Journal of Thrombosis and Haemostasis, 8: 10041011

DOI: 10.1111/j.1538-7836.2010.03783.x

ORIGINAL ARTICLE

D-dimer testing in pregnant patients: towards determining the next level in the diagnosis of deep vein thrombosis
W.-S.CHAN,* A. LEE, F. A. SPENCER, S. CHUNILAL, M. CROWTHER, W. WU, M . J O H N S T O N , * * M . R O D G E R and J . S . G I N S B E R G
*Department of Medicine, Womens College Hospital, Toronto, ON; Department of Medicine, University of British Columbia, Vancouver, BC; Department of Medicine, McMaster University, Hamilton, ON, Canada; Department of Haematology, North Shore Hospital, Takapuna, New Zealand; Womens College Research Institute and Womens College Hospital; **Hemostasis Reference Laboratory, Henderson Hospital, Hamilton, ON; and Department of Medicine, University of Ottawa, Ottawa, ON, Canada

To cite this article: Chan W-S, Lee A, Spencer FA, Chunilal S, Crowther M, Wu W, Johnston M, Rodger M, Ginsberg JS. D-dimer testing in pregnant patients: towards determining the next level in the diagnosis of deep vein thrombosis. J Thromb Haemost 2010; 8: 100411.

Summary. Background: The role of D-dimer in excluding deep vein thrombosis (DVT) in pregnancy is currently uncertain. We hypothesized that the specicity of sensitive D-dimer assays could be improved without compromising sensitivity by using higher D-dimer cut-o values. Objective: To determine the test characteristics of two rapid enzyme-linked immunosorbent assays and three latex agglutination assays in pregnancy. Method: We recruited consecutive pregnant women who presented to participating centers with suspected DVT for the study. Symptomatic women were investigated with compression ultrasonography, and received 3 months of clinical followup to assess for the presence of venous thrombosis. Plasma samples for D-dimer were collected and frozen at the time of presentation. The median and mean D-dimer values for respective trimesters of pregnancy in patients with and without DVT were calculated. Receiver operating curves (ROCs) were plotted for respective assays to establish the best cut-points. The test characteristics corresponding to standard cut-points and these pregnancy cut-points are presented. Results: The prevalence of DVT in our cohort was 6.6% (95% condence interval 4.010.6%). The mean and median D-dimer values were signicantly increased throughout pregnancy. Overall, women with conrmed DVT had higher D-dimer levels than women without DVT (P < 0.0001). Improved specicities (62 79%) were observed with the use of higher cut-points obtained from ROCs for all ve assays, and high sensitivities were manintained (80100%) for DVT diagnosis. Conclusion: Using higher cut-points than those used in non-pregnant patients, the

specicity of D-dimer assays for the diagnosis of DVT in pregnancy can be improved without compromising sensitivity. Validation in prospective management studies is needed. Keywords: D-dimer, deep vein thrombosis, pregnancy. Background Venous thromboembolism (VTE) is a major cause of maternal morbidity and mortality [14] in developed countries. As physiologic changes occurring during pregnancy can mimic symptoms of VTE, objective testing to determine the presence or absence of deep vein thrombosis (DVT) is of particular importance. In addition to compression ultrasonography, the primary imaging modality for diagnosing DVT, diagnostic algorithms involving D-dimer testing, have been developed to assist clinicians in the management of non-pregnant patients with suspected DVT [57]. In combination with clinicians assessments of pretest probability, D-dimer results have been shown to be useful in stratifying patients into those who do and do not require further diagnostic testing, thereby decreasing the number of compression ultrasound scans that must be performed and minimizing unnecessary treatment when ultrasonography is not immediately available. Unfortunately, the utility of D-dimer testing in pregnancy remains unclear, because most diagnostic studies have excluded pregnant women [57] and because D-dimer levels are often elevated in pregnancy in the absence of DVT [810]. Given the relative frequency with which symptoms suggestive of DVT occur in pregnant women, a better understanding of the utility of Ddimer testing in pregnant women will improve our management in this cohort of patients. Currently available D-dimer assays vary in their accuracy and utility in diagnosing DVT [11]. Quantitative enzyme-linked immunosorbent assay (ELISA)-based assays are highly sensitive and can reliably exclude DVT in patients with suspected DVT. However, the false-positive rates of these
2010 International Society on Thrombosis and Haemostasis

