Selective D‐dimer testing for the diagnosis of acute deep vein thrombosis: a validation study

Abstract

The standard diagnostic algorithm for patients with clinically suspected deep vein thrombosis (DVT) starts with the Wells rule to determine pretest probability, followed by a quantitative D‐dimer test in the case of an unlikely clinical probability 1.Bates S.M. Jaeschke R. Stevens S.M. Goodacre S. Wells P.S. Stevenson M.D. Kearon C. Schunemann H.J. Crowther M. Pauker S.G. Makdissi R. Guyatt G.H. Diagnosis of DVT: antithrombotic therapy and prevention of thrombosis, 9th ed: American college of chest physicians evidence‐based clinical practice guidelines.Chest. 2012; 141: e351S-418SAbstract Full Text Full Text PDF PubMed Scopus (519) Google Scholar, 2.Wells P.S. Anderson D.R. Rodger M. Forgie M. Kearon C. Dreyer J. Kovacs G. Mitchell M. Lewandowski B. Kovacs M.J. Evaluation of D‐dimer in the diagnosis of suspected deep vein thrombosis.N Engl J Med. 2003; 349: 1227-35Crossref PubMed Scopus (1138) Google Scholar. In patients with a ‘DVT unlikely’ score (0–1 point) and a D‐dimer level < 0.5 μg/mL, DVT can be safely ruled out 1.Bates S.M. Jaeschke R. Stevens S.M. Goodacre S. Wells P.S. Stevenson M.D. Kearon C. Schunemann H.J. Crowther M. Pauker S.G. Makdissi R. Guyatt G.H. Diagnosis of DVT: antithrombotic therapy and prevention of thrombosis, 9th ed: American college of chest physicians evidence‐based clinical practice guidelines.Chest. 2012; 141: e351S-418SAbstract Full Text Full Text PDF PubMed Scopus (519) Google Scholar, 3.Linkins L.A. Bates S.M. Lang E. Kahn S.R. Douketis J.D. Julian J. Parpia S. Gross P. Weitz J.I. Spencer F.A. Lee A.Y. O'Donnell M.J. Crowther M.A. Chan H.H. Lim W. Schulman S. Ginsberg J.S. Kearon C. Selective d‐dimer testing for diagnosis of a first suspected episode of deep venous thrombosis: a randomized trial.Ann Intern Med. 2013; 158: 93-100Crossref PubMed Scopus (60) Google Scholar, 4.Schutgens R.E. Ackermark P. Haas F.J. Nieuwenhuis H.K. Peltenburg H.G. Pijlman A.H. Pruijm M. Oltmans R. Kelder J.C. Biesma D.H. Combination of a normal D‐dimer concentration and a non‐high pretest clinical probability score is a safe strategy to exclude deep venous thrombosis.Circulation. 2003; 107: 593-7Crossref PubMed Scopus (162) Google Scholar. In the remaining patients a proximal lower limb compression ultrasonography (CUS) is indicated. For patients with a ‘DVT likely’ score (≥2 points) and a normal CUS, a subsequently assessed D‐dimer level ≥0.5 μg/mL is an indication for a repeated CUS after 1 week 2.Wells P.S. Anderson D.R. Rodger M. Forgie M. Kearon C. Dreyer J. Kovacs G. Mitchell M. Lewandowski B. Kovacs M.J. Evaluation of D‐dimer in the diagnosis of suspected deep vein thrombosis.N Engl J Med. 2003; 349: 1227-35Crossref PubMed Scopus (1138) Google Scholar. This strategy has high sensitivity but low specificity, resulting in a high number of CUS that are negative for DVT 2.Wells P.S. Anderson D.R. Rodger M. Forgie M. Kearon C. Dreyer J. Kovacs G. Mitchell M. Lewandowski B. Kovacs M.J. Evaluation of D‐dimer in the diagnosis of suspected deep vein thrombosis.N Engl J Med. 2003; 349: 1227-35Crossref PubMed Scopus (1138) Google Scholar, 4.Schutgens R.E. Ackermark P. Haas F.J. Nieuwenhuis H.K. Peltenburg H.G. Pijlman A.H. Pruijm M. Oltmans R. Kelder J.C. Biesma D.H. Combination of a normal D‐dimer concentration and a non‐high pretest clinical probability score is a safe strategy to exclude deep venous thrombosis.Circulation. 