C-reactive protein in the diagnosis of deep vein thrombosis

Transcription

1 British Journal of Haematology, 2002, 119, C-reactive protein in the diagnosis of deep vein thrombosis Robert A. Bucek, 1 Markus Reiter, 1 Peter Quehenberger 2 and Erich Minar 1 1 University Clinic of Internal Medicine II, Department of Angiology, and 2 Clinical Institute of Medical and Chemical Laboratory Diagnostics, General Hospital, Vienna, Austria Received 21 March 2002; accepted for publication 5 June 2002 Summary. The role of C-reactive protein (CRP) in the diagnosis of suspected deep vein thrombosis (DVT) and a possible advantage of its additional evaluation with D-dimer has not been clearly evaluated. We therefore studied plasma CRP and D-dimer levels in 233 consecutive patients with suspected DVT; the final diagnosis was based on the results of colour duplex ultrasound or venography. DVT was diagnosed in 31Æ3%. CRP and D-dimer correlated significantly (r ¼ 0Æ64, P <0Æ01); both were increased significantly in patients suffering from DVT (P <0Æ001). Multivariate analysis revealed a significant influence of the presence of DVT (P < 0Æ001), the presence of malignancy (P <0Æ001) and the presence of inflammatory diseases (P ¼ 0Æ009) on plasma CRP, while there was no significant influence of the duration of symptoms (P ¼ 0Æ30). The sensitivity (75% vs 93%) to specificity (69% vs 55%) relationship showed inferior results for CRP compared with D-dimer; its additional evaluation did not improve the diagnostic value of D-dimer. We conclude that CRP can provide additional information neither for the diagnostic process in patients with suspected DVT nor for the differential diagnosis of DVT and inflammatory diseases. Keywords: deep vein thrombosis, C-reactive protein, acutephase reaction, inflammation. Deep vein thrombosis (DVT) is a common disorder with an incidence of approximately 1 per 1000 per year (Nordström et al, 1992, Kahn, 1998). Diagnosing DVT reliably on clinical grounds is a difficult task because several inflammatory diseases, such as cellulitis, present with symptoms of swelling, pain and erythema, which also indicate the presence of DVT. Clinical examination can therefore only classify a pretest probability (Wells et al, 1995, 1997). The gold standard in the diagnosis of DVT is still venography (Wille-Jorgensen et al, 1992), but it is an invasive method, bearing the risk of procedure-related complications and adverse reactions due to the use of iodinated contrast medium. For these reasons, phlebography has been widely replaced by duplex ultrasound (Baxter et al, 1990; Rose et al, 1990; Douglas and Sumner, 1996). The main limitations of this method are high dependency on the investigator s experience and inadequate scans of the calf, for example as a result of swelling or obesity. Numerous publications in recent years have demonstrated that the determination of plasma D-dimer, a degradation product of cross-linked fibrin, has a high sensitivity and negative Correspondence: Robert A. Bucek, MD, Clinic for Internal Medicine II, Department of Angiology, General Hospital, Währinger Gürtel 18 20, A-1090 Wien, Austria. robert.bucek@akhwien.ac.at predictive value, and is, therefore, especially useful in the exclusion of DVT (Heijboer et al, 1992; Bounameaux et al, 1994; Aschwanden et al, 1999). The main problem of D-dimer is the high number of false-positive results, especially in patients with inflammatory diseases and cancer (Bucek et al, 2002). Hence, C-reactive protein (CRP), an acute-phase reactant that is a marker for underlying systemic inflammation, is often additionally evaluated to help the physician in the differential diagnosis and to explain a positive D-dimer result in the absence of DVT. On the other hand, animal studies and in vitro experiments have suggested that DVT and the acute-phase response are