Correspondence: Wee-Shian Chan, Womens College Hospital, Department of Medicine, 76 Grenville Street, Toronto, ON M5G 1B2, Canada. Tel.: +1 416 323 6272; fax: +1 416 323 7739. E-mail: wee-shian.chan@wchospital.ca Received 18 December 2009, accepted 22 January 2010

D-dimer testing in pregnant patients 1005

assays are also high, particularly in patients with alternative reasons for an elevated D-dimer level, such as pregnancy. In one study, most asymptomatic pregnant women were found to have elevated D-dimer levels by a rapid ELISA after 16 weeks of gestation [8]. In another study, only 22% of women in the second trimester and none in the third trimester had normal Ddimer values as measured by a latex agglutination assay using standard non-pregnant cut-points [9]. In contrast, the use of assays with lower sensitivity but higher specicity may reduce false-positive rates, but may not reliably exclude DVT in symptomatic patients. However, a recent study [12] evaluating the performance of a qualitative whole blood agglutination (lower sensitivity and higher specicity) D-dimer assay in pregnant women with suspected DVT showed that the sensitivity was 100% [95% condence interval (CI) 77100%), the specicity was 60% (95% CI 5268%), and the negative predictive value (NPV) was 100% (95% CI 95100%). Although this assay appeared to be promising for use in pregnant patients, it is being rapidly replaced in clinical practice by more sensitive D-dimer assays. The primary objective of this study was to determine the test characteristics of ve D-dimer assays for DVT diagnosis in symptomatic pregnant patients. We hypothesized that the use of higher cut-points to discriminate normal from abnormal results could compensate for the higher baseline D-dimer values in pregnancy and improve the specicity of sensitive D-dimer assays, without lowering the sensitivity, in this population. To test this hypothesis, we conducted a retrospective study using stored plasma samples from a cohort of pregnant women who presented with suspected DVT, and who were all investigated with compression ultrasonography for DVT [12]. Materials and methods
Study population

and failure of patient or attending physician to provide consent.

DVT diagnosis

Details of the diagnostic algorithm used to conrm or refute a diagnosis of DVT are described elsewhere [12]. Briey, all patients underwent compression ultrasonography of the symptomatic leg(s) at presentation. Compression ultrasonography was performed with gentle compression of the common femoral, supercial femoral and popliteal veins, as well as the calf trifurcation. If isolated iliac vein thrombosis was suspected, visualization of the iliac vein by direct imaging and Doppler ow was also obtained. If the initial compression ultrasonography nding was negative, patients received repeated compression ultrasonography testing on days 3 and/or 7, based on the clinicians empirical assessment of pretest probability. DVT was diagnosed on the basis of a non-compressible venous segment and, for the iliac veins, by the absence of ow within the iliac vein and/or the presence of a visible thrombus according to B-mode imaging. All patients with DVT were treated with unfractionated or low molecular weight heparin. Patients whose compression ultrasonography nding was normal had anticoagulants withheld and underwent clinical follow-up for at least 3 months. These patients were either seen or contacted at the end of 3 months to ensure that no intervening event had occurred after initial presentation. This approach has been used successfully by several groups (including our own) [7,13,14] to categorize patients as DVTpositive or DVT-negative in the validation of diagnostic tests for DVT in non-pregnant patients.
Laboratory testing

Consecutive pregnant women presenting to one of ve Canadian centers between June 2000 and August 2008 with suspected DVT were potentially eligible for the study. The protocol was reviewed and approved by the Ethics Research Boards of all participating centers. Written informed consent was obtained from all study participants. The ve centers included: Womens College Hospital, Toronto; Hamilton Health Sciences, McMaster University Medical Centre Site and Henderson Hospital Site, Hamilton; St Josephs Healthcare Centre, Hamilton; and Ottawa Hospital Civic and General Campuses, Ottawa. The Hamilton and Ottawa centers are tertiary referral sites for thrombosis; the Toronto, McMaster, St Josephs and Ottawa sites are tertiary referral centers for obstetric management. Patients with one or more of the following criteria were excluded from the study: a previous history of VTE; treatment with full-dose anticoagulation for 24 h or longer; concomitant symptoms consistent with pulmonary embolism (PE); unable or unwilling to return for follow-up; geographic inaccessibility;
2010 International Society on Thrombosis and Haemostasis

