D-dimer: Preanalytical, analytical, postanalytical variables, and clinical applications

Abstract

D-dimer is a soluble fibrin degradation product deriving from the plasmin-mediated degradation of cross-linked fibrin. D-dimer can hence be considered a biomarker of activation of coagulation and fibrinolysis, and it is routinely used for ruling out venous thromboembolism (VTE). D-dimer is increasingly used to assess the risk of VTE recurrence and to help define the optimal duration of anticoagulation treatment in patients with VTE, for diagnosing disseminated intravascular coagulation, and for screening medical patients at increased risk of VTE. This review is aimed at (1) revising the definition of D-dimer; (2) discussing preanalytical variables affecting the measurement of D-dimer; (3) reviewing and comparing assay performance and some postanalytical variables (e.g. different units and age-adjusted cutoffs); and (4) discussing the use of D-dimer measurement across different clinical settings.

KEYWORDS:

• D-dimer
• preanalytical
• analytical
• postanalytical
• venous thromboembolism

Introduction: what is D-dimer?

D-Dimer is a generic term referring to multiple peptide fragments deriving from plasmin-mediated degradation of cross-linked fibrin. Their presence reflects concomitant activation of both coagulation and fibrinolysis [1–4]. The first step of D-dimer formation involves the thrombin-catalyzed conversion of fibrinogen into fibrin monomers (Figure 1). Human fibrinogen is a soluble plasma glycoprotein composed of three different pairs of polypeptide chains (Aα-, Bβ-, and γ-) connecting two-outer D-domains to the central E-domain [5,6]. Thrombin enzymatically cleaves two cryptic polymerization sites located on the E-domain, thus leading to the generation of both highly self-adhesive fibrin monomers and fibrinopeptides A and B [1]. Fibrin monomers then bind to one another to form a soluble network [1]. Simultaneously, the complex between soluble fibrin polymers, thrombin, and plasma factor XIII promotes the formation of factor XIIIa, which catalyzes covalent cross-linking of fibrin polymer via intermolecular bonds formed between lysine and glutamine residues, thus enabling the generation of stable and insoluble clots [1,4]. Later, the fibrinolytic pathway leads to degradation of stabilized clots through plasmin activation [4]. Plasmin is activated from plasminogen by tissue plasminogen activator (t-PA) at the fibrin surface and cleaves fibrin at specific sites [7]. The products of this reaction are the conventionally-defined fibrin degradation products (FDP), and display a vast array of molecular weights. In the first steps of fibrinolysis, FDP are large. Continual breakdown generates the fragment D-dimer/fragment E complex (DD/E), which has two covalently-bound D-domains [4]. Because of variable degrees of plasmin-mediated proteolysis, plasma samples then contain a mixture of fibrin fragment complexes containing one or several D-dimer motifs whose molecular weights range from 228 kDa (DD/E) to several thousand kDa (X-oligomers) [8,9]. These FDPs can derive both from the insoluble fibrin clot and from soluble fibrin polymers [1]. The breakdown of both fibrinogen and fibrin by plasmin is counterbalanced by multiple enzymatic modulators (thrombin-activatable fibrinolysis inhibitor, α-2-antiplasmin, and α-2-macroglobulin), whose functions are to limit fibrinolysis at the injury site [3].

Figure 1. Mechanism of D-dimer production [1,3,4,164].

The term “fragment D-dimer” was initially used to describe the final plasmin digestion products (resistant to further plasmin breakdown) of the factor XIIIa–cross-linked fibrin clot, which is the fragment DD/E [1,10]. However, the actual D-dimer antigen that can be detected by current immunoassays is not necessarily the DD/E complex [1]. In fact, the term D-dimer comprises a broad mixture of degradation products of cross-linked fibrin with molecular weights likely to range from 190 to over 10,000 kDa [9]. D-dimer fragments are mainly eliminated by renal clearance and reticuloendothelial system catabolism. The plasma half-life of D-dimer is 8 h [11], which is much longer than that of the thrombin-antithrombin complex (10–15 min) or prothrombin fragments 1 + 2 (F 1 + 2; 90 min). In normal physiological conditions, around 2–3% of fibrinogen is converted into fibrin, which instantly enters the fibrinolytic pathway. Thus, D-dimer is measurable in small amounts in healthy subjects and tends to increase with aging [4,12].

Fibrinogenolysis and fibrinolysis have to be considered differently; fibrinogenolysis leads to generation of fragments X, D, Y, and E (fibrinogen degradation products), whilst fibrinolysis leads to more complex fragments called X-oligomers that contain the D and E fragments in various sequences [1,4]. Only fibrin polymers that have undergone factor XIII-mediated cross-linking will produce fragments containing covalent bonds between two adjacent D domains (i.e. D-dimer). Therefore, D-dimer (DD/E fragment) should be considered as a specific FDP that is recognized by most of the reagent antibodies used in the laboratory assessment of thrombosis [3]. Commercial D-dimer immunoassays are currently predicted to detect an epitope on the degradation products of factor XIIIa–cross-linked fibrin. Monoclonal antibodies employed should not detect fibrinogen degradation products (e.g. X, Y, D, E) on non-cross-linked FDP, but they predictably do detect only epitopes present in the factor XIIIa–cross-linked fragment D domain of fibrin [1,4]. Gaffney et al. found that antigenic determinants of D-dimer assays are a part of the polypeptides in the D-domain, which are conformationally reactive only after factor XIIIa and plasmin have modified the protein [10]. Notably, each monoclonal antibody has its own specificity toward FDP [13].

Preanalytical variables

According to ISO 15189:2012 Medical laboratories–Requirements for quality and competence (ISO, International Organization for Standardization), the ISO standard that is used for laboratory accreditation, the preanalytical phase is defined as “processes that start, in chronological order, from the clinician’s request and include the examination request, preparation and identification of the patient, collection of the primary sample(s), and the transport to and within the laboratory, and end when the analytical examination begins” [14].

Preanalytical phase in hemostasis laboratories

It is now clearly established that most errors encountered in the hemostasis laboratory are related to preanalytical activities [15–20]. Preanalytical errors have a frequency of 60–70%, a frequency that is much higher than that occurring in the analytical (e.g. 10–15%) and postanalytical phase (e.g. 15–20%) [20]. Preanalytical errors are related mainly to intensive manual activities [20,21].

During a 2-year follow-up period, Salvagno et al. studied all preanalytical errors detected in the local coagulation laboratory, and found that the most frequent preanalytical problems were related to samples not received in the laboratory (49.3%), hemolysis (19.5%), clotting (14.2%), and inadequate sample volume (13.7%) [22]. They also showed that preanalytical problems could be identified in up to 5.5% of all coagulation samples [22]. Another study performed by Grecu et al. during a 1-year period revealed that clotted samples were the major source of preanalytical errors in their hematology laboratory (43.2%), whilst samples with inappropriate volume were less frequent [23]. Recently, Dikmen et al. performed a similar study during a 1-year follow-up and observed that clotted samples (35%) and inadequate volume (13%) were the major causes of sample rejection [24]. They found that the specimen rejection rate for coagulation tests was higher (13.3%) than all other tests (e.g. 3.2% for biochemistry tests, 9.8% for blood gases and 9.8% for urinalysis) [24].

The source of preanalytical variables impairing sample quality can be divided into three main categories: (i) sample collection (e.g. needle size, collection tubes), (ii) sample delivery to the laboratory (e.g. pneumatic tube system (PTS), temperature), and (iii) sample processing (e.g. centrifugation, hemolyzed samples). The storage and stability of samples, as well as the freeze-thaw effect, are also part of the preanalytical process [25].

Importantly, each preanalytical step is vulnerable to errors. Therefore, following preanalytical recommendations at each step of the process is essential for preserving sample integrity [25]. The impact of each of these preanalytical variables on D-dimer measurement in current hemostasis laboratories will be discussed.

Through a review of the literature of the effects of pre-analytical variables on D-dimer, it appeared that statistical analysis was frequently used to assess a significant pre-analytical change. The absence of significant bias is considered a robust indicator that a pre-analytical variable has not significantly impacted a particular analyte. However, a significant difference does not inherently correspond to clinical significance. Therefore, a clinical criterion is also recommended to accurately assess the effect of preanalytical variables on a particular analyte. Regarding D-dimer, a clinically acceptable cutoff of 10% has frequently been used to assess a significant preanalytical change (especially to assess the effect of interfering substances and for performing stability studies). Unfortunately, this 10% cutoff has not been derived from biological variation data (i.e. reference change values [26]) because such data are still lacking in the hemostasis field. Further validation of this empirical 10% cutoff and establishment of biological variation studies are hence needed [27].

Preanalytical variables affecting the D-dimer measurement

Sample collection

Butterfly devices and needle bore size

It is currently recommended to use straight needles with a diameter ranging from 19 to 22 gauge (G) [21]. Blood samples should be obtained with relatively non-traumatic venipuncture. Excess vein manipulation with the needle should be avoided to limit the risk of developing clots [25], which have been found to potentially impact D-dimer concentration [28]. The use of butterfly devices (small needle attached to flexible plastic wings and connected with a flexing tube) is usually discouraged because the passage of blood along the tubing might be a cause of enhanced shear stress, thus potentially triggering hemostasis activation, erythrocyte injury (i.e. hemolysis); they are also more expensive [21,29,30].

Interestingly, Lippi et al. found that the use of butterfly devices (21 G, 300 mm-long tubing) did not introduce a significant bias in D-dimer values measured with the Mini Vidas^® Immunoanalyzer (bioMérieux, Marcy l’Etoile, France) compared to conventional straight needles (21 G) [31]. This observation implies that the potential activation of the hemostatic system occurring within the tubing of the butterfly device, regardless of its composition, may be negligible for the D-dimer measurement.

Lippi et al. also found a clinically negligible bias using butterfly devices with different needle bore sizes (21, 23, or 25 G), when D-dimer was again measured with the Mini Vidas^® Immunoanalyzer [32]. Taken together, the use of butterfly devices, even with small size needles, seems to represent a viable option to standard straight needles [31,32]. However, straight needle venipunctures are preferable, especially when other coagulation tests are requested [33]. The use of butterfly devices may, therefore, be considered in particular individuals (geriatrics, oncology, pediatrics, or in emergency settings) [21,33]. When blood is collected using a winged collection system (or IV catheters), a discard tube is mandatory for removing air contained within the tubing, which may be associated with the collection of an inadequate volume of blood [34].

Tube material

Non-activating material (silicone-coated glass or polypropylene plastic) is preferred for hemostasis testing because this will prevent initiation of clotting due to spurious hemostasis activation in the blood collection tube [25].

Several studies were performed for comparing the use of different collection tube materials. Leroy-Matheron et al. found no significant differences in measured D-dimer values (Asserachrom^® ELISA assay) using glass or polyethylene terephthalate (PET) plastic collection tubes (Vacutainer^®, Becton Dickinson (BD)) [35]. Gosselin et al. also failed to find significant differences for D-dimer values (Advanced D-Dimer) when using plastic (Vacutainer^®, BD and Vacuette^®, Greiner Bio-one (polypropylene)) and glass citrated collection tubes (3.2%) [36]. More recently, Yavas et al. compared three plastic citrate (3.2%) tubes (Vacutainer^®, BD; Vacutainer^® Plus Plastic, BD; and Vacuette^®, Greiner Bio-one) with a standard glass tube (Vacutainer^®, BD), and also failed to find significant differences [37]. Therefore, the quality of D-dimer measurement seems to be not substantially dependent on the type of collection tube.

