C-reactive protein and D-dimer with clinical probability score in the exclusion of pulmonary embolism

research paper C-reactive protein and D-dimer with clinical probability score in the exclusion of pulmonary embolism N. Steeghs, 1 R. J. Goekoop, 1 R. W. L. M. Niessen, 2 G. J. P. M. Jonkers, 3 H. Dik 4 and M. V. Huisman 1 1 Department of General Internal Medicine, Leiden University Medical Centre, Leiden, 2 Department of Clinical Chemistry, Rijnland Hospital, Leiderdorp, 3 Department of Internal Medicine, Rijnland Hospital, Leiderdorp, and 4 Department of Pulmonary Medicine, Rijnland Hospital, Leiderdorp, the Netherlands Received 23 March 2005; accepted for publication 8 June 2005 Correspondence: N. Steeghs, Department of General Internal Medicine C1-R41, Leiden University Medical Centre, Albinusdreef 2, PO Box 9600, 2300 RC Leiden, the Netherlands. E-mail: n.steeghs@lumc.nl Summary This study evaluated the diagnostic value of C-reactive protein (CRP) combined with a clinical decision rule in the exclusion of pulmonary embolism () and compared this with D-dimer. In 363 consecutive outpatients CRP and D-dimer test were performed and clinical probability of was assessed. Patients with D-dimer levels <500 lg/l and clinical probability indicating unlikely were followed for 3 months. Ventilation-perfusion scan or spiral computerized tomography was performed in patients with D-dimer levels 500 lg/l or clinical probability indicating likely. The CRP had a sensitivity of 95Æ7% [95% confidence interval (CI): 90 100] and negative predictive value (NPV) of 98Æ4% (96 100). CRP <5 mg/l with clinical probability score indicating unlikely (n ¼ 108, 30%), had a sensitivity of 96Æ7% (90 100), a specificity of 43Æ0% (37 49) and NPV of 99Æ1% (97 100). D-dimer <500 lg/l with clinical probability score indicating unlikely (n ¼ 170, 51%), had a sensitivity of 96Æ7% (90 100), a specificity of 67Æ9% (62 74) and NPV of 99Æ4% (98 100). Based on retrospective data it was concluded that a standard CRP test can potentially be used to safely exclude, either as a sole test or combined with clinical probability assessment. Prospective studies are needed to confirm these findings. Keywords: clinical probability assessment, C-reactive protein, D-dimer, pulmonary embolism. Objective testing for pulmonary embolism () is crucial, because clinical assessment alone was found to be incorrect in up to 50% of cases and the mortality rate of untreated is high (Cooper et al, 1992). Algorithms to rule out or diagnose often involve ventilation-perfusion (V/Q) scans, spiral computerized tomography (CT), or pulmonary angiography, which are invasive, complex and time-consuming. On the contrary, an incorrect diagnosis of and therefore unnecessary treatment with anticoagulant therapy is associated with a risk of bleeding. New simple tests are therefore warranted to rule out and to reduce the number of sophisticated imaging techniques needed. Increased levels of serum C-reactive protein (CRP), a marker of systemic inflammation, have been reported in patients with venous thrombosis and (Roumen-Klappe et al, 2002). The CRP assay is cheap and readily available 24 h a day in most hospitals. Few studies, involving either small numbers of patients or lacking appropriate study methodology, have evaluated the diagnostic utility of CRP measurement in the exclusion of. Results have been diverse, with reported sensitivities of the CRP assay ranging from 60 to 100% depending on cut-off levels of between 5 and 10 mg/l (Thomas et al, 1989; Syrjala et al, 1990; Jossang & Runde, 1992; Franco et al, 1994; Wong et al, 1996; Horney & Lagerstedt, 1998; Maskell & Butland, 2001; Bucek et al, 2002; Aujesky et al, 2003). The specificities in these studies ranged from 52 to 78%. Importantly, the negative predictive values (NPVs) ranged from 76 to 100%. Three studies have evaluated the CRP test in patients with clinical suspected (Cooper et al, 1992; Franco et al, 1994; Aujesky et al, 2003), but only one has evaluated the value of the CRP assay combined with clinical probability assessment (Aujesky et al, 2003). In this study, the CRP alone had a sensitivity of 84% and a NPV of 87% at a CRP cut-off level of 5 mg/l. The NPV increased to 97% when combined doi:10.1111/j.1365-2141.2005.05652.x ª 2005 Blackwell Publishing Ltd, British Journal of Haematology, 130, 614 619