Blood was collected into 5-mL Vacutainer tubes containing 0.105 M (3.2%) buffered sodium citrate at the time of enrollment, and processed by laboratory technicians blinded to the clinical status and the results of other diagnostic testing for DVT. Samples were gently mixed, and then centrifuged for 15 min at 1700 g within 1 h of collection. Plasma was removed and recentrifuged for 5 min to obtain platelet-free plasma, which was then stored at ) 70 C in 500-lL aliquots at each participating center. At the end of the study, all aliquots were then shipped on dry ice to the central laboratory (Hemostasis Reference Laboratory at Hamilton Health Sciences Henderson Hospital) for batch assays. Five commercially available D-dimer assays (two ELISAs and three quantitative immunoturbidometric microparticle assays) were tested. The ELISAs were Vidas D-dimer (BioMerieux, Durham, NC, USA) and Asserachrome D-dimer (Stago, Asnieres, France). The three automated quantitative immunoturbidometric microparticle latex assays used were the IL Test (Instrumentation Laboratories, Lexington, MA, USA), the Sta-Lia Test (Stago), and Innovance D-Dimer lg mL)1 dimer units (Siemens, Marburg, Germany). All ndings are reported in lg mL)1 brinogen equivalent units (FEU) except

1006 W.-S. Chan et al

for the IL Test, the ndings of which are reported in lg mL)1 dimer units. Assays were performed according to the manufacturers specications. Technologists performing assays were masked to patient characteristics and DVT status.
Analysis

For the analysis, patients were categorized as DVT-positive if they had a diagnostic compression ultrasonography result at presentation or an objectively conrmed, symptomatic DVT or PE during follow-up. Patients with a negative compression ultrasonography result at presentation and no VTE during 3 months of follow-up were categorized as DVT-negative. The median and mean of D-dimer values, with the corresponding interquartile range and standard deviation, were calculated for each trimester of pregnancy in women with and without DVT. The results for women with and without DVT were compared using the MannWhitney U-test. A receiver operating characteristic (ROC) curve was generated for each of the four assays, by plotting the sensitivity against 1 specicity using a Wilcoxon estimate. For each of the optimal cut-points generated by the ROC curves, the test characteristics (sensitivity, specicity, NPV, and negative likelihood ratios) and their corresponding 95% CIs were calculated. Test characteristics were also calculated for all ve assays using standard cut-points established for non-pregnant patients. All analyses were performed using SAS statistical software version 9.1 (SAS Institute Inc., Cary, NC, USA) and STATSDIRECT statistical software version 2.7.3 (StatsDirect Ltd, Altrincham, UK).
Choice of optimal cut-point for each assay

On the basis of a desired NPV of 98% (consistent with that reported in non-pregnant subjects after normal venography, the reference standard diagnostic test) and an estimated incidence of DVT of  10%, we estimated that a sensitivity of 80% would sufce, provided that the specicity was at least 60%. Therefore, the optimal cut-points were established on the basis of two criteria: an observed NPV of 98%, and a specicity of at least 60%. This would reduce the number of false-positives and improve the utility of the test without reducing safety. Results A total of 249 pregnant women who presented with suspected DVT were enrolled in the study over an 8-year period. Plasma was collected and available for testing in 228 (91.2%) of these patients. Among these patients, 15 were diagnosed with DVT, resulting in an incidence of 6.6% (95% CI 4.010.6). Of the 21 patients in whom D-dimer levels were not determined at presentation, two had DVT. The majority of these patients presented with suspected DVT in the second (36.4%) or third (59.6%) trimesters of pregnancy. The distribution of DVT was 26.6%, 26.7% and 46.7%,