Anticoagulant sample (or tube additive)

The use of collection tubes containing 3.2% (105–109 mmol/L) buffered sodium citrate anticoagulant is now recommended by the Clinical Laboratory Standards Institute (CLSI) and by the World Health Organization (WHO) for the vast majority of hemostasis tests [25,34]. A blood to anticoagulant ratio of 9/1 is recommended because the sodium citrate anticoagulant can be used only in a liquid form [25]. Failure to correctly fill the citrate anticoagulant tube will typically prolong clotting times (PT, APTT, and TT), and may also lead to underestimating D-dimer and fibrinogen [28]. Therefore, the sample should be checked for quality (e.g. clotting, underfilling, or overfilling) [4]. Blood should also flow freely into the tube and be promptly mixed within 30 s after phlebotomy (i.e. from 3 to 6 complete inversions), thus ensuring the complete distribution of anticoagulant activity [25]. Serum and heparinized/EDTA plasma samples cannot be accepted because these additives dramatically interfere with clot-based assays (e.g. PT, APTT, FV, FVIII) [25,28].

However, several D-dimer assays (including POC tests) allow the use of citrated, heparinized or EDTA plasma (Pathfast^®, Tina-quant^® (Roche, Switzerland)), AQT-90^® (Radiometer, Denmark), Simplify^® (Agen Inc, Australia)), whilst others recommend only citrate anticoagulated blood (e.g. Vidas^®, BCS^®, STA-Liatest^® (Diagnostica Stago, Asnières, France), Immulite^® (Siemens, USA)) [38]. Schutgens et al. observed a higher mean D-dimer concentration (Tina-quant^® D-dimer) in heparin collection tubes (2510 µg/L), compared to citrate (2060 µg/L) albeit the difference did not achieve statistical significance [39]. The higher D-dimer concentration in heparin plasma is probably attributable to the introduction of a dilution factor in citrate-containing blood tubes. Although Vukovich et al. reported that heparinized and citrated samples yielded identical D-dimer values (D-dimer Gold^®, Agen Biochemical) samples, they recommended using a correction factor of 0.84 to take into account the dilution of the citrate anticoagulant, a process that may be confusing in routine practice [40]. Local evaluation of other matrices should be validated before they are used in clinical practice, especially when the manufacturer recommends using only citrated samples. Lippi et al. found only a modest bias of D-dimer values (Immulite^® 2000 D-dimer) between lithium-heparin and sodium citrate blood tubes [41].

The main advantage of using other anticoagulants (e.g. heparin) is the possibility to measure other analytes (e.g. electrolytes, cardiac troponins). However, the dilution factor may be a confounding factor and may require setting specific algorithms for correcting data. Therefore, other anticoagulants should be locally validated or their use should be reserved for special circumstances; thus buffered sodium citrate (3.2%, or 105–109 mmol/L) remains the recommended sample matrix [33].

Notably, D-dimer was first measured in serum samples with preliminary removal of fibrinogen to avoid cross-reactivity with polyclonal antibodies [4]. However, false-positive results were frequently encountered in patients under anticoagulant treatment, whilst false negative values were also seen when FDPs were entrapped in the clot. The loss of fibrinopeptide A has also been associated with degradation of non-cross-linked FDP [4].

Tourniquet use

Tourniquets are conventionally used to temporarily obstruct the vein flow and thereby assist the phlebotomist to identify vein access. It is commonly recommended that the tourniquet should be removed as soon as the needle is in the vein or when the first tube starts to fill and that it should never remain in place for more than 1–2 min [21,25]. The most important drawbacks of prolonged tourniquet placement include hemoconcentration and clot formation, which may jeopardize the quality of coagulation testing [21,25,32]. More specifically, Lippi et al. observed that D-dimer values (Mini Vidas^® Immunoanalyzer) were significantly increased, by 13.4%, when measured in samples collected after 3 min of venous stasis [32]. Lower differences were also observed after a 1-min stasis (mean increase of 7.9%) [32].

Sample delivery to the laboratory

Samples should be delivered to the laboratory at ambient temperature (15–22 °C), in the shortest possible time (usually <1 h) after collection [34]. The stability of D-dimer will be discussed below. Blood samples have to be delivered in a vertical position and the cap should not be removed [21,25]. The transport of blood tubes in a vertical rather than a horizontal position limits the extent of in vitro microparticle generation [21,42]. Although the presence of blood collection facilities close to the laboratory is the best option to limit pre-analytical variability, many hospitals deliver blood samples with a PTS [21]. PTS has the great advantage of reducing the turnaround time (TAT), but these systems need to be validated before use, because excessive acceleration/deceleration, radial gravity forces, vibration, and changes in air pressure may trigger platelet activation and hemolysis [21,43,44].

Schutgens et al. failed to find significant difference between D-dimer values (Tina-quant^® D-dimer) measured in samples delivered by PTS (≈100 m long) compared to those carried by hand [39]. Wallin et al. also found no substantial difference in D-dimer values (MediRox^® assay, Nyköping, Sweden) delivered to the laboratory by PTS (≈500 m long) compared to those directly collected in the laboratory [44]. Unlike these studies, Le Quellec et al. observed significant difference of D-dimer values (DDi-HS 500^®, HemosIL^®) in samples conveyed by PTS (≈2000 m long) compared to those transported with a motor vehicle (mean difference of 7.4%) [43]. Nevertheless, a disagreement at the 500 µg/L D-dimer cutoff could be observed only in 1 out of the 39 samples (510 μg/L with PTS and 490 μg/L by motor vehicle) and the overall difference was still within the intra-assay imprecision of the D-dimer assay. Although the data in the literature seemingly is reassuring about the low impact of PTS on D-dimer test results, it is advisable that each laboratory assesses its local PTS because systems are heterogeneous in terms of length, internal diameter, maximal acceleration force, and speed [43].

Sampling processing

Prior to being analyzed, samples should be carefully checked for the presence of inappropriate additive, identification errors, insufficient volume (9/1 ratio not respected), and even for the presence of clots. Unsuitable samples should then be rejected and test results suppressed to preserve patient safety [21].

Centrifugation

Except for point of care (POC) whole blood D-dimer analyzers, samples are centrifuged at room temperature at 1500×g for at least 15 min, so that plasma can be reliably separated from the cellular components [34]. According to Bernard et al., however, no difference could be observed in D-dimer values (Vidas^® assay) when samples were centrifuged at 4500×g for 2 min rather than at 1500×g for 15–20 min [45]. The use of a shorter centrifugation time will be indeed useful in emergency settings to reduce the TAT. The samples can be kept at room temperature for up to 8 h and at 4 °C for up to 24 h (unspun citrated blood samples) before measurement of D-dimer [4,46,47]. Elf et al. centrifuged blood samples at 4 °C and measured D-dimer with two quantitative immunoassays (Innovance^® and AxSYM^® (Abbott, USA)) [48], although there was no evidence supporting cold centrifugation for D-dimer testing.

Interfering substances

The types of interferences occurring in the preanalytical phase of coagulation testing are typically classified into paraproteinemia, icterus, lipemia, and, the most studied, hemolysis [20].

In vitro hemolysis still represents one of the most frequent causes of preanalytical problems in clinical laboratories, with a prevalence ranging between 30–70% of all unsuitable specimens [22,49–52]. Causes of hemolyzed samples could result from patient’s characteristics or diseases (e.g. hemolytic anemia, metabolic disorders, infectious agents, hemoglobin-based blood substitute), phlebotomy (e.g. needle G, tourniquet time, traumatic draw, no mixing or vigorous mixing of blood tubes), sample transport (e.g. PTS, time, emergency or intensive care units), processing (e.g. delay before centrifugation, specimen re-spun, time/temperature of transport), and storage (e.g. temperature and duration) [50,51]. The interference will also depend on the analytical method used (photometric, clotting, immunoassay). The CLSI guideline H21-A5, collection, transport, and processing of blood specimens for testing plasma-based coagulation assays and molecular hemostasis assays, advises against analyzing samples with visible hemolysis because of possible bias in coagulation tests from release of pro-coagulant factors from injured cells [34].

Lippi et al. showed that a significant increase in D-dimer level (Vidas^® assay) could be observed in samples containing a final whole blood lysate concentration of at least 2.7% obtained with a freeze-thaw cycle (−70 °C) [53]. However, a clinically significant variation (10%) was observed only above cell-free hemoglobin concentrations of 13.6 g/L (or final lysate concentration of ≈6.4%) [51,53]. Another study by the same authors evaluated the impact of mechanical hemolysis (syringe equipped with a fine needle (30 G, 0.3 × 8 mm)) on two different D-dimer assays (HemosIL AcuStar^® (Werfen, USA) and HemosIL HS^® (Werfen, USA)) [49]. The D-dimer concentration measured with the HemosIL HS^® reached a clinically significant decrease (−5%) from a hemolysis index between 5.5–7.0 g/L of cell-free hemoglobin. Greater hemolysis (i.e. 11.5–15.0 g/L of cell-free hemoglobin) was necessary to achieve the same level of D-dimer decrease (−5%) with the AcuStar D-dimer assay [49]. However, the bias never exceeded the clinically acceptable cutoff of 10%. D’Angelo et al. more recently reported a significant difference of D-dimer values with mild hemolysis (cell-free hemoglobin of 0.5 g/L) obtained by either heat-shock at −80 °C (+6.2%) or by mechanical injury (rotating blade homogenizer for 30 s) (+14.9%)) when measured on a BCSxp^® assay (Siemens Healthcare) [54].

The comparison of studies assessing hemolysis interference is challenging. First, different D-dimer assays with different reagent formulations were used. Secondly, different techniques to obtain hemolyzed specimens were used (freezing and thawing whole anticoagulated blood, mechanical lysis of whole anticoagulant blood by aspiration through a fine needle or by a rotating blade homogenizer). While spiking samples with hemolysate or pure hemoglobin solution is considered unsuitable (lysis of leukocytes and platelets is taken into account), mechanical hemolysis obtained with a syringe equipped with a fine needle is preferred (i.e. it more closely reproduces the breakdown of cells due to traumatic blood collection) to the freeze-thaw method, which also suffers from a lack of standardization across studies (e.g. different temperatures (−70 °C [53] or −80 °C [54]) and duration of freezing [51]. Thirdly, the use of a clinical cutoff derived from biologic variability studies should be preferred to the analytical bias assessed [51]. Fourth, the selection of blood samples may produce different results and conclusions; data generated in healthy volunteers may not necessarily be similar to that obtained in intensive care unit patients, or in those undergoing multiple therapies (namely, antiplatelet or anticoagulant drugs) [51].

The rejection of all hemolyzed samples is recommended by the CLSI. The majority of hemolyzed samples (±95%) from clinical laboratories are only mildly hemolyzed (cell-free hemoglobin 0.3–0.6 g/L) [49,50,55]. Thus, when the concentration of cell-free hemoglobin is below a non-interference limit (i.e. <3 g/L), D-dimer values may still be reliable and may be safely reported to the clinicians [49]. Notably, avoiding rejection of all hemolyzed specimens will reduce additional blood sampling and may shorten the clinical decision-making process, thereby favorably impacting staff workload, patient comfort, and costs [49,51].

The effect of lipemia, icterus, and paraproteinemia has been less widely discussed in the literature [51,56,57]. In an early report, Pittet et al. failed to find any impact of lipemia (triglycerides concentration of 9.7 g/L) and icterus (bilirubin concentration of 485 mg/L, 829.5 µmol/L) on the Vidas D-dimer assay [58]. More recently, La’ulu et al. found similar findings on the AxSYM D-dimer assay, with no clinical significant difference (<10% of deviation from the baseline) following addition of bilirubin (final concentration of 292 mg/L, 499.4 µmol/L) and triglycerides (final concentration of 41.6 g/L) [59].

The Innovance D-dimer assay on the Sysmex CA-7000 also appeared to be unaffected by bilirubin (both free and conjugated bilirubin up to 1000 mg/L, 1710.4 µmol/L) [60]. However, spurious D-dimer results were observed on the STA analyzer after adding high concentrations of free bilirubin (from 800 mg/L, 1368.3 µmol/L) [60]. According to Chen et al., D-dimer, as measured with the Sysmex CS-5100 analyzer, was also free from triglycerides and total bilirubin interferences [61]. Moreover, the HIL check was able to detect lipemic samples (triglycerides concentrations >4 g/L) on the CS-5100 D-dimer assay in three out of four samples [62]. The undetected sample had a very high concentration of triglycerides (9.43 g/L).

