THE ASSOCIATION BETWEEN D-DIMER AND DEEP VEIN THROMBOSIS LOCATION AND ITS ABILITY TO PREDICT PULMONARY EMBOLISM, a retrospective pilot study

Background In lower limb deep vein thrombosis; it is important to identify proximal from distal deep vein thrombosis as it carries the highest risk of pulmonary embolism. It is known that D-dimer has a great role in deep vein thrombosis diagnosis. Yet, the use of D-dimer to predict the location of deep vein thrombosis and the risk of pulmonary embolism in deep vein thrombosis patients has not been investigated before. Objective To address the correlation between D-dimer and the location of deep vein thrombosis and to study the efficacy of D-dimer to predict risk of PE in patients with proximal or extensive deep vein thrombosis. Method We included 110 consecutive patients who were hospitalized with the diagnosis of deep vein thrombosis, with or without a concomitant diagnosis of PE, and with D-dimer measured at initial presentation. We categorized the location of deep vein thrombosis as: distal, proximal, and extensive. In the analysis, patients were grouped into high-risk (patients with Proximal or Extensive deep vein thrombosis and pulmonary embolism) and low risk group (patients without pulmonary embolism).

2 Abstract Background In lower limb deep vein thrombosis; it is important to identify proximal from distal deep vein thrombosis as it carries the highest risk of pulmonary embolism. It is known that Ddimer has a great role in deep vein thrombosis diagnosis. Yet, the use of D-dimer to predict the location of deep vein thrombosis and the risk of pulmonary embolism in deep vein thrombosis patients has not been investigated before.

Objective
To address the correlation between D-dimer and the location of deep vein thrombosis and to study the efficacy of D-dimer to predict risk of PE in patients with proximal or extensive deep vein thrombosis.

Method
We included 110 consecutive patients who were hospitalized with the diagnosis of deep vein thrombosis, with or without a concomitant diagnosis of PE, and with D-dimer measured at initial presentation. We categorized the location of deep vein thrombosis as: distal, proximal, and extensive. In the analysis, patients were grouped into high-risk (patients with Proximal or Extensive deep vein thrombosis and pulmonary embolism) and low risk group (patients without pulmonary embolism).

Results
There was no significant association between D-dimer level and the location of deep vein thrombosis (p=0.519). However, D-dimer level was greater among patients with pulmonary embolism (9.6mg/L) than among patients without pulmonary embolism (7.4mg/L), (p=0.027). D-dimer was a significant predictor of pulmonary embolism as patients with proximal or extensive deep vein thrombosis had 8-folds increased risk of pulmonary embolism than patients with D-dimer less than 4.75mg/L (OR=7.9, p=0.013).
3 Conclusion Though D-dimer was not significantly associated with the location of deep vein thrombosis, it was a significant predictor of pulmonary embolism in patients hospitalized with proximal or extensive deep vein thrombosis.

Background
The annual overall incidence of venous thromboembolism (VTE) is estimated to be approximately 100 per 100.000 persons (half of them with deep vein thrombosis) [1]. In Saudi Arabia, it is estimated that 25,000 people develop VTE annually [2]. Lower limb deep vein thrombosis (DVT) is the main source of pulmonary embolism (PE). Patients with proximal (i.e. above the knee) carry the highest risk of PE (approximately 66%) compared to patients with distal DVT (i.e. below the knee) (33%) [3]. D-dimer results from fibrin degradation and it is the laboratory indicator of thrombus formation and lysis anywhere in the body [4]. it is an inexpensive and feasible test that has a major role in the diagnosis of Venous Thromboembolism. A negative D-dimer is sufficient to rule out a diagnosis of DVT in patients with low or intermediate pretest probability on Well's score [5].D-dimer test has always been interpreted as positive (usually >0.5mg/L] which is highly suggestive of DVT or negative (<0.5 mg/L) which makes DVT is a less likely diagnosis. Little concern was made regarding how positive the test is (i.e. the level of D-dimer). Several Studies have investigated D-dimer beyond its diagnostic value in patients with PE; in attempt to use D-dimer as a biomarker or predictor of the severity of PE [6.7.8]. These studies have shown an association between higher level of D-dimer and the severity of PE, with a higher level of D-dimer indicating a more extensive PE [6,9]. Furthermore, studies proved the prognostic value of D-dimer in predicting mortality in patients with pulmonary embolism [7,8] when high level of D-dimer was associated with increased mortality. One study showed that the higher the D-dimer level the higher the overall mortality in patients 4 with pulmonary embolism. According to that study for every 1000ng/ml increase in Ddimer the overall mortality increased by 10% among patients with PE [7].
The utility of D-dimer in patients with DVT is still confined to its diagnostic value (i.e to rule out DVT diagnosis) giving its high sensitivity and negative predictive value [10]. Until now the use of D-dimer to predict the location of DVT or the risk of developing PE in DVT patients has not been investigated. To our knowledge; there is no study that examined the use of D-dimer as a predictor of the most serious complication associated with lower limb DVT; i.e. pulmonary embolism. We conducted this study to shine a light on the correlation between D-dimer level and the location of DVT and the ability of D-dimer of to predict pulmonary embolism among patients who are diagnosed with proximal or extensive DVT.