respectively, in the rst, second and third trimesters of pregnancy. DVT was diagnosed on initial presentation in 80% (12/15) of patients; in the remaining three patients, DVT was diagnosed on serial testing over 7 days. The rst patient initially had a great saphenous vein thrombosis, which progressed into the femoral vein over 6 days, the second patient had two negative compression ultrasonography results, followed by a positive third compression ultrasonography result on day 7 that demonstrated extensive DVT involving the ileofemoropopliteal veins, and the third patient had a popliteal vein DVT diagnosed 4 days after the initial negative ultrasound result. No patients developed PE during follow-up. The mean and median D-dimer values for each assay are presented by trimester and the presence or absence of DVT in Table 1. As expected, both mean and median D-dimer levels increased with progressive trimesters of pregnancy among women who did not have acute thrombosis. D-dimer values were also signicantly higher in those women who were diagnosed with DVT than in those who were not (Mann Whitney U-test, P < 0.0001), for each of the ve assays. Correlations for all ve assays were good (data available on request). The ROC curves associated with respective D-dimer assay were plotted and are shown in Figs 15. The optimal cut-points were 1.89 lg mL)1 FEU for the Vidas assay, 1.51 lg mL)1 FEU for the Asserachrome assay, 0.57 lg mL)1 dimer units for the IL assay, 1.38 lg mL)1 FEU for the Sta-Lia assay, and 1.50 lg mL)1 FEU for the Innovance assay. The area under the curves ranged between 0.82 and 0.87 for all ve assays, suggesting that the ability of these assays to discriminate between pregnant women with and without DVT is good. The test characteristics corresponding to established and pregnancy cut-points for all ve assays are shown in Table 2. In all ve assays, the sensitivities are high with the use of standard cut-points (100%) as derived from published studies [1519]; however, the specicities in all cases are poor (range, 623%). Conversely, using the optimal cut-points for pregnant patients derived from the ROC curves, there is improved specicity (range, 6179%), with only a small reduction in sensitivity in four of the ve assays (range, 93100%) and a modest reduction in the fth assay (80%). With the low incidence of DVT in the cohort, the NPV is, as expected, high in all cases, at more than 98%. Discussion Our ndings suggest that the currently available sensitive D-dimer assays that are used for the exclusion of DVT in symptomatic non-pregnant patients have the potential to exclude DVT in symptomatic pregnant patients with the application of higher cut-points. We showed that even as D-dimer levels increase several-fold during pregnancy, we can still use higher D-dimer cut-points to exclude a DVT without worrying about missing one in pregnancy. The ndings from our study are important because we are unaware of similar studies investigating the potential use of
2010 International Society on Thrombosis and Haemostasis

D-dimer testing in pregnant patients 1007

available D-dimer assays for the diagnosis of DVT in pregnant patients. In non-pregnant patients with suspected DVT, use of a clinical prediction tool and the D-dimer assay is sufcient to safely exclude DVT in low-risk patients. Whether this will prove true in pregnant patients requires further careful and cautious study. At the very least, D-dimer testing may allow for minimization and/or elimination of the need for serial ultrasound testing in patients with an initial negative study. In addition, increased understanding of the utility of D-dimer assays for the exclusion of DVT in pregnant women will serve as a basis for subsequent studies evaluating its use in pregnant patients with suspected PE. This will be of great importance, as D-dimer assays in this setting may allow for omission or at least minimization of radiologic testing (e.g. computed tomography, ventilation/perfusion scan) in this high-risk population. With the use of current established cut-points, many pregnant women would test positive (with abnormally elevated D-dimer levels) in the absence of DVT. Raising the cut-points can improve the specicity of these tests, but a high degree of sensitivity must be preserved to avoid misdiagnosis. The criteria that we used to decide upon the optimal cut-points were those providing an NPV of at least 98% and a specicity of at least 60%.We recognize the limitations of such an approach, because some clinicians are adamant that missing any VTE during pregnancy is unacceptable and that the sensitivity and NPV of the assays should both be 100%. Such an approach is impractical for some assays, as the specicity falls too low for the assays to be useful. Furthermore, we believe that it is an irrational emotional response that is not consistent with day-to-day practice. Finally, we consider these cut-points to represent a good start for prospective evaluation, in which the D-dimer assays could be combined with clinical pretest probability or compression ultrasonography, and should not be used as stand-alone tests. Our study demonstrates that such a strategy is possible for at least four of the ve commercially available D-dimer assays evaluated. With use of the cut-points identied with ROC curves, the false-positive rate of D-dimer testing (proportion of pregnant women who test positive on the basis of elevated D-dimer levels who do not have DVT) would decrease from about 75% to about 2535%. Raising the cut-points would improve the utility of these available commercial assays. Whether setting different cutpoints for each trimester of pregnancy will ne-tune and improve accuracy is an unanswered question, but we were unable to address this, owing to the small number of patients and events in each trimester. Our study has several limitations. The sample size is small and the incidence of DVT among symptomatic patients is low, so the CIs associated with our point estimates of the accuracy indexes (particularly the sensitivities) are wide. However, the inclusion of unselected patients and the use of objective testing to conrm DVT diagnosis reduce the likelihood of sampling and diagnostic biases in our study. The incidence of DVT in our study is also consistent with other studies in pregnant women, suggesting that the NPVs are not likely to be inated