Regarding the paraproteinemia interference, Mugler et al. reported a case of a patient with Castleman disease associated with an elevated D-dimer level [57]. During investigation of this finding, no D-dimer band could be noticed on Western blot compared to D-dimer measured with the STAR coagulation instrument. The patient’s Castleman disease-associated monoclonal gammopathy was identified to be the source of the false positive D-dimer result. Paraproteinemia interference was also hypothesized in another report published by Roller et al. and it could not be excluded from the reports of Wu et al. and Huang et al. [63–65].

Spurious D-dimer concentrations due to heterophilic antibodies have also been reported in the literature [56,66,67]. Potential interferences in D-dimer testing should always be suspected whenever clinical discrepancies arise. Heterophilic blocking reagents could be used to overcome the problem [56,57]. The comparison with another method to point out interferences represents also a good choice to search for interference [57,68].

Stability, storage, and freeze-thaw effects

It is currently recommended that samples should be maintained at room temperature (15–25 °C) for no more than 4 h before testing [21]. However, and as discussed below, several studies have shown that D-dimer may be stable for a longer period. This is of particular concern because clinicians frequently ask to add the D-dimer analysis to samples already delivered from the emergency room.

Detailed information on D-dimer stability is shown in Table 1. With the exception of the publication of Schutgens et al., citrated specimens were used to assess the stability of this analyte [39]. The D-dimer stability was studied extensively in both plasma and whole blood. In brief, D-dimer is stable for at least 24 h (in plasma/whole blood) at room temperature (RT) or at 2–8 °C with many different D-dimer immunoassays. Only Böhm-Weigert et al. and Toulon et al. suggested, respectively a 6- and 8-h D-dimer plasma stability, but this was because they did not extend their analysis for longer periods [47,69]. Frozen samples may also be used when longer storage periods (i.e. months or years) are needed, especially for research purposes [69–72].

Table 1. Stability of D-dimer.

Stability	Conditions	Anticoagulant	Plasma/whole blood	D-dimer assay	Subjects	Stability criteria	Reference
24 h	RT	Heparin	Plasma	Tina-quant^® (Roche)	17 patients	Student t-test and regression equation	[39]
24 h	RT	Citrate	Plasma	Tina-quant^® (Roche)	15 patients	Student t-test and regression equation	[39]
6 h	RT	Citrate	Plasma	Innovance^® (Siemens)	40 patients	10% deviation from baseline, regression equation and discordance at the cutoff level of 0.5 mg/L FEU	[69]
24 h	RT	Citrate	Plasma	Innovance^® (Siemens)^a	80 patients	10% deviation from baseline, analysis of variance, regression equation and Pearson correlation coefficient	[78]
24 h	RT	Citrate	Whole blood	Vidas^® (bioMérieux)	117 patients	Spearman correlation coefficient, regression equation and discordance at the cutoff level of 500 ng/mL FEU	[79]
24 h	RT	Citrate	Whole blood	Innovance^® (Siemens)	44 patients	10–20% deviation from baseline, Student t-test, and regression equation	[76]
24 h	RT	Citrate	Whole blood	ACL-TOP^® (Werfen)	26 patients	Wilcoxon’s paired t-test, regression equation, and bias plot	[46]
52 h	RT	Citrate	Whole blood	Asserachrom^® (Stago)	59 patients	Analysis of variance, 10% deviation from baseline	[75]
8 h	RT	Citrate	Whole blood	ACL-TOP^® (Werfen)	144 patients	Analysis of variance, Student t-test or Wilcoxon signed rank test, Bland-Altman plot and discordance at the cutoff level of 0.5 ng/mL FEU	[47]
24 h	2–8 °C	Citrate	Plasma	Innovance^® (Siemens)	40 patients	10% deviation from baseline, regression equation, and discordance at the cutoff level of 0.5 mg/L FEU	[69]
24 h	4 °C	Citrate	Plasma	Innovance^® (Siemens)^a	80 patients	10% deviation from baseline, analysis of variance, regression equation and Pearson correlation coefficient	[78]
24 h	4 °C	Citrate	Plasma	Vidas^® (bioMérieux)	20 patients	Wilcoxon’s paired t-test, 10% deviation from baseline	[77]
24 h	4 °C	Citrate	Whole blood	ACL-TOP^® (Werfen)	26 patients	Wilcoxon’s paired t-test, regression equation, and bias plot	[46]
24 months	−24 and −75 °C	Citrate	Plasma	STA-Liatest^® (Stago)	Plasma pool (6 patients)	Statistical change^b, 5–10% deviation from baseline	[72]
2 weeks	−20 °C	Citrate	Plasma	STA-Liatest^® (Stago)	23 hV and 18 patients	Paired t-test, 10% deviation from baseline	[71]
36 months	−60 °C (or less)	Citrate	Plasma	Innovance^® (Siemens)	40 patients	10% deviation from baseline, regression equation, and discordance at the cutoff level of 0.5 mg/L FEU	[69]
9 years	−80 °C	Citrate	Plasma	STA-Liatest^® (Stago)	60 patients	Wilcoxon’s paired t-test	[70]

RT: room temperature; FEU: fibrinogen equivalent units; HV: healthy volunteers.

^ausing the Sysmex^® CA 7000 platform.

^bspecific test not mentioned.

Several criteria are available to assess stability of analytes [73]. As discussed above, statistical analysis is frequently used, and the absence of significant bias is considered a robust indicator of stability. However, a significant difference does not inherently correspond to clinical significance [74]. Therefore, a clinical criterion is also recommended to accurately assess stability of a particular analyte. It has also been demonstrated that the use of different stability criteria may strongly impact the result of stability studies [73]. Therefore, the use of multiple criteria is encouraged. All studies published so far used multiple stability criteria except that performed by Betsou et al. [70]. The review of D-dimer stability studies showed that a clinical cutoff of 10% was most frequently used [69,71,72,75–78]. However, this 10% cutoff has not been derived from biological variation data. Further validation of this empirical 10% cutoff is needed, even though no discordance could be found in the statistical approach available in the literature. The analysis of discordance based on the cutoff used to rule out venous thromboembolism (VTE), as evaluated by Caliezi et al., Böhm-Weigert et al., and Toulon et al. [47,69,79], is also a criterion that seems reasonable.

The studies of Schutgens, Zürcher, and Gosselin also showed that the freeze-thaw procedure did not significantly affect D-dimer concentration measured with three different D-dimer assays [39,75,80]. Böhm-Weigert et al. evaluated the impact of four freeze-thaw cycles (−60 °C or less) and failed to find clinically significant changes of plasma D-dimer values (mean deviation always <10%) [69].

In conclusion, while D-dimer is mainly used for emergency diagnoses and when rapid measurement is needed, stability information is required when samples are stored in central laboratories and when they are used in a research environment [33]. Table 2 summarizes specific pre-analytical data regarding D-dimer testing.

Table 2. Specific pre-analytical data regarding D-dimer testing.

Pre-analytical variables	General recommendations in hemostasis laboratories	Specific data regarding D-dimer
Sample collection
Needle bore size	19–22 G	23–25 G also tolerated
Butterfly devices	Discouraged	Tolerated
Tube material	Non-activating material (silicone-coated glass or polypropylene plastic)	Glass or plastic
Anticoagulant sample	Sodium citrate 3.2% (105–109 mmol/L)	Sodium citrate or heparin^a
Tourniquet use	Removed as soon as the needle is in the vein (max 1–2 min)	Longer tourniquet use (i.e. 3 min) not tolerated
Sample delivery to the laboratory	At RT (15–22 °C), in vertical position, usually <1 h	PTS tolerated
Sample processing
Centrifugation	At RT, 1500×g for at least 15 min	Faster protocol allowed (at RT, 4500×g for 2 min)
Interfering substances	Do not analyze samples with hemolysis	Cell-free hemoglobin i.e. <3 g/L tolerated
Stability, storage and F/T effects	At RT (15–22 °C), no more than 4 h	At least 24 h at RT or at 2–8 °C or years at −60 to −80 °C No impact of F/T procedure

RT: gauge; RT: room temperature; PTS: pneumatic tube system; F/T: freezing/thawing.

^acorrection factor needed (dilution).

Analytical variables

D-dimer assays

The first generation of D-dimer assays, which were developed in the 1970s, was capable to detect both fibrinogen and FDP by means of polyclonal antibodies [2]. These immunoassays could be performed only in serum to avoid the cross-reactivity of fibrinogen present in high concentrations in plasma. However, false-positive results were encountered in patients taking anticoagulants, and false negative results were encountered in samples that developed clots and when degradation products absorbed into the clot.

The loss of fibrinopeptide A during serum preparation impairs detection of non-cross-linked FDP [4]. The measurement of FDP was initially performed using (among others) methods based on latex fixation and agglutination, hemagglutinin inhibition, staphylococcal clumping, immunoelectrophoretic, and immunodiffusion [3].

A significant gain of specificity and sensitivity was reached with the introduction of monoclonal antibodies targeting D-domains in the early 1980s [3]. These immunoassays allowed the specific targeting of D-dimer epitopes, which were lacking in both FDP and non-cross-linked fragments of fibrin [4,81]. The use of plasma was, therefore, possible because of minimal fibrinogen (or FDP) cross-reactivity. Since the first monoclonal antibody developed by Rylat et al. in 1983 (3B6), more than 20 monoclonal antibodies have been produced and used in laboratory practice [9,81]. First generation methods were represented mainly by qualitative latex agglutination immunoassays using antibody-coated latex microparticles and needing visual inspection of agglutination [1,3,4,82]. The development of automated second-generation latex agglutination immunoassays (or the latex-enhanced immunoturbidimetric assay) allowed quantitative measurement of D-dimer through recording aggregation rate in response to D-dimer [1,2]. Once D-dimer was added, latex microparticles agglutinated, thus preventing the light to pass through the solution. The increase of light absorbance measured with a photometer was directly proportional to the D-dimer concentration [3]. Latex-enhanced immunoturbidimetric assays are rapid and achieve comparable sensitivity to conventional enzyme-linked immunosorbent assays (ELISA) [3].

Microplate ELISA assays were employed before latex agglutination assays, mainly for research purposes [1]. In these assays, a capture antibody binds D-dimer antigen on a plate. After incubation, a labeled antibody is added to the well and binds the immobilized D-dimer antigen (sandwich assay) to promote a colorimetric reaction [1,2]. Although microplate ELISAs are very sensitive to D-dimer and have long been considered the reference method [1–3], they are manual, require technical skills, are time-consuming (i.e. around 2–4 h), and are plagued by a high degree of analytical imprecision [2,83].

In the mid-1990s, bioMérieux developed an ELISA-based assay with fluorescence end-point detection (enzyme-linked immunofluorescence assay (ELFA)) [3,58]. This assay displayed sensitivity and specificity similar to those of microplate ELISAs, with the great advantage of being automated, thus generating more precise results in a shorter time (30 min) [3,58]. The Vidas^® assay is still considered the reference commercial quantitative immunoassay [84–86] and is the most clinically validated D-dimer measurement technique [38,48,85,87–89].

More recently, chemiluminescent enzyme immunometric assays that display similar sensitivity to ELISAs and latex-enhanced immunoturbidimetric assays have been developed [3,41,84]. Magnetic particles coated with monoclonal antibodies specific to D-dimer are used; incubation of anti-D-dimer antibody labeled with isoluminol generates a chemiluminescent reaction that is directly proportional to the D-dimer concentration [3].

