Venous thromboembolic disease (VTD) refers to both deep venous thrombosis (DVT) and pulmonary embolism (PE). VTD occurs at an incidence of 1/1000 individuals/year, and is the third most common cardiovascular disease, after acute coronary syndrome and stroke.1

VTD can have different underlying causes, and the pathophysiological mechanism involves a number of risk factors associated with the individual predisposition, be it genetic or acquired, of each subject. When these risk factors and predisposing conditions are known (e.g. surgery, immobilization, hormone treatment, pregnancy, puerperium, and cancer), VTD is said to be provoked. Even so, approximately 50% of VTDs occur in the absence of a clear precipitating factor, in which case they are classified as unprovoked or idiopathic.2–4

The optimal duration of anticoagulant treatment in patients with unprovoked VTD has not yet been determined, since the cumulative risk of recurrent VTD at 5 years in patients who discontinue anticoagulation is 25%–30%.4–6 For this reason, continued anticoagulant treatment is recommended if the risk of bleeding is low.4–7 However, current evidence points to risk factors, such as sex and D-dimer, that can modulate the risk of recurrence following discontinuation of anticoagulation.5,6,8 Practically speaking, we must analyze the case fatality rate of major bleeding, which is 2–3 times higher than the fatality rate of recurrent VTD,9,10 for which reason discontinuation of extended anticoagulant therapy is recommended when the risk of recurrence during the first year is less than 5%–8%.11–13

Several clinical prediction models3,14 have been developed and validated to individually assess the risk of recurrent VTD in patients with unprovoked VTD who discontinue anticoagulant treatment, although none has been validated in Spain. Two of the most widely known validated scales are the DASH prediction score13,15 and the Venice nomogram.16,17 The aim of this study was to perform an external validation in our setting of the DASH score and the Vienna nomogram, reevaluating the calibration of the predictions (e.g. the precision of the probability of recurrent VTD as predicted by the models compared with those observed in the new cohort).

During the study period, we assessed 353 patients, of whom 195 met the inclusion criteria (flow chart, Fig. 1). All patients were Caucasian, with a mean age (±SD) of 53.5 (±19) years and a predominance of women (57%). VTD presented as: DVT (75.4%), PE (14.9%), and DVT plus PE (9.7%). A family history of VTD was recorded in 7.2% (n=14) of the patients, and 7% (n=11) had prior VTD. Of the 11 patients with a personal history of VTD, 80% (n=9) had previously had provoked VTD, and 20% (n=2) unprovoked VTD but with a high risk of bleeding for whom discontinuation of anticoagulant treatment was proposed. Table 1 shows the main characteristics of the study patients.

Fig. 1.

Flow chart.

(0.2MB).

Bleeding was the reason for proposing the discontinuation of anticoagulant treatment in 5.2%. Mean time (±) receiving anticoagulation and mean follow-up after discontinuing anticoagulant treatment was 14.1 (±11.9) and 48.3 (±35.4) months, respectively.

We evaluated the 195 patients using the DASH score, and 42% and 58% of patients were classified as high and low risk, respectively. To apply the Vienna nomogram, we excluded patients receiving hormone treatment (n=36), giving a final sample of 159 patients who were classified as low vs high risk, 30% and 70%, respectively. Fig. 2 shows the ROC curve evaluating the probability of each of the scales predicting recurrent VTD at 12 months.

Fig. 2.

Receiver operating characteristics (ROC) curve based on the DASH and Vienna models applied in our series.

(0.18MB).

Twenty-one recurrences were observed in the validation of the DASH score (n=195) (10.8%, 95% CI: 6.8%–16%) during the first 12 months after anticoagulant treatment was discontinued. In 95% of cases (n=20), recurrences presented as DVT. The cumulative incidence in low and high-risk patients was 4.9% (95% CI: 1.3%–12%), and 15% (95% CI: 9%–23%), respectively (P<.05) (Fig. 3). Twenty recurrences were observed in the validation of the Vienna nomogram (n=159) (12.6%, 95% CI: 7.9%–18.8%) during the first 12 months after anticoagulant treatment was discontinued. After completing the Vienna nomogram, we divided the sample by quartiles (Q), observing a cumulative incidence of recurrence of 4.2% (95% CI: 0.5%–14%) in Q1 vs. 16.2% (95% CI: 9.9%–24.4%) in Q2–4 (P<.05) (Fig. 4). We calibrated both models by comparing the observed incidence of recurrent VTD with the incidence predicted in the validation cohorts (Table 2).

Fig. 3.

Time-to-event curve at 12 months in patients with low-risk (≤1) vs. high risk (≥2) DASH scores.

(0.11MB).

Fig. 4.

Time-to-event curve at 12 months in patients with Vienna nomogram Q1 vs. Q2–4.

(0.1MB).

Forty recurrences (20.5%; 95% CI: 15.1–26.9) occurred during follow-up (median 3.9 years; interquartile range: 1.1–6.3 years). After application of the DASH score, recurrences in low-risk and high-risk patients were 13.4% (95%; 6.9%–22,7%) and 25.7% (95% CI: 17.9%–34.7%), respectively (P=.05) (Table 3, Fig. 5). For the Vienna nomogram, the cumulative incidence in low-risk and high-risk patients was 14.6% (95% CI: 6.1%–27.8%) and 28.8% (95% CI: 20.6%–38.2%), respectively (P<.05) (Fig. 6).

