Regular article
Accuracy and usefulness of a clinical prediction rule and D-dimer testing in excluding deep vein thrombosis in cancer patients

https://doi.org/10.1016/j.thromres.2008.05.002Get rights and content

Abstract

Introduction

Deep vein thrombosis (DVT) can be safely and reliably excluded in patients with a low clinical probability and a negative D-dimer result but the accuracy and utility of such a strategy is less certain in cancer patients. We sough to compare the performance of the Wells pretest probability (PTP) model and D-dimer testing between patients with and without cancer and to examine the utility of the two PTP model classification schemes (low/moderate/high versus unlikely/likely) in excluding DVT in patients with cancer.

Materials and methods

Pooled analysis of databases from three prospective diagnostic studies evaluating consecutive outpatients with suspected DVT.

Results

A total of 2696 patients were evaluated. DVT was diagnosed in 403 (15%) patients overall and in 83 of 200 (41.5%) cancer patients. The PTP distribution and the prevalence of DVT in each PTP category were significantly different between patients with and without cancer, regardless of the classification used (p < 0.01). In patients with cancer, the negative predictive values of a low or unlikely PTP score in combination with a negative D-dimer result were 100% (95% CI 69.8%–100%) and 100% (95% CI 82.8%–96.6%), respectively. However, the specificities ranged from 46.2% (95%CI 27.1%–66.3%) to 57.1% (95%CI 41.1%–71.9%). Further testing was required in 94% of cancer patients using the low/moderate/high PTP classification and in 88% using the unlikely/likely stratification.

Conclusions

As in patients without cancer, the combination of a low or unlikely PTP with a negative D-dimer result can exclude DVT in patients with cancer. However, this strategy has limited utility because very few cancer patients present with this combination.

Introduction

Patients with malignancy have a high risk of developing venous thrombosis [1], [2], [3] and the treatment of thrombosis with anticoagulants is complicated by a substantial risk of bleeding [3], [4], [5], [6]. Therefore, accurate diagnosis of deep venous thrombosis (DVT) is required to appropriately treat patients with the disease and avoid unnecessary anticoagulation in those without venous thrombosis.

Clinical prediction models combine components of the history and physical examination to categorize a patient's probability of having a disease. The Wells pretest probability (PTP) model is validated and well established for diagnosing DVT in symptomatic patients. This model originally stratified patients as having a low, moderate, or high likelihood of having DVT but subsequent analyses showed that the model could be simplified by stratifying patients into either a likely or unlikely risk group [7], [8], [9], [10]. When used in combination with D-dimer testing [10], both of these stratifications can safely manage patients with suspected DVT. The negative predictive value, or the likelihood of not having disease, when a patient has a low or unlikely PTP and a negative D-dimer result, ranges from 99.1% to 99. 6% [8], [9], [10]. It has also been demonstrated that it is safe to withhold anticoagulant therapy in patients who have a negative D-dimer result and either a low or unlikely PTP at initial presentation [8], [9], [10].

But the clinical utility of such a strategy for ruling out DVT in patients with cancer is less certain. The studies that validated PTP models did include cancer patients but the scoring does not take into account the strong risk factors for DVT that are unique to cancer patients. Also, while some studies showed that D-dimer testing had comparable negative predictive values in patients with cancer compared to patients without cancer [11], others have not [12]. Finally, the high prevalence of DVT in cancer patients may sufficiently reduce the negative predictive value of these diagnostic tools and render them less useful [12].

In a retrospective study that evaluated the application of a PTP model and D-dimer testing in cancer patients with suspected DVT, a combination of a negative D-dimer result and a low or low-moderate PTP was useful in excluding DVT [13]. However, given that only 9% of cancer patients scored a low PTP, the authors suggested that the unlikely/likely PTP stratification might be more useful in clinical practice. Also, the confidence interval of the negative predictive value of this combination was as low as 88%, suggesting that up to 12% of patients may be missed diagnosed as not having DVT.

To further investigate the utility of PTP modeling and D-dimer testing in cancer patients, we performed a combined analysis of 3 large prospective cohort studies to compare the performance of the Wells PTP model and D-dimer testing between patients with and without cancer and to examine the utility of the two PTP model classification schemes (low/moderate/high versus unlikely/likely) in excluding DVT in patients with cancer.

Section snippets

Materials and methods

We combined the databases from three prospective diagnostic studies in consecutive outpatients with suspected DVT that were performed at 7 Canadian tertiary care centers between 1997 and 2002 (see Appendix A for names of centers) [8], [9], [10]. Patients were included in these studies if they presented with symptoms of DVT at either the emergency departments or the Thromboembolism clinics. Each study evaluated patients using the Wells PTP model (Table 1) [7], [10] and performed D-dimer testing

Results

A total of 2696 patients were evaluated. The characteristics of the study participants are displayed in Table 2. Deep vein thrombosis was diagnosed in 403 (15%) patients; 13 of these were confirmed during the 3-month follow-up. A total of 200 cancer patients were included. Of these patients, 83 (41.5%) were diagnosed with DVT. A total of 463 patients did not have D-dimer performed. Thirty patients from the 3 original studies did not have sufficient information in the databases and were excluded

Discussion

We investigated the safety and utility of excluding DVT in patients with cancer using different PTP categories (low or unlikely) in combination with D-dimer testing. As in patients without cancer, low or unlikely PTP categories in combination with a negative D-dimer result have very high negative predictive values and should be able to safely rule out DVT in patients with cancer. However, because D-dimer levels are often elevated in cancer patients in the absence of DVT, the D-dimer test has

Acknowledgements

P. Wells is a recipient of a Canada Research Chair. M. Carrier is a recipient of a Canadian Institute for Health Research Fellowship.

References (15)

There are more references available in the full text version of this article.

Cited by (0)

View full text