Utilization of POC D-dimer assays is especially attractive for general practitioners (GP) because quantitative D-dimer assays are not always readily available in rural areas, during weekends, or at home visits [90–92]. In such settings, the GP had to refer the patient to a central laboratory facility [91]. The main interest in POC D-dimer assays is the possibility to rapidly screen patients for thromboembolic disease, thus leading to a decrease in overcrowding in urgent care facilities [3,90,93,94]. The majority of POC tests use whole blood, are homogenous, use monoclonal antibodies and have a short TAT [3,90]. In hemagglutination assays (e.g. SimplyRED^®), bivalent antibodies are specific to red blood cells and D-dimer. In the presence of D-dimer in the blood, red cell agglutination occurs and provides a qualitative result [3,4]. Detection based on immunochromatography (e.g. Clearview Simplify^®), fluorescence (e.g. Vidas^® mini analyzer, Triage^®, Stratus CS^®), and chemiluminescence (e.g. Pathfast^®) also exist. A few semi-quantitative assays remain in use (e.g. Dade Dimertest, NyoCard^®) [3,95]. The recent review of Riley et al. compiled the different POC D-dimer assays available on the market [3]. Table 3 summarizes the main characteristics of D-dimer assays.

Table 3. Characteristics of D-dimer assays [2,3,41,84,91,165,174].

	ELISA	ELFA	Unenhanced latex agglutination assay	CLIA	Latex-enhanced immunoturbidimetric assay	POC assay
Type	Quantitative	Quantitative	Qualitative/semi-quantitative	Quantitative	Quantitative	Qualitative/quantitative
TAT	2–4 h	35–40 min	Rapid	25–40 min	15 min	2–20 min
Pros	Considered as the gold standard, sensitivity, observer-independent	Considered as reference method, most validated method, sensitivity, automation, wider linear range, automated, observer-independent	Rapid, inexpensive	Sensitivity, rapid, automated, observer-independent	Sensitivity, automated, rapid, observer-independent	Readily available, fast, higher specificity, whole blood
Cons	Highly manual, technical skills, time-consuming, not optimal linear range, moderate specificity	Moderate specificity	Moderate sensitivity, manual, observer dependent	Lack clinical validation, moderate specificity	Moderate specificity	Sensitivity, not all FDA cleared, observer dependent, manual
Examples	Asserachrome^® (Stago), Enzygnost^® (Dade Behring)	Vidas^® (bioMérieux)	Dimertest latex^® (IL), Fibrinosticon^® (bioMérieux), Dade Dimertest^® (Siemens)	AcuStar^® (Werfen), Immulite^® (Siemens)	Tina-quant^® (Roche), STA-Liatest^® (Stago), HemosIL HS^® (Werfen) Innovance^® (Dade-Behring)	SimpliRed^® (Agen), Clearview Simplify^® (Agen)

ELISA: Enzyme-linked immunosorbent assay; ELFA: Enzyme-linked immunofluorescence assays; CLIA: Chemiluminescent enzyme immunometric assay; POC: Point of care.

Inter-laboratory variation

Variability across assays

The fibrin assay comparison trial (FACT) study, published in 2001, evaluated 23 quantitative D-dimer assays (15 latex-enhanced immunoassays, 6 ELISAs, and 2 membrane-based immunoassays) [96]. The authors found that the mean values obtained in 39 samples varied from 630 µg/L to 13,350 µg/L (≈21-fold), with two assays displaying significant cross-reactivity toward fibrinogen degradation products. It was also found that ELISAs and latex-enhanced immunoassays were more reactive to low-molecular-weight cross-linked fibrin and high molecular weight fibrin, respectively [96]. Accordingly, the study of Meijer et al., based on 357 laboratories using the 7 frequently-used D-dimer immunoassays, found that some assays gave 20 times higher D-dimer concentrations compared to others [97]. Another study from 423 laboratories comparing D-dimer values also showed high variability in results close to the cutoff used for VTE exclusion [98]. In 2014, in a survey involving 3800 laboratories, the Coagulation Resource Committee of the College of American Pathologists (CAP) found that the inter-method coefficient of variation (CV) was as high as 42% [99]. Based on four UKNEQAS reports (April 2017, July 2017, September 2017, and January 2018), the inter-method CVs (after the exclusion of outliers) were reported to be 33.4%, 38.2%, 41.8%, 30.9% and 18%, 19.3%, 17.7% and 19% for non-FEU and FEU units, respectively; these four surveys included between 736 and 752 laboratories.

Calibrators

The controlled lysis of fibrin clots is used mainly to obtain calibration materials [9]. Hence, manufacturers have to ensure that lysis is reproducible in order to maintain the same variety of size degradation products, because the sensitivity of the assay may change according to the relative amounts of high molecular weight fibrinogen (HMWF) or low molecular weight fibrinogen (LMWF) [9,96]. D-dimer assays may be calibrated with purified fibrin fragment D-dimer equivalents or according to the fibrinogen amount used to prepare the calibrator [9]. D-dimer units thus may vary according to the type of calibrator used, i.e. D-dimer units (DDU) or fibrinogen equivalent units (FEU) [9,96,100].

Standardization versus harmonization

The heterogeneity of fragments derived from plasmin digestion of cross-linked fibrin (from LMWF to HMWF) [9], the use of monoclonal antibodies with different specificities towards D-dimer epitopes [1,96,101,102], the lack of international certified internal controls or calibrators [103] and the use of different units and clinical cutoffs are the leading sources of the large inter-laboratory variability, thus making D-dimer assay standardization a challenging, if not impossible, target [9,13,99,101,102,104–108]. Because the requirements for D-dimer assay standardization cannot be met, less stringent harmonization procedures, based on the use of mathematical models, have been proposed to make results obtained with different assays more comparable [96,97,100,102].

The first attempt to achieve D-dimer assays harmonization was published in 1997 by Nieuwenhuizen et al. [102]. Different pools of patients with various diseases were prepared and tested with 5 different D-dimer assays (one microlatex and four microplate ELISA), and a “pool consensus value” (corresponding to the mean D-dimer result of each assay) was assigned to each pool. Results of each pool obtain from the individual assays were then divided by the corresponding “pool consensus value” and averaged. Samples used to prepare pools were also measured to verify that this model could be broadly implemented. The squared regression coefficient values ranged from 0.7 to 0.92, which led the authors to conclude that the use of a conversion factor may be a feasible approach to harmonized data generated with different assays [102].

The second attempt was published by Dempfle et al. in 2001. Overall, 39 individual samples with 23 D-dimer assays (including microlatex-enhanced, membrane-based and ELISA assays) were tested [96]. For each assay, a conversion factor was calculated by using the median value for each sample measured with all assays, and for each assay, the median value obtained with all samples. The multiplication of individual sample assay values by the corresponding conversion factor was found to be effective to improve the correlations among most assays.

The third attempt was conducted by Meijer et al. in 2006 [97]. A plasma pool of 50 patients diluted with normal plasma was used to prepare five different samples that were then distributed in an external quality control survey to 502 participants using seven different D-dimer assays. For each D-dimer assay, the mean results of each sample were plotted against the amount of pool added (assay-specific regression line). The median results of the mean results obtained with each D-dimer assay were also plotted against the amount of pool added (reference regression line). Harmonized results could, therefore, be calculated from the measured result by using the intercept and the slope of both regression lines (assay-specific and reference). The authors showed that a significant decrease in between-assays variation was observed by introducing this model.

The fourth attempt was published by Jennings et al. in 2007 [100]. Three calibrators and two test samples were delivered to more than 500 laboratories that used 9 different D-dimer techniques and that participated in the UK NEQAS external quality survey. Individual laboratory results for calibrators were plotted against the median results obtained with all the D-dimer immunoassays. The individual regression line was used to convert data generated on the two test samples into harmonized results. This approach was effective in improving the between-center agreement after calibration, with significant improvement in inter-laboratory variability for the two samples (from 25.9% to 11.6% and from 22.4% to 7.7% for results reported as FEU; from 55.3% to 21.6% and from 40.8% to 11.6% for results reported as DDU).

Harmonization of D-dimer assays may be particularly useful for diagnosis and monitoring of disseminated intravascular coagulation (DIC), because the high inter-variability between D-dimer assays [96,98] may directly impact the DIC score and lead to misclassifications. Harmonization of D-dimer results would provide better consistency and improve the calculation of DIC scores [9]. However, the inherent heterogeneity and the challenges of the harmonization efforts make the identification of a unique diagnostic cutoff even more difficult and clinically questionable, because false-negative results have an unfavorable impact on patient management [9].

Nevertheless, the above-mentioned approaches indicate that harmonization is feasible, and it should be considered on a larger scale [103]. Further validation of these models with more assay systems is needed to confirm whether such models may be generalized to all D-dimer immunoassays [9,97,100]. Recently, a call for harmonization was raised by the Fibrinolysis and DIC Standardization Subcommittees of the ISTH [103]. Utilization of a large number of samples from patients with various clinical conditions may be useful to produce a stable freeze-dried reference material containing high concentrations of D-dimer derived from both LMWF and HMWF [103]. A range of D-dimer values could be generated from this reference material, to obtain a reference regression line for D-dimer immunoassays. The use of monoclonal antibodies displaying high affinity to low and middle molecular weight FDP species combined with monoclonal antibodies targeting higher molecular weight forms may also represent a feasible alternative [101]. Continuing discussion among manufacturers, scientists, and clinicians is essential to achieve better harmonization.

Recommendations on the performance of D-dimer assays

Because of the heterogeneity of the different D-dimer immunoassays (i.e. format, antibodies) and the lack of standardization/harmonization, the analytical and clinical performances of the local method should always be evaluated before implementation in management strategies for VTE [109,110].

The cutoff value used to exclude VTE needs to be confirmed locally, using, as suggested by the British Committee for Standards in Haematology guidelines, a minimum of 200 subjects [1,111]. Several manufacturers also recommend local validation of cutoffs [1]. However, this approach may not be practical for all laboratories; thus the cutoff proposed by the manufacturer may be used [4], provided that reliable validation studies have been conducted elsewhere and that no significant batch-to-batch variability has been found [9]. Cut-offs have been clinically validated in prospective studies for a number of assays (e.g. Vidas^®, AxSYM^®, STA-Liatest^®). This may be a consideration when choosing a D-dimer assay [1,4,48,85,87,88,107,112]. Otherwise, comparison with validated assays is encouraged. Manufacturers should stay abreast of the recent literature regarding the use of their immunoassays and update their cutoffs when indicated [113].

A CAP survey showed that 488 laboratories out of 1506 in the USA were using cutoff values higher than those recommended in the literature or by the manufacturer [99]. A European survey also highlighted that 24% and 55% of participants used lower or higher cutoffs, respectively, than those recommended [9].

A recent Italian consensus document recommends the use of certified, quantitative immunoassays for testing the plasma of emergency department patients [33]. It was also recommended that D-dimer immunoassays should have an imprecision <10% close to the diagnostic cutoff and a measurement range and linearity of 50–5000 µg/L FEU. Importantly, D-dimer assays should not react with fibrinogen or FDP, and preferably should not react with fibrin and fibrinogen fragments released from proteolysis mediated by various enzymes [103].

Clinicians need to be aware of the performance characteristics of the D-dimer assay they are using because D-dimer comprises a mixture of degradation products of cross-linked fibrin with heterogeneous molecular weights [1,9,105]. Moreover, the laboratory should always use clinically validated cutoff values, because this diagnostic threshold plays a crucial role in clinical decision making [1].

Postanalytical variables

According to the ISO 15189:2012 standard, the postanalytical phase is defined as “processes following the examination including systematic review, formatting and interpretation, authorization for release, reporting and transmission of the results, and storage of samples of the examinations” [14].

Although it is now clearly established that most problems in the hemostasis laboratory emerge from the preanalytical phase (60–70%), [20] the postanalytical phase is still the cause of a large number of diagnostic errors (15–20%) that can jeopardize patient safety [105]. Here we discuss postanalytical issues that are specific to D-dimer measurement.

