can be designed according to these One way they can vary is by the in which tests are in the different clinical will have to according to and patient population. on the evidence in this possible is shown in One possible diagnostic for patients with a suspected vein thrombosis based on the with a moderate pretest probability could the of the low probability patients a sensitive D-dimer test is used and assessment this to be they should all have an initial ultrasound as for the high probability patients. work is to the for D-dimer assays. to and for the diagnosis of DVT such as and imaging will assessment. these can imaging for DVT with imaging for they may be The diagnosis of and approaches could be shown to estimate clot diagnosis be the and information in these is to be and at the time of to the nor the can for or that may have been The Haemostasis and Thrombosis and will review these occur or by in at the The and Task Force to the and on this made by members of the British Society for Haematology. A after the negative test can estimate the probability of DVT to be