C-reactive Protein in Pulmonary Embolism with a low clinical probability. However, the lower 95% confidence limit (CI) of 89% led the authors to conclude that the CRP test, either alone or in combination with a clinical prediction rule, cannot be used as a safe test to exclude without further testing. In view of the very limited evaluation of CRP, especially in combination with other non-invasive tests, additional studies to validate the use of CRP testing in patients with suspected are needed. The aim of our study was to determine whether CRP alone or in combination with a clinical prediction rule is an alternative to D-dimer as a laboratory marker in the diagnostic work-up of patients with suspected. This study was performed as part of a large prospective trial testing a new diagnostic algorithm for, based on clinical probability assessment and D-dimer test (unpublished observations). Patients and methods Patients Between March 2002 and December 2003 we prospectively screened consecutive patients presenting with suspected to the emergency department or outpatient clinic of a teaching hospital in Leiderdorp, the Netherlands. Inclusion criteria were as follows: (i) presentation to the emergency department or outpatient clinic and (ii) suspected. Exclusion criteria were as follows: (i) age <18 years, (ii) ongoing anticoagulation for >24 h, (ii) pregnancy, (iv) allergy to contrast material, (v) expected survival <3 months, (vi) venous thromboembolism in the previous 6 months and (vii) refusal or inability to consent. Patients with all inclusion criteria and no exclusion criteria were eligible. The hospital s institutional review board approved the protocol. Diagnostic work-up Doctors of the emergency department or outpatient clinic assessed the clinical probability of in all eligible patients, using the clinical decision rule described by Wells et al (1998). A score of 4Æ0 was designated as unlikely and a score of >4Æ0 as likely, based on the previous use of the Wells score in combination with D-dimer testing (Wells et al, 2005). In all patients a common immunoturbidimetric CRP test (Tinaquant, Roche diagnostics, Mannheim, Germany) was performed. In patients with a clinical probability score indicating unlikely a quantitative rapid enzyme-linked immunosorbent assay (ELISA) D-dimer assay (Vidas DD, biomérieux, Marcy l Etoile, France) was performed. The CRP cut-off level was 5 mg/l and the D-dimer cut-off level was 500 lg/l according to previous studies (Ginsberg et al, 1995; Aujesky et al, 2003; Perrier et al, 2004). All analyses were done at admittance on fresh plasma and were performed by technicians, who were unaware of the results of other tests for or clinical status of the patients. A was ruled out in patients with a clinical probability score indicating unlikely and D-dimer level <500 lg/l; these patients did not undergo further diagnostic procedures and anticoagulation treatment was withheld in these cases. In patients with a clinical probability score indicating likely or a D-dimer level 500 lg/l, either V/Q lung scanning or spiral CT was performed. A normal perfusion scan ruled out, and a high probability V/Q scan established the diagnosis (The PIOD Investigators, 1990). In patients with a low/ moderate probability V/Q scan a CT was performed. A CT was positive for if contrast material outlined a central intraluminal defect or if a vessel was totally occluded in at least two different projections (Perrier et al, 2001). All patients with objectively confirmed were hospitalized and treated with anticoagulants. Follow up All patients were followed by their primary care doctors. Patients were instructed to contact their doctor when signs or symptoms of venous thromboembolism recurred. Upon suspicion, objective testing for deep venous thrombosis (compression ultrasonography) or (V/Q lung scanning or spiral CT) was carried out. Patients with a clinical probability score indicating unlikely and D-dimer <500 lg/l were seen in the outpatient clinic after 3 months, or were interviewed by telephone after at least 3 months when a visit to the hospital was not possible. Of the patients who died in the 3-month follow up, the cause of death, including fatal, was determined from the medical record or through contact with the treating doctor. For both the CRP and the D-dimer test the 3-month follow-up information was used as the