D-dimer units

Two units of D-dimer measurement are currently used in clinical laboratories: FEU and DDU. On the one hand, FEU compares the mass of D-dimer to that of fibrinogen, and the calibrators are prepared from plasmin degradation of purified fibrinogen clotted in the presence of factor XIII. On the other hand, DDU determine the mass of the estimated weight of D-dimer, and the calibrators are composed of purified D-dimer [4]. FEU (340 kD) and DDU (195 kD) may be used interchangeably if multiplied by 2 as FEU is roughly two times the mass of one DDU [2–4,9,33]. Another difficulty arises from using as many as 7 different measure units (i.e. ng/mL, mg/L, μg/L, μg/mL, g/L, g/mL, and mg/dL) for reporting D-dimer results [3]. Furthermore, the same D-dimer assay is currently reported in different units [2,114,115]. Therefore, at least 14 combinations for D-dimer measurement coexist.

As recently studied by Olson and Lippi, the majority of clinical laboratories (59% and 60%, respectively) used FEU for D-dimer measurement, whilst the leading measure unit was mg/L, followed by ng/mL, in both surveys [99,105]. Among the measure units that can be adopted, “µg/L” (or “ng/mL”) is probably the unit that best approximates the Système Internationale (SI) and is also recommended by the Italian Consensus document [33,105]. Some laboratories (8% in the CAP survey) did not even know the type of measurement unit they were using [9,99,116].

The use of different units (FEU or DDU, as well as the different measure units) is challenging for clinicians, causes confusing and potentially leads to the misclassification or misdiagnosis of patients [3,4,105]. The fact that 33% of 1500 US laboratories changed their D-dimer units from those recommended by the manufacturer made the situation even more complicated [99]. These variations add additional confusion to diagnostic reasoning, especially when neighbor laboratories use other units [99]. Clinical laboratories should communicate any change in D-dimer measurement units to their stakeholders in a timely manner. Even more importantly, reaching a universal agreement on a uniform measure unit is a compelling need to achieve harmonization of D-dimer immunoassays [99,103].

Age-specific cutoffs

D-dimer values typically increase in parallel with aging, thus leading to a high proportion of elderly patients with D-dimer levels higher than the conventional cutoff set at 500 μg/L FEU [1,4,33,99,103,105,117,118]. The use of age-adjusted cutoffs (i.e. [age-adjusted cutoff, μg/L FEU] = [age in years]) × 10) is now universally recommended for reporting results of D-dimer testing (see below) [33,118–120]. Use of these cutoffs would enable a substantial increase in the positive predictive value (PPV) without significantly impairing the negative predictive value (NPV), and ultimately improve the clinical usefulness of D-dimer measurement in elderly patients (i.e. aged 50 years or older) with low clinical probability [12,33,118–120]. The test most validated for this purpose is the Vidas^® assay (bioMérieux) [120]. However, the age-adjusted strategy seems to be less useful with the STA-Liatest^® assay [121]. The analysis of different populations with different immunoassays may explain, at least in part, such discordances. This emphasizes that the results obtained with specific D-dimer assays cannot be extrapolated from one test to another [121].

However, an international survey on the reporting of D-dimer test results showed that published recommendations on the use of age-adjusted cutoffs have not been broadly implemented (i.e. less than 10% of clinical laboratories have implemented this approach) [105].

Along with the 14 combinations of D-dimer units, the use of age-adjusted cutoffs further complicates clinical decision making – there are nearly 30 different possibilities for reporting D-dimer test results [105]. Worldwide efforts are needed to standardize D-dimer test result reporting, to pave the way for widespread implementation of age-adjusted cutoffs [106]. Specific cutoffs to be used for pregnant women and children have also been proposed [106,122,123]. However, as D-dimer increases physiologically throughout pregnancy, the identification of an accurate D-dimer diagnostic threshold will be challenging [124]. The Diagnosis of PE in Pregnancy (DiPEP) biomarker study recently showed that no biomarker has diagnostic efficiency for diagnosing VTE in pregnancy or in the puerperium [125].

Turnaround time

The TAT is a crucial aspect in D-dimer reporting because this test is mostly used in urgent clinical conditions. The Italian consensus document has recently set an overall TAT <1 h, which seems suitable for managing the vast majority of urgent test requests [33].

The use of D-dimer immunoassays characterized by a wide linearity range (i.e. up to 5000 µg/L) without performing additional external dilutions, along with the use of faster and validated centrifugation processes, PTS, or reliable POC analyzers present valuable approaches to reduce the TAT [3,33,39,43–45,92,107]. Manual ELISAs do not meet the recommended TAT, whilst recent central laboratory D-dimer assays, with analytical process times ranging between 15–40 min, are more likely to fulfill the 1-h criterion (Table 3).

Notably, owing to a half-life of 6–8 h for D-dimer, repeating the test within this time interval has no clinical basis [33]. In a European study, 81% of participants stated that they measure D-dimer 24 h per day [98].

Clinical applications

More than 30 years ago, D-dimer was found to be a reliable biomarker of both coagulation activation and fibrin digestion [8]. Since then, D-dimer has been used in several clinical conditions. Most widely, D-dimer is used for ruling out VTE in patients with low-medium clinical probability, where D-dimer is now considered the biochemical gold standard [1,4,33,98,126]. D-dimer testing is also useful for assessing the risk of recurrent thrombosis and for guiding anticoagulant therapy, for diagnosing and monitoring DIC, for excluding acute aortic dissection (AAD), and for predicting and managing thrombotic complications in patients with severe infections and sepsis [1,4,127]. Other applications have been studied (e.g. prognostication of peripheral artery disease, identification of vaso-occlusive crisis in sickle cell disease, screening of intracardiac thrombus, prediction of VTE in sleep apnea), but no clinical validation has so far been achieved for these conditions [4,128,129].

Below, we discuss the role of D-dimer for exclusion of VTE, for prediction of VTE recurrence, for prediction of thrombosis in medically hospitalized patients, and for diagnosis and monitoring of DIC.

We briefly present here some indications discussed in the literature, namely cerebral venous thrombosis (CVT), AAD, and acute mesenteric ischemia (AMI).

Some indications for measuring D-dimer

Cerebral venous thrombosis

According to the statement of the American Heart Association (AHA) and of the American Stroke Association (ASA) guidelines, a normal D-dimer level obtained with a sensitive assay may help to identify patients with a low probability of CVT. However, the class and level of evidence are low (Class IIb; Level of Evidence B) and the use of D-dimer is considered superfluous when the clinical suspicion of CVT is high [130]. Because of the variability of presentation, the diagnosis of CVT is rarely confirmed by neuroimaging investigations, which remains, however, the gold standard for the diagnosis of CVT [131]. Pretest clinical probability scores validated to assist clinician in the diagnosis of CVT are also lacking [131]. The meta-analysis of Dentali et al. (14 studies included; 1134 patients) found a mean sensitivity and specificity for D-dimer of 93.9% and 89.7%, respectively, in patients with suspected CVT. The risk of false-negative D-dimer results included longer duration of symptoms, limited sinus involvement and isolated headache [131]. Future prospective studies are needed to confirm these findings. The recent meta-analysis performed by Alons et al. (eight studies included; 636 patients) showed that D-dimer has a high NPV in low-risk patients with isolated headache for excluding CVT [132]. Low-risk patients were defined by normal neurological examination, normal standard head CT (computed tomography), and absence of risk factors such as puerperium and pregnancy. Sensitivity, specificity, PPV, and NPV for diagnosing CVT were 97.8%, 84.9%, 33.1%, and 99.8%, respectively [132]. Normal D-dimer levels may, therefore, reduce unnecessary neuroimaging investigations.

Acute aortic dissection

Symptoms of AAD typically include back and/or abdominal pain, acute onset of tearing chest, asymmetric blood pressure and widened mediastinum on chest x-ray [133]. However, in many patients, symptoms are nonspecific and a missed diagnosis may be fatal [133]. The diagnosis of AAD currently relies on imaging techniques (i.e. magnetic resonance imaging, echocardiography, contrast-enhanced CT) (Class I; Level of evidence B) [133,134]. The potential usefulness of D-dimer lies in the possibility to rule out AAD in patients with low clinical probability because D-dimer has been reported to be persistently elevated in AAD [4,133,135]. D-dimer testing is also rapid, economical, and more accessible than imaging investigations [136]. The meta-analysis of Cui et al. (five studies included; 743 subjects) found a sensitivity and specificity for the diagnostic of AAD of 94.5% and 69.1%, respectively [133]. Similar performances were obtained in the meta-analysis of Watanabe et al. (12 studies included; 2827 subjects; sensitivity and specificity of 95.2% and 60.4%, respectively) [137]. D-dimer levels can, therefore, be used to rule out AAD in patients with low likelihood of the disease [133,137]. According to the American College of Cardiology Foundation (ACCF) and to the AHA guideline, D-dimer cannot be used to rule out the disease in high-risk patients and they do not recommend D-dimer screening for all patients being evaluated for aortic dissection (AD) [134]. The sensitivity and the failure rate of D-dimer in patients with AD detection risk score (ADD-RS) 0, ≤1, and 1 (high risk of AD) were 100% and 0%, 98.7% and 0.8%, and 97.5% and 4.2%, respectively [138]. The ADD-RS allows standardized assessment of pretest probability for acute aortic syndromes [139]. The combined ADD-RS (0 or ≤1) with negative D-dimer was superior to D-dimer alone in AAD diagnosis [138,140,141]. However, the accuracy of D-dimer was found to be lower (4% failure rate) in patients at high risk (ADD-RS of 1); therefore D-dimer was not acceptable to rule out a diagnosis of AD in such settings [141]. In conclusion, the absence of ADD risk markers combined with a negative D-dimer result argues strongly against the diagnosis of AD [138,141]. Integration of ADD-RS with D-dimer may be considered to standardize the diagnosis of AD [141].

Acute mesenteric ischemia

The interest in D-dimer in the diagnosis of AMI has also been acknowledged in the literature [142–145]. Four etiologies of AMI exist: arterial thrombosis (15–20%), venous thrombosis (5%), arterial embolism (50%), and non-occlusive mesenteric ischemia (20–30%) [146,147]. Failure to recognize AMI is responsible for a high mortality rate [147]. The sensitivity and specificity of spiral CT for the detection of AMI have been determined to be 93.3% and 95.9%, respectively [148]. According to the guideline of the European Society for Trauma and Emergency Surgery (ESTES), published in 2016, D-dimer does not discriminate patients with AMI from non-AMI patients [147]. Moreover, no correlation between serum D-dimer levels and the severity of AMI was observed [143]. However, the guideline cited only the paper of Chiu et al. (prospective study on 67 patients), which was published in 2009 [143,147]. In 2017, a meta-analysis performed by Sun et al. included more studies (12 studies; 1300 patients), eight of which concerned AMI. The four other studies dealt with acute strangulated intestinal obstruction, acute intestinal necrosis and mixed types of acute intestinal ischemia, which could be gathered into the term “acute intestinal ischemia”; the calculated sensitivity and specificity were 94% and 50%, respectively [149]. D-dimer measurement could hence be useful for identifying patients with AMI. Further validations of these results in large multi-center clinical studies are needed.

Exclusion of the diagnosis of venous thromboembolism

Although D-dimer levels are increased in almost all cases of acute VTE, any other condition that increases fibrin production or breakdown may also generate increased D-dimer values [2]. These conditions, listed in Table 4, include infections, cancer, chronic inflammation, aging, pregnancy, recent surgery, and trauma. A large retrospective study of 1647 patients showed that the most common causes of positive D-dimer results were related to infections, followed by VTE, syncope, heart failure, trauma, and cancer [126]. This is of particular concern for hospitalized patients, because D-dimer may be increased for reasons other than VTE and DIC [1]. Brotman et al. found that only 22% of hospitalized patients had normal D-dimer levels [150]. Therefore, D-dimer testing should not be considered specific for VTE [1,2,4,33], and it should be used mainly to exclude VTE because low D-dimer values reflect the lack of significant (ongoing) activation of blood coagulation [1,151].