Objectives:
To assess weather D-dimer level is associated with the location of lower limb DVT (distal, Proximal, Extensive) and the incidence of symptomatic PE in these patients. In addition, to determine the efficacy of D-dimer to risk stratify patients with lower limb DVT to high risk group (patients who developed PE on top of proximal or extensive DVT) and low risk group (those who are diagnosed proximal or extensive DVT but did not develop PE), and to find a D-dimer level above which patients can be considered high risk of developing PE.

Materials & Methods
This is a retrospective study which included 110 consecutive patients, between January 2015 and December 2017, who were admitted to the hospital with a diagnosis of DVT confirmed by ultrasonography, with or without a concomitant diagnosis of PE at initial presentation, and D-dimer level measured at initial presentation. Inpatients who developed DVT during hospitalization were excluded.
Data which consisted of (patient's age and gender, symptoms of DVT(as unilateral leg swelling, pain, hotness), duration of symptoms, length of hospital stay, location of DVT (distal, proximal, or extensive), location of proximal DVT (popliteal, femoral, or iIiac), presence of malignancy, pregnancy or recent surgery which are considered as confounder that may interfere with D-dimer level, D-dimer level and finally the occurrence of pulmonary embolism either at initial presentation or later during hospitalization) was collected from patients' electronic medical records and medical files into a designed data sheet. The extracted data was de-identified (without names, National ID number, contact number, or geographical data) to assure patients confidentiality.
D-dimer was measured using the rapid quantitative turbidimetric D-dimer assay which has a negative predictive value of 97.7% [11] and was reported in mg/L. The diagnosis of DVT was confirmed by whole leg compression ultrasonography. "Distal DVT" was defined [as anterior tibial, posterior tibial or peroneal vein thrombosis]; "Proximal DVT" was defined [as popliteal, femoral, or external iliac vein thrombosis], while "Extensive DVT" was defined [as thrombosis involving the whole leg (extending from below to above the popliteal vein)]. The diagnosis of pulmonary embolism was confirmed by Spiral CT angiography. During the analysis we grouped the patients into a high-risk group (patients who developed PE on top of proximal or extensive DVT) and low risk group (those who are diagnosed proximal or extensive DVT but did not develop PE).

Statistical Analysis:
The power analysis was performed using G*power software (version 3.1.9.2] to determine the minimum sample size needed for this study. For a medium effect size (f2 = 0.15) and an alpha level = 0.05, the minimum sample size needed to achieve a .80 power was 55.
Descriptive data was expressed by mean, standard deviation, median, minimum and 6 maximum for continuous variables and frequencies for categorical variables.
For bivariate analysis; the non-parametric Spearmen correlation was used to assess the association between D-dimer and length of hospital stay and duration of symptoms. The Kruskal-Wallis test was used to establish the association between D-dimer and location of DVT and site of proximal DVT. Finally, the Mann-Whitney test was used to assess the differences in D-dimer values between the patients with PE and without PE.
A multinomial logistic regression was conducted to model the relationship between the independent variable D-dimer and membership to the three groups of DVT location (distal, proximal, extensive) and it was adjusted for age and gender. Malignancy, pregnancy and surgery could not be included as covariates since there were not cases of malignancy, pregnancy or surgery for the distal category of DVT. We ran two models where we used Ddimer as continuous and then as two groups: below and above the value its median (=4.39) We used ROC curves to study the feasibility of using Dimer level to identify patients at high risk. In particular, the area under the curve (AUC) is used to estimate the accuracy of Dimer levels to separate high risk from moderate/low risk patients. An area of 1 represents a perfect diagnostic test and an area of .5 represents a worthless test, in the sense that flipping a coin would provide the same level of accuracy. The following point system is were used to establish a cutoff that will provide optimal values of sensitivity and specificity. [12] Finally, the correlation between D-dimer and occurrence of PE was addressed with a logistic regression to model the likelihood of developing a Pulmonary Embolism based on D-dimer.