reference test for the diagnosis of. In patients with a clinical probability score indicating unlikely and D-dimer levels <500 lg/l, an uneventful 3-month follow up was interpreted as confirmation of exclusion of. The CRP testing was not used in the management strategy, but results were prospectively collected. Analysis of the data were performed retrospectively. Statistical analysis Statistical analysis was performed with the Statistical Package for the Social Sciences (spss) version 11.0 (SPSS, Chicago, IL, USA). For the CRP and D-dimer, the test characteristics and predictive values with their 95% confidence intervals (CIs) were calculated according to the clinical probability class. The areas beneath the receiver operating characteristics (ROC) curves were used to compare the discriminative power of the CRP and D-dimer test in diagnosing or ruling out. Results During the investigation period 363 consecutive patients with suspected were identified. Twenty-one patients (5Æ8%) were ª 2005 Blackwell Publishing Ltd, British Journal of Haematology, 130, 614 619 615

N. Steeghs et al excluded because of age <18 years (n ¼ 3), ongoing anticoagulation for >24 h (n ¼ 9), pregnancy (n ¼ 2), allergy to contrast material (n ¼ 2), expected survival <3 month (n ¼ 1), venous thromboembolism in the previous 6 months (n ¼ 5), and refusal or inability to give consent (n ¼ 1). Two patients each had two excluding factors present. In eight patients (2Æ2%) no CRP measurement was performed and one patient (0Æ3%) did not have a D-dimer measurement performed. Two patients were lost to follow up (0Æ6%). Baseline characteristics of the study population of 331 patients are shown in Table I. Forty-seven patients were found to have (14Æ2%). was confirmed by CT in 19 patients (40%), V/Q in 26 patients (55%) and leg ultrasound in one (2%). One patient with symptoms of possible after 3 months follow up refused additional diagnostic procedures and was judged to have had. Nine of 331 patients died (2Æ7%). In one patient had been diagnosed at study entry and this patient was treated with anticoagulant therapy; the cause of death based on postmortem results was malignancy. One patient had a clinical probability score of 1Æ5 and D-dimer of 387 lg/l at entry; this patient committed suicide and had no reported signs or symptoms of before death. In seven patients CT or VQ scanning at study entry ruled out. The clinical probability score was 4 in 279 patients (84%) and >4 in 52 patients (16%). Of the 279 patients with a clinical probability score of 4, 108 patients (39%) had CRP levels <5 mg/l. In this group 94 patients (87%) had a D-dimer <500 lg/l and, according to the study protocol, no further diagnostic tests were performed and anticoagulation was withheld. Follow up was performed by visits to the outpatient clinic in 122 of the 170 patients (72%) and was done by telephone in all other patients. Table I. Patient baseline characteristics (n ¼ 331). Characteristic Number (%) or median (range) Age (years) 51 (18 95) Sex (female) 205 (61Æ9) Risk factors Paralysis or cast within 4 weeks 8 (2Æ4) Immobilization >3 d or surgery 19 (5Æ7) within 1 month Previous DVT or 30 (9Æ1) Immobilization in travel 26 (7Æ9) Known heart failure 10 (3Æ0) Known COPD 44 (13Æ3) Oral contraceptives or 50 (15Æ1) hormone replacement therapy Malignancy 3 (0Æ9) DVT, deep venous thrombosis;, pulmonary embolism; COPD, chronic obstructive pulmonary disease. During the 3-month follow-up period none of the 170 patients with a clinical probability score of 4 and D-dimer <500 lg/l had signs or symptoms of venous thromboembolism for which they contacted the primary care doctor or our hospital. At the 3-month follow-up date two patients had signs or symptoms of possible. In one patient V/Q lung scanning was normal. In the other patient, who refused further diagnostic testing, the clinical probability score had been 0-points, the CRP level 4 mg/l and the D-dimer level 351 lg/l. This patient was not treated with anticoagulants and was alive 1 year after study inclusion; the primary care doctor had diagnosed hyperventilation syndrome. For the study, this patient was judged to have had by lack of a definitive diagnostic exclusion. A total of 14 of the 108 patients with a clinical probability score of 4 and CRP levels <5 mg/l had D-dimer levels of 500 lg/l or more. None of these 14 patients