Table 4. Causes of D-dimer elevation [1,126,127,164].

Acute respiratory distress syndrome	Disseminated intravascular coagulation	Pancreatitis
Advance age	Heart failure	Post transplantation complications
Alzheimer	HELLP syndrome	Pregnancy or puerperium
Aneurysm	Hemolysis (falciform anemia)	Recent surgery
Aortic dissection	Hemorrhage	Renal disease
Arthritis	Hospitalization	Severe chronic urticaria
Atrial fibrillation	Inflammatory bowel disease	Thrombolytic therapy
Burns	Ischemic cardiopathy	Trauma
Cancer	Liver disease	Venous or arterial thrombosis
Chronic inflammation	Localized or systemic Infection
Disability	Neonatal period

Epidemiology of VTE

VTE, including both DVT and PE, is a life-threatening condition, with an incidence ranging between 104 to 183 per 100,000 person-years in Europe [152]. The incidence rate is generally higher in men after 45 years, as well as in women in the childbearing years [152]. The prevalence of VTE has been estimated at 422 cases per 100,000 individuals with isolated DVT (69.9%), PE (23.7%) or both (6.4%) [117]. The frequency of VTE cases appears to be slightly higher in women than in men in the United States [117]. The incidence of VTE is also known to increase sharply with age [153].

Of note, during the last decades, the overall incidence of VTE decreased significantly [154]. The wider introduction of thromboprophylaxis in hospitalized patients may, at least in part, explain this result. Specifically, the incidence of isolated DVT also decreased while the incidence of isolated PE increased [154]. This discordance may be explained by the lack of major improvement in the accuracy of ultrasonographic diagnosis of DVT whereas improved resolution imaging has been used for the diagnosis of PE (i.e. multi-detector CT scan) [154].

Several factors have been associated with an increased risk of VTE, including major surgery, hospitalization for acute medical illness (e.g. heart failure, diabetes or pneumonia), trauma/fracture, thrombophilia, obesity, active cancer, pregnancy or postpartum, extended periods of immobility, and oral contraceptives [152,153,155]. Overall, the survival after PE is predictably worse than after DVT alone. The survival rate for DVT at 3 months and 5 years is 91.9% and 72.7%, respectively, whilst for PE, it is 62.8% and 47.5%, respectively [152,156].

VTE is a major healthcare issue worldwide, with an estimated incidence of 3.0–3.3 cases per 100 hospitalizations per year and a rate of hospital readmission of 5–14% [152,157–159]. Compared to hospitalized controls, adjusted mean predicted costs were 1.5–2.5 times higher for patient with VTE. Cost differences were greatest within the first 3 months [160,161]. The total annual health care cost for VTE in 2007 ranged from $7594 to $16,644 per patient in the US [158]. The VTE-attributable cost has been estimated to be up to $10 billion [162]. It has also been estimated that the number of cases of VTE among adults will at least double by 2050 (0.95 million in 2006 to 1.82 million predicted in 2050) [117].

D-dimer assay performance

According to the US Food and Drug Administration (FDA), the specifications for sensitivity and NPV of a D-dimer assay applied to rule out VTE should be ≥95% (with lower limit of CI ≥90%) and ≥97% (with lower limit of CI ≥95%), respectively [163]. The CLSI provides slightly more stringent recommendations for sensitivity and NPV, entailing ≥97% (with lower limit of CI ≥90%) and ≥98% (with lower limit of CI ≥95%), respectively [163]. Based on these requirements, D-dimer assays can be classified into two main categories: (1) those with high sensitivity (>95%) and low specificity (<40%), and (2) those with moderate sensitivity (80–94%) and high specificity (up to 70%) [164]. Quantitative laboratory assays are more likely to have a higher NPV compared to semi-quantitative and qualitative assays [3]. The meta-analysis published by Di Nisio et al., which included 113 studies, showed that ELFAs (e.g. Vidas^®, Stratus DS^® (Siemens, USA), AxSYM^®), microplate ELISAs (e.g. Asserachrom^®, Enzygnost^®), and latex quantitative immunoassays (second generation of latex-based assays [2]) (e.g. Tina-quant^®, STA-Liatest^®) had higher sensitivity (median, ≥95%) but lower specificity (median, ±50%) [165]. They also found that whole blood agglutination, latex semi-quantitative or qualitative assays had low sensitivity (median, 84%) and high specificity (median, 76%) [165]. These results were also confirmed in the meta-analysis of Stein et al., which included 78 studies [151]. Similarly, Roy et al. found that ELFA assays exhibit the highest exclusion value (the lowest negative likelihood ratio, see below) and whole blood agglutination assays, the lowest [166]. The specificity of whole blood POC D-dimer assays is overall higher than central laboratory D-dimer assays (71% and 69% for DVT and PE, respectively), meaning that the number of patients in whom VTE could be excluded increases [91]. However, the sensitivity is lower (83% and 87% for DVT and PE, respectively) as compared to central laboratory D-dimer assays [165]. More recent studies also found optimal diagnostic performance for current laboratory D-dimer immunoassays (e.g. STA-Liatest^®, HemosIL HS 500^®, HemosIL AcuStar^®, Innovance^® (Siemens, USA), Tina-Quant^®) [38,48,84,112,114,167,168]. Not all POC D-dimer assays are appropriate to be used to exclude VTE [107]. Some POC assays have been approved only as an aid to diagnosis and still require additional tests to rule-out VTE reliably [107]. It is also important to know if the FDA has cleared the assay for exclusion of VTE [107]. Up to now, few POC D-dimer assays meet the performance characteristics recommended by CLSI [95,107]. Quantitative POC assays have been found to be more reliable than qualitative POC assays, especially in patients with low pretest probability of VTE [91,107,169]. Taken together, GPs should weigh the benefit of rapid POC D-dimer assay results over the drawback of overall lower performance [91].

Notably, the NPV is directly influenced by the prevalence of a disease [38]. For example, the prevalence of DVT ranges between 10–50% [38]. An accurate interpretation of D-dimer immunoassay performance is hence needed. The negative likelihood ratio may be the most accurate criterion to assess the diagnostic performance of a D-dimer test [170,171]. The negative likelihood ratio represents the extent of change in the odds or the probability of the disease after a test result [172]. The number needed to test (NNT) index, or the number of patients in whom D-dimer must be measured to rule out one VTE, may also be used to compare the diagnostic yield of commercial D-dimer immunoassays [168,173,174].

Clinical prediction rules (CPR)

It has been demonstrated extensively that assessment by means of a clinical score should be combined with D-dimer testing to improve the diagnostic specificity of the D-dimer result and enhance the diagnostic accuracy [2,4,33,120,153,175]. The aim of clinical probability assessment is to deal with the prevalence dependence for accurate interpretation of diagnostic tests of D-dimer assays by identifying subgroups of patients according to different levels of VTE prevalence: low (<10%), moderate (around 30%), high (>50%), or “unlikely” (±10%) and “likely” (±35%) [155]. There are additional aspects that support the use of D-dimer in combination with CPR, such as the fact that some patients with thrombosis may present a false-negative D-dimer concentration due to a hypofibrinolytic state, small thrombi (i.e. distal DVT or isolated subsegmental PE), anticoagulant therapy or D-dimer testing performed too early or too late after the thrombosis [1,2,4,33,48].

CPR have been derived from clinical studies and include a number of signs, symptoms, and risk factors of VTE [2]. The use of two-level scores is typically preferred, because it is straightforward and because three-level scores depend on the sensitivity of the D-dimer assay (which is rarely known by the clinicians) [153,175]. The Well’s and Geneva scores (modified or not) are the two CPRs mainly used for PE, whilst the Well’s score is especially used for DVT [155,169,174,176]. These scores for PE are shown in Table 5. The diagnostic performances of these CPR were found to be almost comparable [119,177–179]. Other CPR have also been validated and may be used routinely (e.g. Empiric, Charlotte or Simplified Geneva) [1,177,180]. A good CPR should not contain subjective variables, and it should be accurate, reproducible, easy to remember, and offer a standardized approach compared to clinical assessment (especially for inexperienced physicians) [1,155]. However, pretest probability calculation may still be subjective [181,182]. These algorithms are considered clinically validated when the number of thromboembolic event after 3 months is as high as 1–2% compared to a gold standard test (pulmonary angiography for PE and venography for DVT) [183].

Table 5. Clinical prediction rules for PE (adapted from Righini et al. [174]).

Wells score		Geneva score		Revised Geneva score
Items	Score	Items	Score	Items	Score
Previous PE or DVT	1.5	Previous PE or DVT	2	Age >65	1
Heart Rate >100	1.5	Heart Rate >100	1	Previous PE or DVT	3
Recent surgery or immobilization	1.5	Recent Surgery	3	Surgery or fracture within 1 month	2
Clinical signs of DVT	3	Age		Active malignancy	2
Alternative diagnosis less likely than PE	3	60–79	1	Unilateral lower limb pain	3
Hemoptysis	1	≥80	2	Hemoptysis	2
Cancer	1	Arterial blood gases		Heart rate
		CO2 (kPa)		75–94	3
		<4.8	2	≥95	5
		4.8–5.19	1	Pain on lower limb deep vein palpation and unilateral edema	4
		O2 (kPa)
		<6.5	4
		6.5–7.99	3
		8–9.49	2
		9.5–10.99	1
		Chest X-ray
		Atelectasis	1
		Elevated hemidiaphragm	1
Clinical probability	Clinical probability	Clinical probability
Low	<2	Low	0–4	Low	0–3
Intermediate	2–6	Intermediate	5–8	Intermediate	4–10
High	>6	High	≥9	High	≥11
Dichotomized
PE unlikely	≤4
PE likely	>4

Different algorithms combining CPR and D-dimer measurement have been proposed and endorsed by international guidelines [33,155,164,184]. A “low” or “unlikely” probability is generally followed by D-dimer testing. VTE can be ruled out when D-dimer values are below the method-specific cutoff. In these conditions, the use of time-consuming, expensive and less safe (i.e. due to radiation exposure and/or injection of contrast products) tests would be unnecessary [2]. However, when the clinical probability is “high or likely” or “moderate/high”, D-dimer testing is not useful and radiologic imaging should be used [2,185]. A positive D-dimer test result may also trigger imaging tests [2]. For DVT, this would entail ultrasonography [183] or Doppler flow studies [183]. For PE, the most used techniques include CT pulmonary angiogram (CTPA) [183] or contrast-enhanced or unenhanced magnetic resonance imaging, especially when CTPA is unadvisable [183]. Some studies based on specific algorithms concluded that a negative D-dimer result in patients with a “moderate” score could also safely rule out a VTE episode, thus enhancing the usefulness of D-dimer testing and decreasing the use of imaging testing [155,180]. Several other studies, including a meta-analysis of 1660 patients, showed that a normal D-dimer concentration along with a low CPR score had a NPV of up to 95% [3,116,119,186].

However, in current practice, such clinical algorithms are not always followed [181,182]. In a recent survey that included 487 clinicians from six countries, only 70.3% used pretest probability scores, and 10% excluded or confirmed DVT based only on D-dimer test results. Moreover, a significant number of clinicians still order D-dimer testing in patients with a high probability of VTE, whilst others order imaging testing in cases of low pretest probability. This waste of resources may lead to a negative impact on health care resources [181,182]. Moreover, in a large study that included 117 centers and 1529 patients with suspected PE, the diagnostic management was inappropriate in 662 patients (43%): in 36 of 429 (8%) patients with confirmed PE and in 626 of 110 (57%) patients in whom PE was ruled out [187]. One of the independent risk factors for inappropriate management was the lack of written algorithms that included a clinical probability score. Major efforts are hence needed to harmonize the use of the different CPR algorithms.