7
The model was adjusted for age, gender, surgery and DVT (proximal vs extensive). D-dimer was categorized into two groups according to the suggested cutoff level from the ROC curve: below and above the value of 4.75.

Main Patient Characteristics:
Sixty-four of patients were female and the average age was 44.8 with 65.5% of the patients had proximal DVT, a 28.2% had extensive DVT and only 6.4% were diagnosed as having distal DVT. Most of the patients with proximal DVT had thrombosis at the femoral vein 52.8%, 26.4% at the popliteal and 20.8% at the iliac vein. The incidence of PE in the sample was 11%, (12 out of 110) either upon presentation or during hospital stay (average length of stay was 6 days). All patients with PE were from the proximal or the extensive group (50% each). All of them were hemodynamically stable. Only two of our patients had malignancy, 24 % developed DVT post-operatively and 8% during pregnancy. The duration of DVT symptoms ranged from 1 day to 20 days (Table 1). Table 1.

Descriptive Characteristics of Study Sample (N=110)
The association between D-dimer level and DVT Location: Even on multinomial logistic regression analysis which was adjusted for age, gender, and surgery using D-dimer as continuous variable (Model 1), we found that there was not a significant improvement in the model with predictors compared to the model with only the intercept (H=4.43, p=.619). As seen in (Table 2), the unique contribution of D-dimer was not statistically significant, in other words, D-dimer does not significantly predict the location of DVT (H=.688, p=.709). Table 2 Unique contributions of predictors in multinomial logistic regression predicting DVT location As reported in (Table 3) (Table 3), we obtained similar results when D-dimer was categorized into 2 categories: above and below the median value (4.39 mg/L). When using D-dimer quartiles to create 4 groups (<=2.72, 272 to 4.39, 4.39 to 9.48 and >9.48) results did not provide any significant findings either.
(See Appendix B for further details). Table 3 Parameter estimates of multinomial logistic regression predicting DVT location The correlation between D-dimer and High-Risk Group (defined as patients who developed PE on top of proximal or extensive DVT): We risk-stratified the patients based on the risk of pulmonary embolism; into high risk group (those are the patients with proximal or extensive DVT who had symptomatic PE upon presentation or later during hospital stay) and low risk group (patients with DVT who did not develop PE) to try to find its association with D-dimer level . We found that higher level of Ddimer was associated with the high-risk group, according to a ROC curve analysis ( Figure 1).
The Area Under the Curve (AUC) suggested that D-dimer is a fair test to classify patients as   Logistic regression for D-dimer's cut-off predicting PE, controlling for age, gender and DVT location.