(1.965.19) (0.711.75) (1.874.89) (0.631.67) (0.571.55) (0.250.58) (1.523.92) (0.461.33) (1.805.70) (0.871.96) Vidas (lg mL FEU) Asserachrome (lg mL)1 FEU) IL Test (lg mL)1 dimer units) Sta-Lia (lg mL)1 FEU) Innovance (lg mL)1 FEU)
)1

Table 1 The mean and median D-dimer values for four D-dimer assays by trimester and deep vein thrombosis status

2010 International Society on Thrombosis and Haemostasis

D-dimer assay (units)

DVT, deep vein thrombosis; FEU, brinogen equivalent units; IQR, interquartile range; Negative, absence of deep vein thrombosis; Positive, presence of deep vein thrombosis; SD, standard deviation.

Median (IQR) Mean (SD) All n Median (IQR) Mean (SD) n Median (IQR) Mean (SD) n Median (IQR) Mean (SD) n DVT Trimester 3 (28 weeks or later) Trimester 2 (12 to less than 28 weeks) Trimester 1 (less than 12 weeks)

Positive Negative Positive Negative Positive Negative Positive Negative Positive Negative

4 5 4 5 4 5 4 5 4 5

8.53 0.90 4.98 0.90 2.12 0.78 10.27 0.76 8.88 1.03

(9.79) (0.91) (3.08) (0.89) (2.05) (0.36) (12.76) (0.69) (8.24) (0.96)

4.60 0.60 5.41 0.65 1.59 0.82 5.40 0.48 7.23 0.80

(2.5014.56) (0.151.35) (2.387.58) (0.191.17) (0.583.66) (0.471.04) (1.9918.55) (0.221.06) (2.3515.41) (0.211.51)

4 79 4 79 4 79 4 79 4 79

2.00 0.98 2.13 0.97 0.69 0.40 1.62 0.97 2.03 1.30

(0.37) (1.21) (0.35) (1.23) (0.19) (0.77) (0.73) (2.15) (0.58) (1.93)

1.98 0.71 2.26 0.66 0.72 0.27 1.60 0.52 2.02 0.93

(1.762.23) (0.550.95) (1.912.34) (0.510.89) (0.540.84 (0.210.34) (1.092.15) (0.330.82) (1.532.54) (0.741.17)

7 129 7 129 7 129 7 129 7 129

3.77 3.39 3.47 1.90 1.13 0.75 3.09 2.50 4.35 3.70

(1.52) (7.94) (2.09) (1.80) (0.68) (1.04) (1.39) (6.49) (2.11) (8.49)

3.74 1.48 2.75 1.25 1.04 0.42 2.76 0.99 4.67 1.56

(2.105.19) (0.972.05) (1.864.89) (0.842.19) (0.571.55) (0.290.70) (1.723.92) (0.671.77) (2.055.70) (1.102.67)

15 212 15 211 15 210 15 213 15 209

4.57 2.43 3.52 1.53 1.28 0.62 4.62 1.89 4.94 2.74

(5.32) (6.32) (2.26) (1.66) (1.19) (0.95) (6.98) (5.26) (4.85) (6.79)

3.03 1.04 2.52 0.96 0.79 0.35 2.61 0.82 3.04 1.26

1008 W.-S. Chan et al

Sensitivity 1.00

Optimum cut-off point selected = 1.89 g mL1 FEU

0.75

0.50

DVT 0.25 Dd Pos Neg

Pos 14 1

Neg 45 167

0.00 0.00

0.25

0.50

0.75

1.00 1-Specificity

Fig. 1. Receiver operating characteristic (ROC) curve for Vidas D-dimer assay. Dd, D-dimer; DVT, deep vein thrombosis; FEU, brinogen equivalent units; Pos, Positive; Neg, negative; Dd Pos, level at or above cut-point; Dd Neg, level below cut-point. Area under ROC curve by Wilcoxon estimate = 0.87 (95% condence interval 0.820.92).