2003; 107: 593-7Crossref PubMed Scopus (162) Google Scholar. One of the main challenges in improving this algorithm is to safely diminish the number of required CUS. Recently, a diagnostic algorithm applying selective D‐dimer thresholds depending on clinical probability was investigated in a randomized clinical trial, in which the original three risk categories by Wells were reintroduced 3.Linkins L.A. Bates S.M. Lang E. Kahn S.R. Douketis J.D. Julian J. Parpia S. Gross P. Weitz J.I. Spencer F.A. Lee A.Y. O'Donnell M.J. Crowther M.A. Chan H.H. Lim W. Schulman S. Ginsberg J.S. Kearon C. Selective d‐dimer testing for diagnosis of a first suspected episode of deep venous thrombosis: a randomized trial.Ann Intern Med. 2013; 158: 93-100Crossref PubMed Scopus (60) Google Scholar. In patients with a low risk score (0 points) the D‐dimer threshold was increased to 1.0 μg/mL, while in patients with a moderate risk score (1–2 points) the threshold remained 0.5 μg/mL to rule out DVT without further imaging. In all other patients CUS was performed. Repeated CUS was indicated in patients with moderate or high risk scores and an initial negative CUS in combination with a D‐dimer ≥0.5 μg/mL 1.Bates S.M. Jaeschke R. Stevens S.M. Goodacre S. Wells P.S. Stevenson M.D. Kearon C. Schunemann H.J. Crowther M. Pauker S.G. Makdissi R. Guyatt G.H. Diagnosis of DVT: antithrombotic therapy and prevention of thrombosis, 9th ed: American college of chest physicians evidence‐based clinical practice guidelines.Chest. 2012; 141: e351S-418SAbstract Full Text Full Text PDF PubMed Scopus (519) Google Scholar. This strategy resulted in a relative reduction of 13.3% (95% CI 10.4–16.5%; absolute reduction 7.6%) of CUS with an absolute reduction of 22% in the number of D‐dimer tests without compromising the safety of the algorithm. Nonetheless, due to two characteristics of this study, a validation study would be of interest. First, the overall DVT prevalence was 7%, which is lower than the 16–39% reported in European studies 2.Wells P.S. Anderson D.R. Rodger M. Forgie M. Kearon C. Dreyer J. Kovacs G. Mitchell M. Lewandowski B. Kovacs M.J. Evaluation of D‐dimer in the diagnosis of suspected deep vein thrombosis.N Engl J Med. 2003; 349: 1227-35Crossref PubMed Scopus (1138) Google Scholar, 4.Schutgens R.E. Ackermark P. Haas F.J. Nieuwenhuis H.K. Peltenburg H.G. Pijlman A.H. Pruijm M. Oltmans R. Kelder J.C. Biesma D.H. Combination of a normal D‐dimer concentration and a non‐high pretest clinical probability score is a safe strategy to exclude deep venous thrombosis.Circulation. 2003; 107: 593-7Crossref PubMed Scopus (162) Google Scholar. Second, management in the reference group differed from the standard algorithm 1.Bates S.M. Jaeschke R. Stevens S.M. Goodacre S. Wells P.S. Stevenson M.D. Kearon C. Schunemann H.J. Crowther M. Pauker S.G. Makdissi R. Guyatt G.H. Diagnosis of DVT: antithrombotic therapy and prevention of thrombosis, 9th ed: American college of chest physicians evidence‐based clinical practice guidelines.Chest. 2012; 141: e351S-418SAbstract Full Text Full Text PDF PubMed Scopus (519) Google Scholar, 2.Wells P.S. Anderson D.R. Rodger M. Forgie M. Kearon C. Dreyer J. Kovacs G. Mitchell M. Lewandowski B. Kovacs M.J. Evaluation of D‐dimer in the diagnosis of suspected deep vein thrombosis.N Engl J Med. 2003; 349: 1227-35Crossref PubMed Scopus (1138) Google Scholar: all patients underwent D‐dimer testing at presentation regardless of the Wells score and in the case of a normal D‐dimer test DVT was excluded. Therefore, we assessed the safety and performance of the selective D‐dimer algorithm in a post‐hoc analysis of a prospective DVT management study performed in the Netherlands from 2009 until 2010 with a higher DVT prevalence and compared the failure rate and number of required CUS with the standard algorithm 5.Tan M. Abstracts of the XXIII congress of the international society on thrombosis and haemostasis with the 57th annual SSC (Scientific and Standardization Committee) meeting.J Thromb Haemost. 