interrelated (Wakefield et al, 1993, 1997, 2000; Downing et al, 1998; Londy et al, 1999; van Aken et al, 2000, 2002). Compared with this extensive number of experimental studies, little is known about the relevance of the determination of acute-phase reaction parameters in patients with DVT. Two small studies have proposed that the measurement of CRP alone, which is a cheap, quick and readily available test, can also be used to reliably exclude DVT, but three further trials failed to confirm these promising results (Thomas et al, 1989; Syrjala et al, 1990; Jossang and Runde, 1992, Wong et al, 1996; Maskell and Butland 2001). All of these studies suffered from important limitations, including size, no evaluation of possible influencing factors on CRP levels and most importantly no comparison with D-dimer. Ó 2002 Blackwell Publishing Ltd 385

2 386 R. A. Bucek et al To help to resolve the uncertainty regarding the exact role of CRP in the diagnosis of DVT, we performed this prospective study. The main objectives were the evaluation of sensitivity and specificity of CRP, and the possible advantages of the additional assessment of CRP with D-dimer, especially concerning a potential improvement of the low specificity of D-dimer. MATERIALS AND METHODS Consecutive patients with suspected DVT were referred to the outpatient department of Angiology of Vienna General Hospital and were included in this prospective study. All patients gave informed consent. A complete history was taken and after clinical investigation a venous blood sample was drawn from the cubital vein. A blank and a citrate tube were taken to determine plasma CRP and D-dimer levels. Exclusion criteria were clinically suspected superficial phlebitis, gravidity and antithrombotic or antiinflammatory treatment prior to admission. We used Tina-quant Ò CRP (Roche Diagnostics, Mannheim, Germany) for the evaluation of plasma CRP and STA-LIA Ò Test (Diagnostica Stago, Asnieres, France) for the evaluation of D-dimer. Tina-quant CRP is an immunoturbidimetric assay with a lower detection limit of 3Æ0 mg/l and a measurement range of 3Æ0 24 mg/l. STA- LIA D-dimer is also an immunoturbidimetric assay with a lower detection limit of 200 lg/l and a measurement range of lg/l. Assays were performed and interpreted by an independent operator without knowledge of the results of the clinical evaluation. CRP was classified as positive above a cut-off point of 10 mg/l and D-dimer above 500 lg/l, which are the standard thresholds in our local laboratory. These cut-off points were both established in 250 healthy individuals. The final diagnosis of DVT was confirmed by colour duplex sonography or ascending venography. Colour duplex ultrasound examinations of the symptomatic leg were performed by experienced technicians who were blinded to the clinical assessment and the laboratory results. Sonography was performed using a 128 XP scanner (Acuson, US) with a 5-MHz linear array probe. The pelvic, inguinal and femoral veins were investigated with the patients in a supine position, while the popliteal vein and the calf veins were scanned in prone position. DVT was diagnosed if at least one of the two following diagnostic criteria were present: visualization of an intraluminal thrombus in a deep vein and lack of, or incomplete, compressibility. The confidence for the exclusion of DVT of a single negative ultrasound examination of our local ultrasound laboratory was evaluated previously (Bucek et al, 2002). In this trial, none of 90 consecutive patients suffered from symptomatic DVT within a follow-up period of 3 months. Ascending venography was performed after exclusion of contraindications (renal insufficiency or contrast agent allergy) by an experienced investigator, who was blinded to the clinical assessment and the laboratory results. At least one of the following