Sensitivity 1.00 Optimum cut-off point selected = 1.51 g mL1 FEU 0.75

0.50

DVT 0.25 Dd Pos Neg

Pos 15 0

Neg 55 156

0.00 0.00

0.25

0.50

0.75

1.00 1-Specificity

Fig. 2. Receiver operating characteristic (ROC) curve for Asserachrome D-dimer assay. Dd, D-dimer; DVT, deep vein thrombosis; FEU, brinogen equivalent units; Pos, Positive; Neg, negative; Dd Pos, level at or above cut-point; Dd Neg, level below cut-point. Area under ROC curve by Wilcoxon estimate = 0.87 (95% condence interval 0.810.92).

by an unusually low incidence of DVT in our study sample. Also, we have minimized diagnostic suspicion bias by batchtesting our plasma samples and blinding our laboratory technicians to the disease status of the study patients. Another limitation of this study is that stored plasma samples were used. Owing to budgetary constraints, we did not perform these assays at the time of presentation. Blood samples were collected at the time of presentation for the specic purpose of D-dimer testing. Owing to the lengthy enrollment period that was necessary to conduct the diagnostic study,

samples were frozen and batch-tested for up to 8 years after the collection date. Although great care was taken to process and store these samples at ) 70 C during this time, it is possible that some degradation of the samples may have occurred. However, D-dimers are end-products of brinolysis, and levels in frozen samples should remain stable because there is no in vitro clearance or production. The mean levels in our samples were also consistent with those in another published study [8,9]. Performance of the assays at a central laboratory also allowed for better control of testing conditions and
2010 International Society on Thrombosis and Haemostasis

D-dimer testing in pregnant patients 1009

Sensitivity 1.00

0.75

Optimum cut-off point selected = 0.57 g mL1 dimer units

0.50 DVT 0.25 Dd Pos Neg

Pos 12 3

Neg 53 157

0.00 0.00

0.25

0.50

0.75

1.00 1-Specificity

Fig. 3. Receiver operating characteristic curve (ROC) for IL D-dimer assay. Dd, D-dimer; DVT, deep vein thrombosis; Pos, Positive; Neg, negative; Dd Pos, level at or above cut-point; Dd Neg, level below cut-point. Area under ROC curve by Wilcoxon estimate = 0.82 (95% condence interval 0.740.90).

Sensitivity 1.00
Optimum cut-off point selected = 1.50 g mL1 FEU

0.75

0.50

0.25

0.00 0.00

0.25

0.50

0.75

1.00 1-Specificity

Fig. 4. Receiver operating characteristic curve (ROC) for Sta-Lia D-dimer assay. Dd, D-dimer; DVT, deep vein thrombosis; FEU, brinogen equivalent units; Pos, Positive; Neg, negative; Dd Pos, level at or above cut-point; Dd Neg, level below cut-point. Area under ROC curve by Wilcoxon estimate = 0.85 (95% condence interval 0.780.92).

eliminated variability in assay performance over time. Nevetheless, prospective studies are needed to validate these results before the results of this study can be applied to clinical practice. It is established that D-dimer testing, combined with pretest probability using a prediction rule, is important in the diagnosis of VTE in the general population, and there is also increasing evidence supporting the use of D-dimer testing in excluding recurrent DVT [6] as well as for the prediction of future DVT recurrence [20]. We have shown that D-dimer testing could be extended to pregnant women by using higher cut-points than in the general population. The use of D-dimer testing (at higher
2010 International Society on Thrombosis and Haemostasis

cut-points), combined with pretest probability assessment, in pregnant patients for the diagnosis of DVT should now be evaluated prospectively. Conclusion With the use of higher cut-points, currently available sensitive D-dimer assays appear to be sufciently sensitive and specic for use in pregnant women with suspected DVT. Combined with pretest probability assessment, D-dimer testing may prove to be useful for the exclusion of DVT in pregnant women.