was diagnosed to have by V/Q scan or helical CT scan. The results of the CRP and D-dimer tests according to the clinical risk category and the presence of are shown in Table II. Table III shows the diagnostic performance of the CRP test according to the clinical probability score. A full comparison of CRP and D-dimer data was not possible, while in patients with a clinical probability score of >4 no D-dimer assay was performed. In these high clinical probability patients a considerable chance of a possible was present and therefore V/Q or CT scanning was performed as soon as possible, regardless of laboratory results. The sensitivity for the CRP assay alone with a cut-off level of 5 mg/l was 95Æ7% (95% CI: 90 100); 44 of 46 patients with diagnosed had CRP levels 5 mg/l. Specificity was 43Æ2% (95% CI: 37 49); 123 of 285 patients with ruled out had CRP levels <5 mg/l. The NPV for the CRP assay was 98Æ4% (95% CI: 96 100); of the 125 patients with CRP level <5 mg/l two patients (one in each clinical risk category) were shown to have. When combined with a clinical probability score of 4 the sensitivity increased to 96Æ7% (95% CI: 90 100). Specificity was 43Æ0% (95% CI: 37 49) and NPV increased to 99Æ1% (95% CI: 97 100). The negative likelihood ratio was 12Æ9. The sensitivity for the D-dimer assay with a cut-off level of 500 lg/l when combined with a clinical probability score of 4 was 96Æ7% (95% CI: 90 100); 29 of 30 patients with diagnosed had D-dimer levels 500 lg/l. Specificity was 67Æ9% (95% CI: 62 74); 169 of 249 patients with ruled out had D-dimer levels <500 lg/l. The NPV was 99Æ4% (95% CI: 98 100); of the 170 patients with D-dimer level <500 lg/l one patient was shown to have. The negative likelihood ratio was 20Æ4. The areas beneath the ROC curves, used to compare the discriminative power of the CRP and D-dimer test in the ruling out or diagnosing of were determined (Fig 1). The area under the ROC curve was 82Æ4% (95% CI: 77 88) for the CRP test and 92Æ5% (95% CI: 88 97) for the D-dimer test, implicating that the D-dimer assay was more accurate in ruling out or diagnosing. 616 ª 2005 Blackwell Publishing Ltd, British Journal of Haematology, 130, 614 619

C-reactive Protein in Pulmonary Embolism Table II. C-reactive protein and D-dimer in patients with and without by clinical probability score. C-reactive protein (n ¼ 331) D-dimer (n ¼ 279) <5 mg/l (n ¼ 125) 5 mg/l (n ¼ 206) <500 lg/l (n ¼ 170) 500 lg/l (n ¼ 109) (n ¼ 2) No (n ¼ 123) (n ¼ 44) No (n ¼ 162) (n ¼ 1) No (n ¼ 169) (n ¼ 29) No (n ¼ 80) Clinical probability score Unlikely 1 107 29 142 1 169 29 80 Likely 1 16 15 20 n.m. n.m. n.m. n.m. Values are expressed as number. n.m., not measured;, pulmonary embolism. Table III. Diagnostic performance of C-reactive protein (cut-off level 5 mg/l) by clinical probability score. Parameter, % (95% CI) C-reactive protein All patients (n ¼ 331) unlikely (n ¼ 279) likely (n ¼ 52) Sensitivity 95Æ7 (90 100) 96Æ7 (90 100) 93Æ8 (82 100) Specificity 43Æ2 (37 49) 43Æ0 (37 49) 44Æ4 (28 61) Positive predictive value 21Æ4 (16 27) 17Æ0 (11 23) 42Æ9 (26 59) Negative predictive value 98Æ4 (96 100) 99Æ1 (97 100) 94Æ1 (83 100), pulmonary embolism; CI, confidence interval. Discussion In this retrospective study, we have demonstrated that CRP, using a cut-off level of 5 mg/l, has a high sensitivity for the diagnosis of, comparable with D-dimer. This high sensitivity and high NPV in combination with a clinical probability score are in contrast to a recently published study, in which a more modest NPV of 87% was observed (Aujesky et al, 2003). The observed differences in NPVs are most likely partly due to the result of the lower prevalence of in our study population, i.e. 14% vs. 30% in the previous study (Aujesky et al, 2003). The negative likelihood ratio of the CRP assay was 12Æ9 compared with 20Æ4 for the D-dimer assay. Both the CRP and D-dimer tests had 99% NPV when combined with a clinical probability score. However, the specificity of the CRP test in patients with clinical probability score indicating unlikely was 43Æ0% compared with a specificity of 67Æ9% for the D-dimer test. This indicates that less invasive, time-consuming and expensive additional diagnostic procedures are needed when a D-dimer test is used compared with the CRP test and therefore this seems a better option in our opinion than the use of CRP. This is