2011; 9: 499-711Google Scholar. Exclusion criteria were pregnancy, anticoagulant therapy and inability for follow‐up. A diagnosis of DVT was established by CUS. The primary endpoint was the objective diagnosis of symptomatic venous thromboembolism. A rejected DVT diagnosis at baseline followed by an uneventful 3‐month follow‐up was the reference standard for the true absence of DVT. We calculated failure rates, the number of required CUS and D‐dimer tests, test characteristics and predictive values. Of 606 eligible patients, 128 were excluded because of unavailable D‐dimer test results due to diagnosis of DVT at the initially performed CUS and 89 patients were excluded because a D‐dimer test result was not available due to protocol violation. The baseline characteristics of the remaining 389 patients and the 217 excluded patients are shown in Table 1. Of the included patients, overall DVT prevalence was 22% (79 diagnosed at baseline, eight after repeated CUS, none during follow‐up). One patient died during follow‐up in whom serial CUS was negative for DVT, with an unknown cause of death. Five patients were lost to follow‐up, of which three were managed without CUS in both strategies. The excluded patients had a considerably higher DVT prevalence and higher Wells scores (Table 1).Table 1Baseline characteristicsCharacteristicsIncluded patients (n = 389)Excluded patients (n = 217)TotalStandard algorithmSelective D‐dimer algorithmTotalStandard algorithmSelective D‐dimer algorithmUnlikely (n = 175) 45%Likely (n = 214) 55%Low (n = 76) 20%Moderate (n = 175) 45%High (n = 138) 35%Unlikely (n = 37) 17%Likely (n = 180) 83%Low (n = 13) 6%Moderate (n = 86) 40%High (n = 118) 54%Mean age in years (SD)58 (18)54 (18)62 (17)51 (17)57 (18)64 (16)60 (18)58 (18)60 (18)57 (20)58 (18)61 (18)Female sex (n, %)237 (61)107 (61)130 (61)48 (63)105 (60)84 (61)114 (53)24 (65)90 (50)9 (69)42 (49)63 (53)Estrogen use (n, % of females)35 (15)22 (21)13 (10)12 (25)16 (15)7 (8)20 (18)5 (21)15 (17)2 (22)5 (12)13 (21)Immobilization > 3 days or surgery (n, %)78 (20)14 (8)64 (30)2 (3)28 (16)48 (35)63 (29)5 (14)58 (32)1 (8)14 (16)48 (41)History of VTE (n, %)87 (22)46 (26)41 (19)18 (24)42 (24)27 (20)38 (18)6 (16)32 (18)3 (23)15 (17)20 (17)Malignancy (n, %)23 (6)1 (0.6)22 (10)0 (0)5 (3)18 (13)26 (12)1 (3)25 (14)0 (0)9 (11)17 (14)DVT at baseline (n, %)79 (20)20 (11)59 (28)6 (8)30 (17)43 (31)139 (64)11 (30)128 (71)2 (8)47 (55)90 (76)SD, standard deviation; VTE, venous thromboembolism; DVT, deep venous thrombosis. Open table in a new tab SD, standard deviation; VTE, venous thromboembolism; DVT, deep venous thrombosis. Using the standard algorithm, 45% of the patients had a ‘DVT unlikely’ and 55% a ‘DVT likely’ score, with a DVT prevalence of 11% and 28%. A D‐dimer test was indicated in 84% (325/389), CUS was required in 79% (306/389), and in 125 patients repeated CUS was performed. Fourteen repeated CUS were performed without indication by the algorithm, of which three were positive for DVT. In total, 445 CUS were performed. The failure rate in patients managed without CUS was 0.0% (0/83; 95% CI 0.0–4.4%) and 0.0% (0/302; 95% CI 0.0–1.2%) in the total cohort. The sensitivity of the algorithm was 100% (95% CI 0.96–100%), the specificity 27% (95% CI 23–33%), the negative predictive value (NPV) 100% (95% CI 96–100%) and the positive predictive value (PPV) 28% (95% CI 23–34%). Applying the selective D‐dimer algorithm, 20% of the patients had a low, 45% a moderate and 35% a high risk score with a DVT prevalence of 8%, 17% and 31%, respectively. CUS was needed in 72% (281/389) at baseline and 165 repeated CUS were indicated, resulting in a total of 446 CUS. This is a relative reduction of 8.2% (95% CI 5.4–11.8%) for CUS needed at baseline, but an increase of 0.2% (95% CI 0.0–1.2%) of the total number of CUS. D‐dimer