A recent prospective study based on 808 patients found that a low D-dimer level (<750 μg/L FEU) measured with the MDA D-dimer assay (quantitative latex agglutination assay, bioMérieux) could be used alone for ruling out PE, without the need of a CPR score [188]. The theoretical advantage of limiting the use of CPR is a decrease in the cost attributable to imaging techniques (e.g. CTPA and ventilation-perfusion lung scanning) and prevention of radiation exposure [188]. However, the MDA assay is no longer being marketed. Moreover, the number of patients with a high pretest probability was quite low in that study [188]. On the other hand, D-dimer testing has limited clinical value in patients with a high CPR score in whom the prevalence of VTE is high (i.e. low NPV) [164]. Additional studies with other D-dimer assays are needed to confirm these findings. Nevertheless, the use of a D-dimer assay as a stand-alone test has not been widely endorsed by most guidelines and consensus documents [1,2,4,33,38,127,164,175,183,189]. This is understandable because the high percentage of deaths in the untreated patients is the major risk of underdiagnosing VTE [183,185].

Age-adjusted cutoff

As earlier discussed, D-dimer values tend to increase with age, thus leading to a greater proportion of older patients (≈60%) with D-dimer values higher than the conventional cutoff of 500 μg/L FEU [1,4,33,105,117,118,190]. The specificity of D-dimer for VTE in elderly patients (0–18%, ≥80 years) is lower compared to younger patients (49–67%, <50 years). Elderly patients are therefore less likely to have negative D-dimer results when VTE is absent [12,118], and a high number of these patients with a low clinical score may undergo unnecessary imaging testing because of high D-dimer values [118]. This is because consolidated evidence shows that VTE prevalence increases with age [12,119,152,174].

Age-adjusted cutoffs have subsequently been validated and recommended because their application increases diagnostic specificity without significantly impairing the NPV [33,119,173,191–193]. The calculation of the age-adjusted cutoff is based on the following equation: “age-adjusted cutoff, μg/L FEU = age in years ×10 [119]”. Thus, the age-adjusted cutoff of a 70 year-old patient is 700 μg/L FEU (rather than 500 μg/L FEU). The multiplication factor of “10” has been slightly rounded to facilitate practicability and clinical usefulness [119].

The ADJUST-PE study, which included over 3000 patients, prospectively demonstrated that the combination of age-adjusted cutoff and a clinical score (Wells or Revised Geneva score) was associated with a higher proportion of patients in whom PE could be safely ruled out without additional CTPA [120,194]. Many other studies have demonstrated that the number of patients in whom VTE could be ruled out increased remarkably after applying age-adjusted cutoffs, and this approach has also been found to be cost-effective because a lower number of patients with low clinical score will need to undergo expensive imaging tests [120,168,174,183,194].

A recent meta-analysis evaluated the clinical effectiveness of age-adjusted cutoffs in 12497 older patients and concluded that the sensitivity was >97% in all age categories [118]. The gain in specificity (up to 23% better specificity compared to a fixed cutoff) was especially favorable for patients aged 75 years or older [118,120]. This strategy has been validated for both PE and DVT with many commercial D-dimer immunoassays [38,119,120,168,191]. Specifically, Oude Elferink et al. found that the NPV of 9 D-dimer immunoassays was always >97% [38]. Another study found a NPV >99% in using five different D-dimer immunoassays [168].

More recently, Farm et al. reported that 4 D-dimer assays display sensitivity >95% and NPV >97%, whilst only one method (MediRox^®, Sweden) displayed slightly lower sensitivity (94%) [191]. Although age-adjusted cutoffs seem straightforward in most cases, caution should be used with non- or insufficiently validated D-dimer immunoassays [106]. More recently, the use of age-adjusted cutoffs for ruling out PE has also been endorsed by the European Society of Cardiology (ESC) [184].

Interestingly, a recent study demonstrated that age-adjusted cutoffs may also be useful in younger patients. Prochaska et al. [192] performed a clinical study that included 500 patients with suspected DVT, and reported that a D-dimer cutoff of 500 µg/L FEU had lower sensitivity in patients <60 years (76.1%) compared to older patients (92.9%) A lower sensitivity was found especially in women, and in patients with unprovoked DVT, with low thrombotic burden, and with distal DVT [192]. However, the use of a lower cutoff in younger patients (250 µg/L FEU) was effective in increasing the sensitivity of this test remarkably, up to 92.4%. Nevertheless, further studies will be needed to confirm the results of this preliminary observational study.

Clinical probability-adjusted cutoffs

The use of clinical probability-adjusted cutoffs has also been proposed [2,195–197]. Linkins et al. used a low (200 µg/L FEU), an intermediate or “conventional” (500 µg/L FEU) and a high (2000 µg/L FEU) D-dimer cutoff (MDA D-dimer assay) for high, moderate, and low pretest probability patients, respectively [196]. Overall, 571 patients were studied retrospectively. Although the authors failed to find different sensitivities and NPVs across the groups for diagnosing VTE, the use of clinical probability-adjusted cutoffs increased the D-dimer test specificity compared to the conventional approach (i.e. 44.7% versus 60.4%) [196]. Another study published by Kabrhel et al. in 7940 patients with suspected PE and based on six different D-dimer assays reported similar data [195]. The use of higher (conventional ×2), conventional (identical), and lower (conventional/2) cutoffs for low, intermediate, and high pretest probability patients displayed a higher specificity (75% versus 58%) and a similar NPV (99.1% versus 99.5%) [195]. Kline et al. also suggested using a higher D-dimer cutoff (1000 µg/L instead of 500 µg/L) in younger patients (i.e. aged <50 years) with a low clinical score for PE (clinical probability-adjusted strategy). The authors found a significant decrease in the need for performing unnecessary pulmonary angiograms but also observed a slight decrease in diagnosis of isolated subsegmental PE [198]. Likewise, Linkins et al. reported that modifying D-dimer cutoffs according to the clinical pretest probability of DVT (<500 µg/L and <1000 µg/L for moderate and low pretest probability, respectively) may be associated with a reduced need to perform ultrasonography, without impairing the NPV of the test [185].

Recently, van der Hulle et al. prospectively validated a diagnostic algorithm incorporating the YEARS clinical decision rule (CDR) (clinical signs of DVT, hemoptysis, PE as the most likely diagnosis) and various D-dimer cutoffs in 3465 patients with suspected PE. Among the 2946 (85%) patients in whom PE was ruled out (no YEARS CDR and D-dimer <1000 µg/L or at least one YEARS CDR and D-dimer <500 µg/L), the rate of VTE during a 3-month follow-up period was only 0.61% (95% CI: 0.36%–0.96%) [180]. Furthermore, a 14% decrease in CTPA was observed [180]. Based on this latter study, van der Pol et al. found that the age-adjusted cutoff added no value to the YEARS algorithm [199].

Takach Lapner et al. recently compared both the age-adjusted and the clinical probability-adjusted strategies [197]. For that purpose, 1649 patients with low or moderate clinical probability after assessment for a first VTE event were studied retrospectively. The authors found that the NPV was similar in both groups (99.7% versus 99.6%), but also observed a significantly higher rate of negative D-dimer test results (STA-Liatest^®) in the clinical probability-adjusted strategy compared to the age-adjusted strategy (56.1% versus 50.9%) [197]. According to these recent findings, it can be concluded that the clinical probability-adjusted cutoffs may be more efficient than the age-adjusted cutoffs for limiting the number of imaging tests. However, this conclusion needs to be supported by further prospective studies using different D-dimer immunoassays.

VTE diagnosis in specific populations

In patients with recent surgery, D-dimers are usually increased and the diagnosis of VTE is hence challenging. No CPR is validated and only the global clinical judgment of the clinician can be used to assess CPR [200]. However, clinical judgment is criticized for its lack of standardization and the difficulty in teaching. Therefore, clinicians perform imaging whenever VTE is suspected. Nevertheless, Penaloza et al. developed a score composed of 9 variables to assess the usefulness of D-dimer measurement, using data from 4537 patients that had high sensitive D-dimer testing (Vidas^® for 2079 patients, STA-Liatest^® for 1783 patients, and MDA^® for 675 patients) [201]. These 9 variables were independently associated with a risk of false positive D-dimer value (sex female: +1, age 65–84 years: +4, age ≥85 years: +8, heart rate ≥95/min: +1, pulse oxygen saturation <95%: +2, temperature ≥38.5 °C: +3, personal history of VTE: +1, surgery under general anesthesia within 4 weeks: +2, active malignancy: +3, pregnancy or postpartum within 4 weeks: +4). In patients with non-high CPR and a relevance score ≤8, at least 10% had a quantitative D-dimer assay below 500 µg/L, meaning that, in patients with recent surgery, D-dimer measurement may still be useful, at least if they have no other risk factors for false positive results. Indeed, a negative D-dimer result keeps its sensitivity and NPV [200].

Similarly, D-dimer levels increased physiologically in pregnancy and in the postpartum period. In a study that included 1343 pregnant women with D-dimer measurements using a turbidimetric method (STA-Liatest^®), the rate of D-dimer test results below the usual cutoff (500 µg/L) in healthy pregnant women was 85%, 29% and 4.1% during the first, the second and the third trimesters, respectively [202]. Similar results were observed with other assays (MDA^® and HemosIL HS assay^®) [203,204]. In the post-partum period, D-dimer returns to normal levels around the 6th week. In case of suspicion of PE, because imaging tests may expose the mother and the fetus to radiation, the ability to rule-out PE with non-radiologic tests is crucial [107]. To improve the usefulness of D-dimer results, specific cutoffs have been proposed, but not yet validated, in a large prospective study [124]. However, once again, a negative D-dimer result keeps its sensitivity and NPV [107]. Two large studies are currently ongoing on this topic and may give important information on the best algorithm to exclude VTE in pregnant women [205]. Currently and in order to avoid imaging tests, D-dimer measurement is still recommended in pregnant women with suspicion of VTE and a non-high CPR, at least in the suspicion of PE [206].

In patients known to have cancer, the relevance of D-dimer in the diagnosis of VTE is decreased [107]. The prevalence of VTE is increased (up to 20% of cancer patients develop VTE) and the NPV is therefore reduced [107,207]. A large meta-analysis of 10002 patients showed that the prevalence of both a low Wells score and a negative D-dimer value among patients with cancer was only 9% [208]. It has also been shown that 88–94% of patients with malignancy require additional tests to rule out VTE [209]. However, the D-dimer measurement as included in a score may be useful to identify cancer patients at low or high risk of VTE [210].

The exclusion of VTE in patients with renal disease is also challenging, given that D-dimer fragments are eliminated mainly by renal clearance and reticuloendothelial system catabolism [11]. Accordingly, Robert-Ebadi et al. showed that D-dimer concentrations increased according to renal status, and the proportion of negative D-dimer results decreased from 46% to 11% with normal (≥90 mL/min) and moderate renal failure (30–59 mL/min), respectively [211]. In addition, Lindner et al. showed that 100% of patients with low eGFR (<30 mL/min; n = 29) had a positive D-dimer value [212]. The specificity will therefore proportionally decrease according to the renal failure status [212,213]. The possibility of using renal function-adjusted cutoffs should be verified in dedicated studies [212].