Fig. 5.

Long-term time-to-event curve in patients with low-risk (≤1) vs. high risk (≥2) DASH scores.

(0.14MB).

Fig. 6.

Long-term time-to-event curve in patients with Vienna nomogram Q1 vs. Q2–4.

(0.14MB).

We performed a retrospective validation of the use of the DASH prediction score and the Vienna nomogram in our setting. These scales are used to identify patients with unprovoked VTD and a low risk of recurrence, in order to safely discontinue anticoagulant treatment indefinitely after at least 3 months of treatment. The implications of these results are that the duration of anticoagulant treatment may be optimized by safely discontinuing it in patients classified as low risk of recurrence, as shown by previous studies.11–13

The discontinuation of anticoagulant treatment in patients with unprovoked VTD is a hotly debated and controversial topic.12,27–29 In patients with unprovoked VTD, the guidelines recommend indefinite anticoagulant treatment unless the risk of bleeding is high.6,30 The International Society of Thrombosis and Hemostasis (ISTH) suggests that the discontinuation of anticoagulant treatment is appropriate in subgroups of patients with less than 5% risk of recurrent VTD at 1 year.12 Our study has validated both scores in a population with an incidence in the low-risk group of less than 5%. Given that the DASH score classifies more patients as low-risk than the Vienna nomogram (42% vs. 30%) and that it is easier to use, this would be the tool of choice.

Our study has several strengths. First, both scores are analyzed in the same patient cohort, so they can not only be validated, but also compared to see which includes more patients in the low-risk group, that is to say, the group that will benefit from the safe discontinuation of anticoagulation. To our knowledge, this is the only study that has validated these scales in Spain, giving them external validity and justifying their use in our setting (known as portability). Second, the clinical characteristics of our patients are similar to those of other previously published series. In our population the mean age was 54±19 years; in other studies, age ranged from 54 to 68 years, depending on the study.13,15–17 Third, all recurrences were documented objectively and reviewed by an independent committee (SMR and LJP).

Our study has some limitations. First, it is a retrospective study, with the inherent limitations of this design. However, it analyzes unselected consecutive patients with VTD. The DASH score and Vienna nomogram were also validated in retrospective patient series.13,17 It must be borne in mind that in some cases there may have been a selection bias, since in patients with a high risk of recurrent VTD, the clinician did not discontinue anticoagulation, although this limitation is also a feature of other previously published studies, due to their retrospective nature.13,17 In our study, we documented the number of patients who chose to continue anticoagulant treatment (14%), those in whom the clinician decided to maintain treatment (11%), and those in whom treatment was reintroduced after the detection of very high D dimer levels (3%). Similarly, in the validation study of the Vienna model, only 904 patients from a database of 1818 (50%) could be included in the external validation of the model.17 In the external validation of the DASH prediction score, no mention is made and no information is provided regarding patients who were not analyzed or who were excluded.13 Although it is preferable to perform the external validation prospectively from a methodological perspective, from a practical point of view we wished to determine the applicability and validity of both scales in our environment. Second, we have not analyzed other models, such as the Men-HERDOO2 rule,31 which evaluates sex (all men continue to receive anticoagulation), hyperpigmentation, edema or redness in the affected leg, D dimer during anticoagulation, and age. Although this method was validated prospectively, it has limitations, including the subjective assessment of signs in the affected limb and the fact that all men must continue anticoagulation indefinitely, parameters that do not figure in the other 2 scales. We could not validate this scale in our setting, as we lacked data on hyperpigmentation, edema or redness in the affected leg. Third, all our patients were Caucasian, which potentially limits generalization to other races. Fourth, we do not know the potential role of direct-acting oral anticoagulants when considering discontinuing anticoagulants, as these drugs have demonstrated a good safety profile, to the extent that some experts argue that if the annual risk of bleeding is greater than 3%, treatment should be discontinued, and if it is less than 3% a year, discontinuation should be considered if the patient is at low risk of recurrent VTD after evaluating their score.28 Nevertheless, this approach needs to be evaluated in a clinical trial. Fifth, it would have been interesting to conduct a subgroup analysis on the basis of gender or VTD site. Unfortunately, the sample was calculated to validate both scales, and our study does not have enough power to perform subanalyses.

Our study validates the DASH score and the Vienna nomogram in our population. The DASH prediction score may be the most useful, because of both its simplicity and its ability to identify more low-risk patients than the Vienna nomogram (42% vs. 30%).

Luis Jara Palomares has received honoraria for speaking engagements and travel and accommodation expenses from Rovi, Pfizer, Menarini, Leo-Pharma and GSK, unrelated to this manuscript. Maria Isabel Asensio Cruz has received travel and accommodation expenses from Rovi, Novartis, and Teva, unrelated to this manuscript. Raquel Morillo Guerrero has received honorariafor speaking engagements and travel and accommodation expenses from Pfizer, Menarini, and GSK, unrelated to this manuscript. The other authors state that they have no conflict of interests.