direct or indirect criteria had to be present in phlebography to diagnose DVT: a constant defect in filling seen on two views, an abrupt discontinuation of visible filling at a constant site in the vein and the absence of filling in the entire deep vein system, with or without venous flow through collateral veins. Statistical analysis. Statistical analysis was performed with the Statistical Package for the Social Sciences (SPSS) version (SPSS, Chicago, US). The significance of any difference in numeric values between patients (final diagnosis of DVT) and control subjects (final diagnosis of no DVT) was analysed using the Mann Whitney U-test or paired t-test and the significance of any difference in proportions was assessed by the chi-square method. Because CRP and D-dimer values are skewed, we used logarithmic transformation. For correlation analysis of D-dimer with CRP, we used Spearman s correlation coefficient (r). Multivariate analysis was performed using a linear regression model; results state the level of significance and in cases of significant results, the standardized correlation coefficient (B). For all tests, a P-value of < 0Æ05 was considered as significant. Sensitivity and specificity of CRP and D-dimer were compared using the receiver operating characteristic (ROC) curve procedure. A statistician at the local Institute for Medical Statistics was consulted to ensure proper usage of methods and tests. RESULTS In total, we included 233 patients in this prospective trial. Patients characteristics are listed in Table I. DVT was finally diagnosed in 73 patients (31Æ3%). The median plasma CRP level in patients [25 mg/l (interquartile range 56 mg/l)] was significantly increased compared with control subjects [5Æ0 mg/l (8Æ0 mg/l); P < 0Æ001]. Corresponding results (P < 0Æ001) for D-dimer were 3100 lg/l (8000 lg/l) and 500 lg/l (900 lg/l). CRP and D-dimer levels were significantly correlated (r ¼ 0Æ64, P ¼ 0Æ01). Multivariate regression analysis revealed that the presence of DVT (B ¼ 0Æ34; P <0Æ001), the presence of a known malignancy (B ¼ 0Æ26, P < 0Æ001) and presence of an inflammatory disease (B ¼ 0Æ15, P ¼ 0Æ009) were significantly correlated with higher CRP levels while the duration of symptoms (P ¼ 0Æ30), sex (P ¼ 0Æ32) and age (P ¼ 0Æ5) did not influence CRP significantly. A ROC curve, indicating the inferior sensitivity to specificity relationship of CRP compared with D-dimer, is shown in Fig 1. Sensitivity, specificity, the positive and the negative predictive values with 95% confidence intervals are shown in Table II. We performed a subgroup analysis to assess the value of the determination of CRP in suspected calf vein thrombosis, where inferior results of D-dimer have been reported. Sensitivity of CRP for suspected calf vein thrombosis was 63% (17/27) compared with 92Æ6% for D-dimer (25/27). Finally, we calculated a workflow diagram, indicating the distribution of the final diagnosis, based on the results of D-dimer and CRP (see Fig 2) to evaluate the role of the additional assessment of CRP with D-dimer.

3 Table I. Characteristics of enrolled patients and controls distributed to the final diagnosis. CRP in DVT 387 No DVT (160 patients) DVT (73 patients) P-value Sex Male 54 (33Æ7%) 38 (52Æ1%) 0Æ008 Female 106 (66Æ3%) 35 (47Æ9%) Suspected side Left 84 (52Æ5%) 39 (53Æ4%) 0Æ896 Right 76 (47Æ5%) 34 (46Æ6%) Former DVT 19 (11Æ9%) 10 (24Æ4%) 0Æ025 Age (years) 55 (18 88) 59 (21 90) 0Æ051 BMI 25Æ9 (13Æ1 43Æ8) 25Æ9 (13Æ8 36Æ1) 0Æ989 Median duration of symptoms (d) 7 (1 28) 5 (1 28) 0Æ033 Known malignancy 22 (13Æ8%) 26 (35Æ6%) < 0Æ001 Known inflammatory disease 17 (10Æ6%) 12 (16Æ4%) 0Æ212 Additional PE 5 (6Æ8%) Reference method Duplex sonography 122 (76Æ3%) 51 (69Æ9%) 0Æ301 Venography 38 (23Æ7%) 22 (30Æ1%) Proximal extension Iliac veins 18 (24Æ7%) Femoral veins 28 (38Æ4%) Popliteal vein 16 (21Æ9%) Calf veins 11 (15Æ0%) See also the evaluation of possible significant differences between the subgroups. Numbers in parenthesis show minimum and maximum values. DVT, deep vein thrombosis; PE, pulmonary embolism; BMI, body mass index. Fig 1. ROC curve of CRP and D-dimer for the diagnosis of DVT. Note the superior sensitivity to specificity relation of D-dimer compared with CRP for the whole curve. DISCUSSION This is the first trial evaluating the exact in vivo role of an acute-phase parameter, CRP, in the diagnosis of DVT and comparing the results with D-dimer. Until now, only five reports have dealt with the results of CRP in DVT, but conflicting results and conclusions have been reported. Thomas et al (1989) found a sensitivity of 100% and a specificity of 52% for 47 patients with suspected DVT; corresponding results were reported by Jossang et al (1992) in a series of 69 patients. Both concluded that CRP seemed to provide an inexpensive screening test for the exclusion of DVT. Inferior results were reported by Wong et al (1996) (sensitivity of 84% and a specificity of 62%), Syrjala et al (1990) (sensitivity of 72% in patients with iliofemoral thromboses and only 32% in patients with popliteal and calf vein thromboses) and Maskell and Butland (2001) (sensitivity of 60% and a specificity of 70%). In our study, sensitivity was 75% and specificity 69%. Differences in the reported results may partly relate to the differing study populations, especially taking in account that the plasma CRP levels are significantly influenced by the presence of inflammatory and malignant disease. These possible differences cannot be evaluated exactly because none of the above reports included any baseline characteristics or an assessment of possible influences on CRP levels. Multivariate analysis showed that CRP plasma levels were also significantly influenced by the presence of DVT, so this parameter cannot be used in the differential diagnosis of DVT and inflammatory diseases of other origin. In our study, D-dimer showed results comparable to those published in the literature. We found that CRP correlated significantly with D-dimer. However, compared with D-dimer, CRP provided a superior specificity (10%) and positive predictive value (5%). Therefore, we evaluated any possible improvement of the diagnostic

4 388 R. A. Bucek et al Table II. Sensitivity, specificity, positive predictive (PPV) and negative predictive value (NPV) with 95% confidence intervals (CI) of (CRP) and D-dimer. CRP (%) 95% CI D-dimer (%) 95% CI Sensitivity 75Æ3 (55/73) 63Æ9 84Æ7% 93Æ2 (68/73) 84Æ7 97Æ7% Specificity 68Æ8 (110/160) 61Æ0 75Æ8% 55 (88/160) 46Æ9 62Æ9% PPV 52Æ4 (55/105) 42Æ4 62Æ2% 46Æ6 (68/140) 40Æ4 57Æ2% NPV 85Æ9 (110/128) 78Æ7 91Æ4% 94Æ6 (88/93) 87Æ9 98Æ2% Fig 2. Workflow diagram based on the results of D-dimer and CRP assessment. Note that the additional evaluation of negative CRP in patients with negative D-dimer did not increase the negative predictive value. The additional evaluation of a negative CRP in patients with positive D-dimer reduces the risk of false-positive results by one half, but this approach bears the risk of DVT in 6%. performance of D-dimer by additionally assessing CRP. In patients with positive D-dimer and negative CRP, the rate of false-positive results was reduced by one half, but this approach carries the risk overlooking DVT in 6% of cases. In patients with positive D-dimer, additional positive CRP does not increase sensitivity. A negative D-dimer carries the risk of missed DVT in 2%, the additional evaluation of CRP does not increase the negative predictive value. To explain the significant difference of plasma CRP levels in patients compared with controls, several hypotheses have been proposed. One hypothesis, the possibility that increased levels of CRP prior to the event can induce further