Prediction of recurrence of VTE

The risk of recurrence during a 1-year follow up after a first VTE episode is higher in men (9.5%) than in women (5.3%), and tends also to increase over time (9.1% and 19.7%, for men and women after 3 years, respectively) [214]. Interestingly, a high D-dimer value has been associated with increased risk of recurrent VTE [127]. It has hence been suggested that the duration of anticoagulant therapy in VTE patients may be combined with results of D-dimer testing [1,215]. In the PROLONG study, the D-dimer value was increased 1 month after cessation of vitamin K antagonist (VKA) in 608 VTE anticoagulated patients who were treated for at least 3 months [215]. Patients with normal D-dimer values measured with the Clearview Simplify^® D-dimer assay (whole blood qualitative assay) did not resume anti-coagulation, whilst those with abnormal D-dimer values were randomly assigned to receive or not receive a new anticoagulant therapy. An abnormal D-dimer concentration was observed in ≈37% of patients. Among these, a new VTE episode occurred in 15% of patients who stopped VKA, whilst VTE events occurred only in 2.9% of those receiving VKA therapy (adjusted hazard ratio [HR] 4.26; 95% CI 1.23–14.6; p = .02). Legnani et al. reanalyzed the data of the PROLONG study to assess D-dimer cutoff values for several quantitative assays (Vidas^®, Innovance^®, HemosIL^®, and STA-Liatest^®) [216]. Based on this post-hoc analysis, they determine age-specific cutoffs and identified a higher threshold in older patients (i.e. aged 70 years or older) [216]. The use of a strategy based on method-specific cutoff values, adjusted for age and gender, is therefore advisable for more accurate prediction of VTE recurrence [217,218]. The use of quantitative methods rather than qualitative tests results (i.e., positive or negative) may be more suitable to achieve this goal [216–218]. The meta-analysis of Douketis et al., which included 1818 patients from seven prospective studies, found that the D-dimer value was a significant predictor of VTE (HR 2.59; 95% CI 1.90–3.52; p < .001) [214]. Overall, a D-dimer value higher than the diagnostic cutoff after 3-months of anticoagulant therapy in patients with a first unprovoked VTE event was associated with double the risk of recurrence compared to patients with lower D-dimer values [2,219]. The risk of VTE recurrence was similar among young and elderly patients. More recently, Nagler et al. showed that increased D-dimer values (Vidas^® and Innovance^®) observed after discontinuing anticoagulant therapy after 1 month (HR 3.3; 95% CI 1.8–6.1), male sex (HR 2.8; 95% CI 1.5–5.1), and use of oral contraceptives (HR 0.1; 95% CI 0.0–0.9), were significant predictors of recurrent DVT in 479 patients diagnosed with proximal DVT [220]. These authors also found that high factor VIII values (HR 2.2; 95% CI 1.2–4.0) could be predictors of recurrent DVT [220]. The distinction between provoked and unprovoked VTE is also important. Patients with provoked VTE (after surgery, lower limb trauma, and orthopedic immobilization or hospitalization for an acute medical illness) have a very low risk of recurrence as compared to patients with unprovoked VTE who have a higher risk of recurrence at two years (±20%) [221].

Several models have been developed to predict the risk of recurrence of unprovoked VTE. In the real-life study of Tosetto et al., performed with different D-dimer assays in nine centers, an increased D-dimer value after stopping anticoagulation (cutoff 500 µg/L with a quantitative assay or “positive” with a qualitative assay), a younger age (<50 years), male sex, and VTE not associated with hormonal therapy in women were identified as the best predictors of VTE recurrence (area under the curve, 0.71) [222]. The diagnostic performance of a model including all these variables had a higher diagnostic efficiency compared to D-dimer testing alone (area under the curve 0.61; p < .001). This score may hence be useful to assess whether the anticoagulant treatment should be discontinued or resumed after the usual 3-month period. The DASH score has been retrospectively validated in a recent independent cohort study, but the score appeared to perform better in younger subjects (<65 years old) [223]. To our knowledge, the DASH score has not been validated in an interventional prospective study. Additional algorithms, such as the Vienna and HERDOO2 prediction models, have also been developed [224]. The Vienna prediction model included D-dimer (Asserachrom assay^®), sex, and site of index event [224,225]. It has been validated in external populations but, again, not in an interventional study [226]. Conversely, the HERDOO2 model has been recently validated in an interventional study that included 3155 patients from several countries [227]. This model included D-dimer (measured with Vidas^® assay; cutoff, 250 µg/L), age, body mass index (BMI), and post-thrombotic signs [224,228]. In line with these findings, the Scientific and Standardization Committee of the International Society on Thrombosis and Hemostasis recommends that D-dimer testing should be performed in all patients with clinical suspicion of recurrent VTE [229]. However, caution has to be used when translating these findings to other D-dimer immunoassays. Moreover, controversy remains about the most appropriate timing of D-dimer monitoring during or after discontinuing anticoagulation. Finally, the accurate definition of “positive” or “negative” D-dimer (i.e. precise cutoff) is still an unmet requirement [2,214].

Prediction of risk of VTE in hospitalized patients

Recently, a baseline D-dimer level measured in 7441 hospitalized patients was found to be independently associated with symptomatic VTE during a period of ∼3 months (adjusted HR 2.22; 95% CI 1.38–1.58; p < .001) [230,231]. Patients hospitalized for acute illness (ischemic stroke, heart failure, respiratory failure, rheumatic disorders, infection) have an increased risk of developing VTE [231–234]. A risk assessment for VTE is hence needed before initiating thromboprophylaxis [231]. The International Medical Prevention Registry on Venous Thromboembolism (IMPROVE) assessment tool has been developed with the aim of risk stratification in hospitalized, medically ill patients. The score encompasses many clinical variables such as previous VTE, known thrombophilia, current lower-limb paralysis, current cancer, immobilization ≥7 days, intensive care or coronary care units, and age >60 years [231]. Because D-dimer is an independent predictor of symptomatic VTE, Gibson et al. combined D-dimer testing with the IMPROVE assessment tool [231]. A D-dimer level ≥2× the upper reference limit (URL) measured with STA-Liatest platform was assigned two more points in the new scoring system (IMPROVEDD). The population of the APEX study was used to validate this approach. The combination of clinical variables already included in the IMPROVE score with D-dimer test results was effective in substantially improving risk discrimination and patients risk reclassification at both 42 and 77 days.

Diagnosis and monitoring of disseminated intravascular coagulation

DIC is a life-threatening condition characterized by persistent activation of the hemostatic system with intravascular thrombin generation, fibrin formation, and increased fibrinolysis [3]. Patients with DIC may present with bleeding, thrombosis, or both [164]. Early recognition of DIC is paramount to initiate the appropriate treatment, which generally entails eliminating the underlying condition (e.g. sepsis, malignancy, trauma or burns, obstetrical diseases, toxins, drugs, immunological disorders, and other inflammatory diseases) [1]. The International Society of Thrombosis and Haemostasis (ISTH) [235] and the Japanese Ministry of Health and Welfare [236] have proposed two different scoring system for diagnosis and management of DIC. Comparison of these scoring systems yielded similar diagnostic effectiveness for DIC [1,237]. The laboratory parameters included in these scoring systems are platelet count, fibrinogen, prothrombin time, and FDP [235,236,238]. Although D-dimer is the most common FDP measured in clinical laboratories, some authors suggested that soluble fibrin may be more specific to detect intravascular clot generation. D-dimer is considered positive when its value is ≥2 times the URL [1,4]. As for VTE, there is now consolidated evidence that a “normal” D-dimer value excludes a diagnosis of DIC [127]. However, a “positive” D-dimer value cannot be considered specific for DIC, because D-dimer may be increased in a vast array of clinical conditions (Table 4). Repeated D-dimer analysis (i.e. 3 times per day) is also important to monitor DIC [236,238,239]. The combination of D-dimer and fibrin monomer allowed the identification of patients with septic shock with a worse survival [240].

Conclusions

Today, D-dimer has become one of the most commonly requested coagulation tests. D-dimer, considered a biomarker of activation of coagulation and fibrinolysis, is mainly employed for the exclusion of VTE. A negative D-dimer value along with a low clinical probability can safety exclude VTE in suspected individuals. Currently, it is recommended to employ D-dimer assays with a very high sensitivity and NPV to safety exclude VTE (≥95% and ≥97%, respectively). However, a positive D-dimer value should always trigger imaging tests to confirm VTE, given that a wide array of diseases and conditions are characterized by an increase in D-dimer (i.e. aging, inflammation, cancer, renal failure). More recently, the use of age-adjusted cutoffs to safely rule out VTE has been proposed to substantially increase the PPV without significantly impairing the NPV, so ultimately improving the clinical usefulness of D-dimer measurement in elderly patients with low clinical probability. Even though these adjusted cutoffs have been endorsed in several guidelines, many laboratories are still not using them. Major efforts are therefore needed for implementation of these recommendations. VTE exclusion in specific populations (i.e. pregnancy, heart failure, cancer, perioperative settings) is also challenging. Utilization of POC D-dimer assays as a rapid screening tool is especially attractive for GPs because central laboratory quantitative D-dimer assays are not always readily available. However, not all POC D-dimer assays are appropriate to be used to exclude VTE, and only POC assays that have been validated in clinical trials and cleared by the FDA should be used.

D-dimer measurement is also useful to predict the risk of VTE recurrence and the risk of VTE in hospitalized patients and to diagnose and monitor DIC. Other emerging indications (i.e. prognostication of peripheral artery disease, screening of intracardiac thrombus) need further investigations.

Pre-analytical requirements are paramount in the hemostasis laboratory. Nevertheless, the literature appears to be quite reassuring regarding several preanalytical variables that affect D-dimer results (needle bore size, butterfly devices, use of PTS, in vitro hemolysis, stability). In some specific settings (i.e. emergency, older patients, delay before centrifugation), D-dimer results may be released to the clinician without impairing patient management.

A major drawback of D-dimer assays is the high variability observed between immunoassays. This variability is explained by the fact that D-dimers comprise a broad mixture of degradation products of cross-linked fibrin, by the utilization of different monoclonal antibodies, by the lack of international certified internal controls or calibrators, and by the use of different units and clinical cutoffs. Further studies are needed to achieve harmonization of D-dimer measurements. Ongoing discussion among manufacturers, scientists and clinicians is essential for achieving this goal.

Abbreviations
AAD	=	acute aortic dissection
ACCF	=	American College of Cardiology Foundation
AD	=	aortic dissection
AHA	=	American Heart Association
AMI	=	acute mesenteric ischemia
APTT	=	activated partial thromboplastin time
ASA	=	American Stroke Association
BMI	=	body mass index
CAP	=	College of American Pathologists
CDR	=	clinical decision rule
CLIA	=	chemiluminescent enzyme immunometric assay
CLSI	=	Clinical Laboratory Standards Institute
CPR	=	clinical prediction rules
CT	=	computed tomography
CTPA	=	CT pulmonary angiogram
CV	=	coefficient of variation
CVT	=	cerebral venous thrombosis
DD/E	=	D-dimer/fragment E complex
DDU	=	D-dimer units
DIC	=	disseminated intravascular coagulation
DiPEP	=	Diagnosis of PE in Pregnancy
DVT	=	deep venous thrombosis
eGFR	=	estimated glomerular filtration rate
ELFA	=	enzyme-linked fluorescence immunoassay
ELISA	=	enzyme-linked immunosorbent assays
ESC	=	European Society of Cardiology
ESTES	=	European Society for trauma and Emergency Surgery
F/T	=	freezing/thawing
F1 + 2	=	prothrombin fragments 1 + 2
FACT	=	fibrin assay comparison trial
FDA	=	Food and Drug Administration
FDP	=	fibrin degradation products
FEU	=	fibrinogen equivalent units
FV	=	factor V
FVIII	=	factor VIII
G	=	Gauge
GP	=	general practitioner
HMWF	=	high molecular weight fibrinogen
HR	=	hazard ratio
HV	=	healthy volunteers
ISO	=	International Organization for Standardization
ISTH	=	International Society on Thrombosis and Haemostasis
IV	=	intravenous
LMWF	=	low molecular weight fibrinogen
NNT	=	number needed to test
NPV	=	negative predictive value
PE	=	pulmonary embolism
PET	=	polyethylene terephthalate
POC	=	point of care
PPV	=	positive predictive value
PT	=	prothrombin time
PTS	=	pneumatic tube system
RS	=	risk score
RT	=	room temperature
TAT	=	turnaround time
t-PA	=	tissue plasminogen activator
TT	=	thrombin time
URL	=	upper reference limit
VKA	=	vitamin K antagonist
VTE	=	venous thromboembolism
WHO	=	World Health Organization

Acknowledgments

The authors would like to thank Mr. Nicolas Bailly for the editing of the figures and tables.

Disclosure statement

No potential conflict of interest was reported by the authors.