Discussion
Most of our patients had proximal DVT and only (6%) were with distal DVT, that can be attributed to the fact that most of the asymptomatic DVT are distal DVT [13]. Patients with extensive DVT had higher D-dimer level than proximal group and distal group, but we could not establish a significant relationship between D-dimer level and location of DVT, or with the site of proximal DVT (popliteal, femoral, and iliac) with P value = 0.519 and 0.336, 11 respectively.
In the contrary to our results, a pilot study [14], included 249 patients with suspected DVT, 50 of which had confirmed DVT, found that high D-dimer was significantly correlated with proximal DVT. They concluded that patients with D-dimer more than 1000 ng/ml are at increased risk of proximal DVT. Though we had a larger sample size (110 patients), the percentage of distal DVT (6%) was lower than their study (34%).
In our study, higher D-dimer level was associated with an increase in length of hospital stay (p=.015), but there was no correlation between level of D-dimer and duration of symptoms (ranging from 1 day to 20 days).
Interestingly, we found that D-dimer level is higher in patients with PE compared to patients who did not develop PE (P=0.027). We further proved that when we classified the patients into high risk group (patients who developed PE on top of proximal or extensive DVT) and low risk group and we found that D-dimer is a fair test to risk stratify patients with proximal or extensive DVT into high risk of PE (according to ROC curve figure 1). We suggest 4.75 mg/l as a cut-off level above which patients with proximal or extensive DVT can be considered as high-risk for developing PE. This cut-off has a sensitivity of 66% and a specificity of 63%; implying that the probability to correctly identify patients with proximal or extensive DVT as high risk of PE when they have D-dimer above 4.75 mg/L is 66%.
In patients with PE; several studies have found a significant association between D-dimer level and the extend of PE. A prospective study concluded that higher levels of D-dimer was associated with a greater perfusion defect on patients with PE who are hemodynamically stable. They found that D-dimer more than 4000micr/l (4mg/L) was associated with perfusion defect > 50% [2]. another study found that high D-dimer was also significantly associated with the proximity of PE as main pulmonary arteries PE was associated with high D-dimer (9.2 mg/l) while lobar and segmental PE patients had D-dimer of 3.8 and 1.4 mg/l respectively [6].
We examined the predictive value of D-dimer for occurrence PE in patients with proximal or extensive DVT using the cut-off of 4.75mg/l. The logistic regression model shows that Ddimer is a significant predictor of PE, (OR=7.9, p=0.013). In particular, patients with proximal or extensive DVT is at 8-folds increased risk of PE than patients with D-dimer less than 4.75mg/l. To our knowledge, there is no other study examined the ability of D-dimer to predict PE in patients with DVT.
The prognostic value of D-dimer in patients with PE was also examined in two studies derived from the RIETER Registry of VTE. They have linked higher level of D-dimer with increased mortality in patients with acute symptomatic PE [7.8].
Grau E et al showed that PE patients with D-dimer more than 5000ng/ml are at 2.9-fold increased risk of overall mortality compared to patients with D-dimer less than 5000ng/ml [7].
The Second study assessed the association between D-dimer and 15-day mortality in patients with acute PE. Patients with higher D-dimer (4200 ng/ml) had a higher overall mortality rate 7% compared to 2.7% in patients with lesser D-dimer (<1050ng/ml) [8]. Both studies concluded that level of D-dimer is a strong, independent predictor of overall and PErelated mortality (at 15 days and 3-months) among patients with acute symptomatic PE. On the other hand; one retrospective study which involved 292 patients failed to show a similar significant association between high D-dimer and increased rate of in-hospital overall or PErelated mortality among patients with stable acute pulmonary embolism [15]. Though rate of mortality was 2.9% in patient with D-dimer > 5000 ng/ml compared to 0% with D-dimer <5000 ng/ml; it was not statistically significant (P value 0.06). This has been shown in one more study [16]. Though in one prospective study, that included 699 patients with VTE found that high D-dimer level ( >8 mg/l) is associated with decreased overall survival; this finding was limited to the patients with malignancy (P <0.001) while this association was statistically nonsignificant (P<0.282) in patients without malignancy [17].
Limitation of the study: Our study lacks the balance between the three groups of DVT, with distal DVT representing only 6.4% of the sample, that might have affected the variation in the sample and therefore the results. Moreover, the small percentage of PE patients (only 11%) may have influenced the statistical analysis. Large, prospective studies are needed to evaluate the use of Ddimer beyond its diagnostic value as a predictor for DVT location and incidence of PE in DVT patients.

Conclusion:
Though level of D-dimer is not significantly associated with the location of DVT, it is a significant predictor of PE in patients hospitalized with proximal or extensive DVT. Based on our results, this can be applied to the daily practice by being more careful with the patients with proximal or extensive DVT and D-dimer level above 4.75mg/l by admitting those patients into a monitored area and early administration of anticoagulation as they are at increased risk of developing pulmonary embolism. While those patients with D-dimer level less than 4.75 mg/l are at lower risk of developing PE during hospitalization and they are more suitable for early discharge to complete the treatment as outpatients.   Distal DVT is the reference group.
OR: odds ratio Reference categories for gender is male, for DVT location is extensive and for D-dimer is below or equal to 4.75 mg/dl; analysis restricted to patients with proximal or extensive DVT. Figure 1 The value of AUC is .700, which indicates that D-dimer is a fair test to classify patients as high-risk of PE

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.