1010 W.-S. Chan et al

Sensitivity 1.00

0.75

Optimum cut-off point selected = 1.50 g mL1 FEU

0.50 DVT 0.25 Dd Pos Neg Pos 15 0 Neg 81 128

0.00 0.00

0.25

0.50

0.75

1.00 1-specificity

Fig. 5. Receiver operator curve (ROC) for Innovance D-dimer assay. Dd, D-dimer; DVT, deep vein thrombosis; FEU, brinogen equivalent units; Pos, Positive; Neg, negative; Dd Pos, level at or above cut-point; Dd Neg, level below cut-point. Area under ROC curve by Wilcoxon estimate = 0.83 (95% condence interval 0.760.90). Table 2 Test characteristics of the D-dimer assays in symptomatic pregnant patients at non-pregnant and pregnant cut-points Assay Vidas Asserachrome IL Test STA-Lia Innovance D-dimer level cut-points (lg mL)1 FEU) Non-pregnant, 0.50 Pregnant, 1.89 Non-pregnant, 0.50 Pregnant, 1.51 Non-pregnant, 0.23 Pregnant, 0.57 Non-pregnant, 0.50 Pregnant, 1.38 Non-pregnant, 0.50 Pregnant, 1.50 Sensitivity, % (95% CI) 100 93.3 100 100 100 80.0 100 93.3 100 100 (74.7100) (68.199.8) (74.7100) (78.2100) (74.7100) (51.995.7) (74.7100) (68.199.8) (74.7100) (74.7100) Specicity, % (95% CI) 10.3 78.8 12.3 73.9 17.8 74.8 22.9 75.6 6.22 61.2 (6.6315.5) (72.784.1) (8.2817.8) (67.579.7) (13.024.0) (68.380.5) (17.529.4) (69.381.2) (3.4910.6) (54.367.8) NPV, % (95% CI) 100 99.4 100 100 100 98.1 100 99.4 100 100 (80.8100) (96.2100) (83.4100) (97.0100) (88.0100) (94.299.5) (90.6100) (96.1100) (71.7100) (96.4100) Negative likelihood ratio (95% CI) 0.09 (0.010.56) 0 ()* 0.27 (0.100.74) 0.09 (0.010.59) 0 ()*

CI, condence interval; FEU, brinogen equivalent units; NPV, negative predictive value. *One or more cells contains a zero value, so 95% CI cannot be calculated. Measured in dimer units.

Addendum W.-S. Chan, A. Lee, S. Chunilal and J.S. Ginsberg were responsible for overall study design and methodology; A. Lee, F.A. Spencer, J.S. Ginsberg and M. Johnston were responsible for selection of D-dimer assays and testing; W.-S. Chan and W. Wu were responsible for statistical analysis; W.-S. Chan, F.A. Spencer, M. Crowther, M. Rodger and J.S. Ginsberg were responsible for patient recruitment and data collection; W.-S. Chan, A. Lee, F.A. Spencer, J.S. Ginsberg and M. Johnston were responsible for manuscript preparation. All other authors approved the nal manuscript. Acknowledgements The authors wish to acknowledge N. McEwen, J. McGrath and E. Ali for testing the samples. We would also like to thank M. Quilacio, A. M. Clement and P. Stevens for their assistance in patient recruitment.

J. S. Ginsberg and F. A. Spencer are recipients of a Career Investigator Award of the Heart and Stroke Foundation of Ontario. Partial funding was provided in the rst 2 years by the Heart and Stroke Foundation of Ontario (Grant no. NA 5048). No external funding was obtained for D-dimer assays or testing. Disclosure of Conict of Interests The authors state that they have no conict of interest. References
1 Sachs BP, Brown DA, Driscoll SG, Schulman E, Acker D, Ransil BJ, Jewett JF. Maternal mortality in Massachusetts. Trends and prevention. N Engl J Med 1987; 316: 60772. 2 Kaunitz AM, Hughes JM, Grimes DA, Smith JC, Rochat RW, Kafrissen ME. Causes of maternal mortality in the US. Obstet Gynecol 1985; 65: 60512.