DVT, was based on the fact that CRP can induce monocytes to express tissue factor, a membrane glycoprotein important in initiating coagulation (Cermak et al, 1993). The in vivo role of this finding was evaluated by Ridker et al (1997). They found, in a subgroup analysis of the Physicians Health Study, including male physicians, that baseline concentrations of CRP were not significantly higher in patients who subsequently developed venous thrombosis (P ¼ 0Æ34). Syrjala et al (1990) tested whether DVT induced an inflammatory process. They demonstrated that white blood cell counts and axillary temperature were normal, and did not differ in patients and controls. Wong et al (1996) stated that the acute-phase response was only induced by the initial formation of thrombus and delayed presentation of DVT would allow time for CRP to return to normal levels. Thomas et al (1989) reported that CRP concentrations in plasma increased within 6 8 h after an inflammatory stimulus, peaking at around 48 h before falling to normal levels, with a half life of about 48 h. These data are challenged by our results. Although CRP levels were not significantly influenced by the duration of symptoms (P ¼ 0Æ3) in our study, they were significantly increased in patients compared with controls. On the other hand, several animal and in vitro studies have shown that DVT and inflammation are interrelated (Wakefield et al, 1993, 1997, 2000; Downing et al, 1996; Londy et al, 1999; van Aken et al, 2000, 2002), and several pathophysiological explanations have been reported, so these results and their clinical relevance have to be confirmed by further research. In conclusion, CRP was significantly increased in patients suffering from DVT. As multivariate analysis has shown, this fact is not only due to inflammatory or malignant diseases, but also due to the presence of DVT, so this parameter cannot be used in the differential diagnosis for inflammatory diseases of other origins. The assessment of CRP as the sole parameter in patients with suspected DVT provided inferior results compared with D-dimer. The additional evaluation of negative CRP in patients with positive D-dimer reduced the rate of false-positive results by one half, but overlooked a DVT in 6%. Based on these facts, we do not recommend assessing CRP in patients with suspected DVT.

5 REFERENCES van Aken, B.E., den Heijer, M., Bos, G.M., van Deventer, S.J. & Reitsma, P.H. (2000) Recurrent venous thrombosis and markers of inflammation. Thrombosis and Haemostasis, 83, van Aken, B.E., Reitsma, P.H. & Rosendaal, F.R. (2002) Interleukin 8 and venous thrombosis: evidence for a role of inflammation in thrombosis. British Journal of Haematology, 116, Aschwanden, M., Laboratories, K.H., Jeannerer, C., Gehrig, A. & Jaeger, K.A. (1999) The value of rapid D-dimer testing combined with structured clinical evaluation for the diagnosis of deep vein thrombosis. Journal of Vascular Surgery, 30, Baxter, G.M., McKechnie, S. & Duffy, P. (1990) Colour duplex ultrasound in deep venous thrombosis: a comparison with venography. Clinical Radiology, 42, Bounameaux, H., de Moerloose, P., Perrier, A. & Reber, G. (1994) Plasma measurement of D-dimer as diagnostic aid in suspected venous thromboembolism: an overview. Thrombosis and Haemostasis, 71, 1 6. Bucek, R.A., Koca, N., Reiter, M., Haumer, M., Zontsich, T. & Minar, E. (2002) Algorithms for the diagnosis of deep vein thrombosis in patients with low clinical pre-test probability. Thrombosis Research, 105, 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, Douglas, M.G. & Sumner, D.S. (1996) Duplex scanning for deep vein thrombosis: has it replaced both phlebography and noninvasive testing? Seminars in Vascular Surgery, 9, Downing, L.J., Strieter, R.M., Kadell, A.M., Wilke, Ca, Brown, S.L., Wrobleski, S.R., Burdick, M.D., Hulin, M.S., Fowlkes, J.B., Greenfield, L.J. & Wakefield, T.W. (1996) Neutrophils