testing was required in 88% (343/389), an absolute increase of 4.6% (95% CI −0.5 to 9.5%). No additional patients with DVT would have been missed by this algorithm, resulting in failure rates of 0.0% (0/108; 95% CI 0.0–3.4%) in patients managed without CUS, and 0.0% (0/302; 95% CI 0.0–1.2%) in the total cohort. The sensitivity was 100% (95% CI 96–100%), the specificity 36% (95% CI 30–41%), the NPV 100% (95% CI 97–100%) and the PPV 31% (95% CI 26–37%). These results suggest that the selective D‐dimer strategy is indeed safe and reduces the number of initial CUS by 8.2% in a population with a high DVT prevalence of 22%. However, considering repeated CUS, the total number of CUS did not decrease. Of note, in the study by Linkins, the total number of CUS performed in the standard strategy was 839 and in the selective testing strategy 821, a relative reduction of 2.1% (95% CI 1.3–3.4%). which is considerably lower than the 13.3% of CUS at baseline (difference 11.1%; 95% CI 8.4–13.2) 3.Linkins L.A. Bates S.M. Lang E. Kahn S.R. Douketis J.D. Julian J. Parpia S. Gross P. Weitz J.I. Spencer F.A. Lee A.Y. O'Donnell M.J. Crowther M.A. Chan H.H. Lim W. Schulman S. Ginsberg J.S. Kearon C. Selective d‐dimer testing for diagnosis of a first suspected episode of deep venous thrombosis: a randomized trial.Ann Intern Med. 2013; 158: 93-100Crossref PubMed Scopus (60) Google Scholar. The increase of the total number of CUS in our study is caused by the recommendation to perform serial CUS in a larger proportion of patients in the three‐level algorithm compared with the dichotomized algorithm. Also in contrast to the trial by Linkins, the selective D‐dimer strategy would not result in a reduction in the number of D‐dimer tests. This is a result of the smaller proportion of patients in whom a CUS is the first and only test performed after clinical probability assessment, because D‐dimer measurement is necessary in all patients except those with a high probability and with DVT on initial CUS. Other important considerations are the modification and the higher complexity of the reasonably well‐established dichotomized algorithm, which may limit the applicability of this new algorithm in busy emergency wards. An important strength of our analysis is the DVT prevalence of 22%, which is representative for European clinical practice. Also, the comparison with the standard algorithm provides a valuable estimate of the potential benefits for physicians applying the dichotomized algorithm. Limitations are the post‐hoc design and the exclusion of patients in whom no D‐dimer test was performed. Exclusion of these patients may have resulted in an overestimation of the reduction in CUS and may also have influenced the failure rate. Additional limitations are the absence of recurrences in our study and the inclusion of outpatients only. In conclusion, the selective D‐dimer strategy appears to safely exclude DVT, with a reduction of CUS at baseline in a population with a higher DVT prevalence. However, the total number of CUS was not decreased. Therefore, we cannot recommend replacing the standard strategy by the selective D‐dimer thresholds strategy. Nevertheless, due to the limitations of this study a prospective management trial comparing the selective D‐dimer threshold strategy with the standard strategy in a population with a higher DVT prevalence is still needed. T. van der Hulle, M.V. Huisman and F.A. Klok designed the study, performed the analyses and drafted the manuscript. M. Tan and G.C. Mol gathered the data and designed the study. All authors reviewed the manuscript and approved the final version of the manuscript. The study was supported by unrestricted grants from the participating hospitals. The authors state that they have no conflict of interests.