2010 International Society on Thrombosis and Haemostasis

D-dimer testing in pregnant patients 1011

3 Rochat RW, Koorin LM, Atrash HK, Jewett JF. Maternal mortality in the United States: report from the Maternal Mortality Collaborative. Obstet Gynecol 1988; 72: 917. 4 Kierkegaard A. Incidence and diagnosis of deep vein thrombosis associated with pregnancy. Acta Obstet Gynecol Scand 1983; 62: 239 43. 5 Bates SM, Kearon C, Crowther M, Linkins L, ODonnell M, Douketis J, Lee AY, Weitz JI, Johnston M, Ginsberg JS. A diagnostic strategy involving a quantitative latex D-dimer assay reliably excludes deep venous thrombosis. Ann Intern Med 2003; 138: 78794. 6 Rathbun SW, Whitsett TL, Raskob GE. Negative D-dimer result to exclude recurrent deep venous thrombosis: a management trial. Ann Intern Med 2004; 141: 83945. 7 Kearon C, Ginsberg JS, Douketis J, Crowther MA, Turpie AG, Bates SM, Lee A, Brill-Edwards P, Finch T, Gent M. A randomized trial of diagnostic strategies after normal proximal vein ultrasonography for suspected deep venous thrombosis: D-dimer testing compared with repeated ultrasonography. Ann Intern Med 2005; 142: 4906. 8 Morse M. Establishing a normal range for D-dimer levels through pregnancy to aid in the diagnosis of pulmonary embolism and deep vein thrombosis. J Thromb Haemost 2004; 2: 12024. 9 Kline JA, Williams GW, Hernandez-Nino J. D-dimer concentrations in normal pregnancy: new diagnostic thresholds are needed. Clin Chem 2005; 51: 8259. 10 Francalanci I, Comeglio P, Liotta AA, Cellai AP, Fedi S, Parretti E, Mello G, Prisco D, Abbate R. D-dimer concentrations during normal pregnancy, as measured by Elisa. Thromb Res 1995; 78: 399405. 11 Stein PD, Hull RD, Patel KC, Olson RE, Ghali WA, Brant R, Biel RK, Bharadia V, Kalra NK. D-dimer for the exclusion of acute venous thrombosis and pulmonary embolism. Ann Intern Med 2004; 140: 589 602. 12 Chan WS, Chunilal S, Lee A, Crowther M, Rodger M, Ginsberg JS. A red blood cell agglutination D-dimer test for the exclusion of DVT in pregnancy: a prospective cohort study. Ann Intern Med 2007; 147: 165 70. 13 Bates SM, Kearon C, Crowther M, Linkins L, ODonnell M, Douketis J, Lee AY, Weitz JI, Johnston M, Ginsberg JS. A diagnostic strategy involving a quantitative latex D-dimer assay reliably excludes deep venous thrombosis. Ann Intern Med 2003; 138: 78794. 14 Kraaijenhagen RA, Piovella F, Bernardi E, Verlato F, Beckers EA, Koopman MM, Barone M, Camporese G, Potter Van Loon BJ, Prins MH, Prandoni P, Buller HR. Simplication of the diagnostic management of suspected deep vein thrombosis. Arch Intern Med 2002; 162: 90711. nchez J, Serra P, de Moerloose P. 15 Arnout M, Sales B, Arza T, Sa Clinical Management Study of Venous Thromboembolism (VTE) Using HemosILTM D-Dimer. Presented at the XX Congress of The International Society on Thrombosis and Haemostasis, 612 August, Sydney, Australia, 2005. 16 Bournameaux H, de Moerloose P, Perrier A, Reber G. Plasma measurement of D-dimer as diagnostic aid in suspected venous thromboembolism: an overview. Thromb Haemost 1994; 71: 16. 17 Perrier A, Desmarais S, Miron MJ, de Moerloose P, Lepage R, Slosman D, Didier D, Unger PF, Patenaude JV, Bounameaux H. Noninvasive diagnosis of venous thromboembolism in outpatients. Lancet 1999; 353: 1905. 18 Schutgens REG, Haas FJLM, Gerritsen WBM, van der Horst F, Nieuwenhuis HK, Biesma DH. The usefulness of ve D-dimer assays in the exclusion of deep venous thrombosis. Thromb Haemost 2003; 1: 97681. 19 de Moerloose P, Palareti G, Aguilar C, Legnani C, Reber G, Peetz D. A multicenter evaluation of a new quantitative highly sensitive Ddimer assay for exclusion of venous thromboembolism. Thromb Haemost 2008; 100: 50512. 20 Palareti G, Legnani C, Cosmi B, Guazzaloca G, Pancani C, Coccheri S. Risk of venous thromboembolism recurrence: high negative predictive value of D-dimer performed after oral anticoagulation is stopped. Thromb Haemost 2002; 87: 712.

2010 International Society on Thrombosis and Haemostasis