supported by the fact that there was a clear difference in negative likelihood ratios, suggesting that the high NPV of the CRP assay may indeed have been influenced by the lower prevalence of in our population. However, the CRP assay is cheap and readily available 24 h a day in most hospitals, in comparison with the D-dimer assay. As a full cost-effectiveness analysis was not performed as part of this study, future studies should take this into consideration. Several points need comment First, ultrasensitive CRP assays have recently been developed and can accurately detect CRP levels <5 mg/l. Our results demonstrate, however, that the standard CRP test has a 99Æ1% NPV and ultrasensitive CRP assays are in our view not necessary in the management strategy of excluding. Secondly, CRP levels 5 mg/l are non-specific as CRP levels rise in any inflammatory condition, including bacterial, viral or fungal infections, rheumatoid syndromes, cancer, pregnancy and atherosclerotic disease. The mechanism by which CRP is raised in venous thromboembolism is not yet completely understood. It is suggested that its ability to induce tissue factor expression on monocytes plays an important role (Cermak et al, 1993). Clinicians should be aware that in the management strategy of venous thromboembolism only CRP levels <5 mg/l are indicative of absence of ; CRP levels 5 mg/l do not implicate that is present, only that there is an inflammatory condition present. Thirdly, the question arises whether future studies should consider different cut-off levels for CRP. Our ROC analysis showed that a cut-off level of 6 mg/l would have resulted in a sensitivity of 95Æ7%, equal with the cut-off of 5 mg/l, and specificity of 48Æ1%, slightly higher than with the cut-off of 5 mg/l. Higher cut-off levels would have resulted in sensitivities below 95%, which is not acceptable when dealing with such a possibly fatal condition as. Lower cut-off levels would have resulted in very low specificities, leading to too many false-positive results. Also, given the fact that other studies demonstrated a cut-off level of 5 mg/l as optimal in ª 2005 Blackwell Publishing Ltd, British Journal of Haematology, 130, 614 619 617

N. Steeghs et al A Sensitivity 1 0 0 75 0 5 0 25 0 0 0 0 0 25 0 5 0 75 1 0 1- specificity B 1 0 0 75 that the two patients that were lost to follow up indeed developed, so these results are highly speculative. The strengths of our study include the novel design in diagnostics with CRP testing and clinical probability assessment with a validated rule, the prospective collection of follow-up data, and the independent assessment of outcome. Limitations of our study include the retrospective design, the relatively low prevalence of in comparison with other studies and absence of D-dimer data in patients with a clinical probability score >4. In conclusion, based on retrospective data it is concluded that a standard CRP test using a cut-off level of 5 mg/l can potentially be used to safely exclude, either as a sole test or combined with clinical probability assessment. Prospective studies are needed to confirm these findings Acknowledgements The Regio- Study Group consisted of: Rijnland Hospital, Leiderdorp R.W.L.M. Niessen, H. Dik, G.J.P.M. Jonkers and N. Steeghs; Diaconessen Hospital, Leiden R.W.L.M. Niessen, H.C.J. van Klink and E.V. Planken; Bronovo Hospital, Den Haag Th. Vlasveld, E.A.M. Werker-van Gelder, R. Bieger and A. Castel; Leiden University Medical Centre M.V. Huisman, H. Putter, N. Steeghs and R. J. Goekoop. Sensitivity 0 5 0 25 0 0 0 0 0 25 0 5 0 75 1 0 1-specificity Fig 1. Receiver operating characteristics (ROC) curves. (A) C-reactive protein ROC curve; (B) D-dimer ROC curve. their population, we considered 5 mg/l as the optimum CRP cut-off level. In our opinion these data suggest that more research for other CRP cut-off levels is no longer needed. Fourthly, if the two patients that were lost to follow up were included and had been given a diagnosis of, sensitivities of CRP and D-dimer combined with the clinical probability score would have been 96Æ9 (95% CI: 91 100) and 90Æ6% (95% CI: 81 100), specificities 43Æ0 (95% CI: 37 49) and 67Æ9% (95% CI: 62 74) and NPV 99Æ1 (95% CI: 97 100) and 98Æ3% (95% CI: 96 100). This indicates that CRP could have been a better alternative to D-dimer if a larger population could have been included in