are the initial cell type identified in deep venous thrombosis induced vein wall inflammation. ASAIO Journal, 42, M677 M682. Downing, L.J., Strieter, R.M., Kadell, A.M., Wilke, C.A., Austin, J.C., Hare, B.D., Burdick, M.D., Greenfield, L.J. & Wakefield, T.W. (1998) IL-10 regulates thrombus-induced vein wall inflammation and thrombosis. Journal of Immunology, 151, Heijboer, J., Ginsberg, J.S., Büller, H.R., Lensing, A.W.A., Colly, L.P. & ten Cate, J.W. (1992) The use of the D-dimer test in combination with non-invasive testing versus serial non-invasive testing alone for the diagnosis of deep vein thrombosis. Thrombosis and Haemostasis, 67, Jossang, B. & Runde, I. (1992) Diagnostic value of C-reactive protein and D-dimer in deep venous thrombosis. Tidsskrift for Den Norske Laegeforening, 112, Kahn, S.R. (1998) The clinical diagnosis of deep venous thrombosis. Integrating incidence, risk factors, and symptoms and signs. Archives of Internal Medicine, 158, Londy, F.J., Kadell, A.M., Wrobleski, S.K., Prince, M.R., Strieter, R.M. & Wakefield, T.W. (1999) Detection of perivenous inflammation in a rat model of venous thrombosis using MRV. Journal of Investigative Surgery, 12, CRP in DVT 389 Maskell, N.A. & Butland, R.J.A. (2001) A normal serum CRP measurement does not exclude deep vein thrombosis. Thrombosis and Haemostasis, 86, Nordström, M., Lindblad, B., Bergqvist, D. & Kjellström, T. (1992) A prospective study of the incidence of deep-vein thrombosis within a defined urban population. Journal of Internal Medicine, 232, Ridker, P.M., Cushman, M., Stampfer, M.J., Tracy, R.P. & Hennekens, C.H. (1997) Inflammation, aspirin, and the risk of cardiovascular disease in apparently healthy men. The New England Journal of Medicine, 336, Rose, S.C., Zwiebel, W.J., Nelson, B.D., Priest, D.L., Knighton, R.A., Brown, J.W., Lawrence, P.F., Stults, B.M., Reading, J.C. & Miller, F.J. (1990) Symptomatic lower extremity deep venous thrombosis: accuracy, limitations, and role of color duplex flow imaging in diagnosis. Radiology, 175, Syrjala, H., Haukipuro, K. & Kiviniemi, H. (1990) Acute phase response and deep lower limb venous thrombosis. Journal of Clinical Pathology, 43, Thomas, E.A., Cobby, M.J., Rhys Davies, 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, Wakefield, T.W., Greenfield, L.J., Rolfe, M.W., DeLucia, III, A., Strieter, R.M., Abrams, G.D., Kunkel, S.L., Esmon, C.T., Wrobleski, S.K., Kadell, A.M., Burdick, M.D. & Taylor, F.B. (1993) Inflammation and procoagulant mediator interactions in an experimental baboon model of venous thrombosis. Thrombosis and Haemostasis, 69, Wakefield, T.W., Strieter, R.M., Prince, M.R., Downing, L.J. & Greenfield, L.J. (1997) Pathogenesis of venous thrombosis: a new insight. Cardiovascular Surgery, 5, Wakefield, T.W., Strieter, R.M., Schaub, R., Myers, D.D., Prince, M.R., Wrobleski, S.K., Londy, F.J., Kadell, M.A., Brown, S.L., Henke, P.K. & Greenfield, L.J. (2000) Venous thrombosis prophylaxis by inflammatory inhibition without anticoagulation therapy. Journal of Vascular Surgery, 31, Wells, P.S., Hirsh, J., Anderson, D.R., Lensing, A.W., Foster, G., Kearon, C., Weitz, J., D Ovidio, R., Cogo, A. & Prandoni, P. (1995) Accuracy of clinical assessment of deep-vein thrombosis. Lancet, 345, Wells, P.S., Anderson, D.R., Bormanis, J., Guy, F., Mitchell, M., Gray, L., Clement, C., Robinson, K.S. & Lewandowski, B. (1997) Value of assessment of pretest probability of deep-vein thrombosis in clinical management. Lancet, 350, Wille-Jorgensen, P., Borris, L., Jorgensen, L.N., Hauch, G., Lassen, M.R., Nehen, A.M., Kjaer, L. & Jensen, R. (1992) Phlebography as the gold standard in thromboprophylactic studies? A multi-center interobserver variation study. Acta Radiologica, 33, 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,