the study. However, there is no reason to assume References Aujesky, D., Hayoz, D., Yersin, B., Perrier, A., Barghouth, G., Schnyder, P., Bischof-Delaloye, A. & Cornuz, J. (2003) Exclusion of pulmonary embolism using C-reactive protein and D-dimer. A prospective comparison. Thrombosis and Haemostasis, 90, 1198 1203. Bucek, R.A., Reiter, M., Quehenberger, P. & Minar, E. (2002) C-reactive protein in the diagnosis of deep vein thrombosis. British Journal of Haematology, 119, 385 389. Cermak, J., Key, N.S., Bach, R.R., Balla, J., Jacob, H.S. & Vercellotti, G.M. (1993) C-reactive protein induces human peripheral blood monocytes to synthesize tissue factor. Blood, 82, 513 520. Cooper, T.J., Prothero, D.L., Gillett, M.G., Pallister, C.J., Mitchelmore, A.E., Goddard, P.R., Wallington, T.B. & Hartog, M. (1992) Laboratory investigation in the diagnosis of pulmonary thromboembolism. Quarterly Journal of Medicine, 83, 369 379. Franco, J.A., Gonzalez-Mangers, E. & Butler, T.T. (1994) Negative predictive value of C-reactive protein testing. Journal of Nuclear Medicine, 35, 189 190. Ginsberg, J.S., Wells, P.S., Brill-Edwards, P., Donovan, D., Panju, A., van Beek, E.J. & Patel, A. (1995) Application of a novel and rapid whole blood assay for D-dimer in patients with clinically suspected pulmonary embolism. Thrombosis and Haemostasis, 73, 35 38. Horney, E. & Lagerstedt, C. (1998) Evaluation of analyses to exclude suspected thrombosis. Don t rely on the D-dimer test! Lakartidningen, 95, 55 58. Jossang, B. & Runde, I. (1992) Diagnostic value of C-reactive protein and D-dimer in deep venous thrombosis. Tidsskrift for den Norske laegeforening, 112, 1153 1155. 618 ª 2005 Blackwell Publishing Ltd, British Journal of Haematology, 130, 614 619

C-reactive Protein in Pulmonary Embolism Maskell, N.A. & Butland, R.J. (2001) A normal serum CRP measurement does not exclude deep vein thrombosis. Thrombosis and Haemostasis, 86, 1582 1583. Perrier, A., Howarth, N., Didier, D., Loubeyre, P., Unger, P.F., de Moerloose, P., Slosman, D., Junod, A. & Bounameaux, H. (2001) Performance of helical computed tomography in unselected outpatients with suspected pulmonary embolism. Annals of Internal Medicine, 135, 88 97. Perrier, A., Roy, P.M., Aujesky, D., Chagnon, I., Howarth, N., Gourdier, A.L., Leftheriotis, G., Barghouth, G., Cornuz, J., Hayoz, D. & Bounameaux, H. (2004) Diagnosing pulmonary embolism in outpatients with clinical assessment, D-dimer measurement, venous ultrasound, and helical computed tomography: a multicenter management study. American Journal of Medicine, 116, 291 299. Roumen-Klappe, E.M., den Heijer, M., van Uum, S.H., van der Ven- Jongekrijg, J., van der Graaf, F. & Wollersheim, H. (2002) Inflammatory response in the acute phase of deep vein thrombosis. Journal of Vascular Surgery, 35, 701 706. Syrjala, H., Haukipuro, K. & Kiviniemi, H. (1990) Acute phase response and deep lower limb venous thrombosis. Journal of Clinical Pathology, 43, 519 520. The PIOD Investigators (1990) Value of the ventilation/perfusion scan in acute pulmonary embolism. Results of the prospective investigation of pulmonary embolism diagnosis (PIOD). The PIOD Investigators. Journal of the American Medical Association, 263, 2753 2759. Thomas, E.A., Cobby, M.J., Rhys, D.E., Jeans, W.D. & Whicher, J.T. (1989) Liquid crystal thermography and C reactive protein in the detection of deep venous thrombosis. British Medical Journal, 299, 951 952. Wells, P.S., Ginsberg, J.S., Anderson, D.R., Kearon, C., Gent, M., Turpie, A.G., Bormanis, J., Weitz, J., Chamberlain, M., Bowie, D., Barnes, D. & Hirsh, J. (1998) Use of a clinical model for safe management of patients with suspected pulmonary embolism. Annals of Internal Medicine, 129, 997 1005. Wells, P.S., Anderson, D.R., Rodger, M., Ginsberg, J.S., Kearon, C., Gent, M., Turpie, A.G., Bormanis, J., Weitz, J., Chamberlain, M., Bowie, D., Barnes, D. & Hirsh, J. (2005) Derivation of a simple clinical model to categorize patients probability of pulmonary embolism: increasing the models utility with the SimpliRED D-dimer. Thrombosis and Haemostasis, 83, 416 420. Wong, N.A., Laitt, R.D., Goddard, P.R. & Virjee, J. (1996) Serum C reactive protein does not reliably exclude lower limb deep venous thrombosis. Thrombosis and Haemostasis, 76, 816 817. ª 2005 Blackwell Publishing Ltd, British Journal of Haematology, 130, 614 619 619