D-dimer for the exclusion of cerebral venous thrombosis: a meta-analysis of low risk patients with isolated headache

Headache frequently leads to emergency room consultation. It is vital to exclude secondary forms of headache requiring further treatment such as cerebral venous thrombosis (CVT). CVT is accompanied by headache in 89 % of cases [1, 2]. It may present with headache alone in 14 % [3] and acute headache in 3–13 % of cases [4‐6]. Patients with isolated headache seem to have a good prognosis, however, CVT patients with isolated headache who present early (<7 days) are more at risk to deteriorate neurologically than patients with isolated headache who present later, making diagnosis in an early stage important [6].

D-dimers have been proven useful in the diagnosis of pulmonary embolism (PE) and deep venous thrombosis (DVT) [7, 8]. Clinical factors are translated to the Wells score, ranking patients to high or low risk categories [7]. The Wells score for PE is calculated based on clinical parameters and patient history. The score includes clinical signs of DVT, PE being the most likely diagnosis, heart rate over 100 bpm, recent immobility or surgery, hemoptysis, previous PE or DVT or the presence of malignancy, Each item in the Wells score has a rating and patients with 1.5 point or less are rated low risk for PE. When low risk Wells score patients have a normal D-dimer, the post-test probability of DVT or PE is 0.5–2 % when using a sensitive ELISA quantative assay [9]. In these low risk patients it is safe to forego further imaging such as echo venography of the lower limbs or pulmonary spiral CT. Sensitivity of D-dimer varies from 83–96 % in diagnosing DVT patients and 75–97 % in diagnosing PE patients, depending on the performed D-dimer test [10].

Whether the measurement of D-dimer can play a similar role in suspected cerebral vein thrombosis remains uncertain. A recent meta-analysis of the literature concluded that D-dimer may be useful as a diagnostic tool in CVT patients in general, but an analysis for isolated headache was not the aim of this review [11]. According to American guidelines patients with a high clinical suspicion of CVT should receive neuro-imaging regardless of D-dimer results [12]. This makes D-dimer a superfluous test in high risk cases. The definition of low risk patients for CVT requires different criteria from the Wells score for PE of DVT. In current practice the presence of neurological deficit and abnormalities on standard head CT, but also known risk factors for CVT such as pregnancy or use of oral contraceptives are weighed in the decision to perform further diagnostic work-up. Whether D-dimers are sensitive enough to exclude CVT in low risk patients, without these characteristics and isolated headache is unknown. In these patients D-dimer would be valuable to decide whether or not additional imaging is necessary. There are conflicting and limited data on patients with isolated headache. One often cited study found negative D-dimers in five out of 19 (26 %) patients with CVT and isolated headache [13]. Another prospective study found no false negative D-dimers in 20 patients with isolated headache [14].

We aimed to assess whether D-dimer is a valuable test in CVT patients with isolated headache and a low risk of CVT.

Our study suggests that D-dimer is a sensitive diagnostic tool in excluding CVT in low risk patients. We found a high negative predictive value in patients with isolated headache suspected of CVT in our meta-analysis that combined data from our own and three previous series. This is comparable with the values found in PE and DVT when a low risk Wells score, based on physical examination, is combined with negative D-dimer. Our meta-analysis describes the largest group of patients with isolated headache studied for the diagnostic value of D-dimer in recent literature. This is also the only study focusing on low risk CVT patients, making the results valuable for everyday practice.

Our study has limitations. First, insufficient data on isolated headache and concomitant D-dimer levels were available from six potential useful articles [13, 17‐20, 22]. We attempted to obtain additional information from the concerned authors but received information from only one [17]. However, we did collect data on a large number of patients allowing a sufficiently precise estimation of the diagnostic value of D-dimer in this low risk group. Second, we could not determine how many headache patients presented to our hospital in whom D-dimer was not determined in this period. Unfortunately, it is not possible to reliably retrieve this information from the hospital information system. This means that in our evaluation very low risk patients may have been missed compromising the generalizability of the findings. On the other hand, the addition of patients from studies performed in other large hospitals in several countries improved the generalizability of our results..

Finally, the D-dimer determination method varied over the cited articles; however, most articles used quantitative latex assays, with high sensitivity with the same cut-off value. It was not possible to calculate D-dimer cut-off levels in this group as exact values were not given in most articles, in particular not in patients without CVT.

We defined patients to have a low risk of CVT when they had a normal neurological examination and normal standard head CT. The definition of a normal neurological examination might be a point of discussion as papillary edema for instance may be difficult to determine reliably and in many cases is not judged by the treating physician at all. In an earlier study patients with papillary edema were included in a cohort with isolated headache [6]. In our group two patients with papillary edema alone had a negative D-dimer, and were excluded from our group of patients with isolated headache. Also a symptom like tinnitus which may be present in raised intra cranial pressure is often not mentioned and could not be included in our definition.

A raised risk of CVT exists during pregnancy and puerperium and in patients using oral contraceptives, but also in older patients and patients with underlying malignancy. These patients are also at risk of having a false positive D-dimer. In the entire group three patients were reported to be pregnant and 22 were reported to be using contraceptives. However, the use of contraceptives or presence of pregnancy in the patients from literature could only be clearly deduced in those with established CVT. The limited number of patients and insufficient reporting of pregnancy and use of contraceptives makes it difficult to give recommendations. Caution in this high risk group seems warranted and we feel these patients should undergo follow up neuro-imaging when CVT is suspected regardless of D-dimer outcome. Of all consecutive patients with CVT there were 1.6 % false negative D-dimers. Earlier concerns regarding usability of D-dimer in this patient group was based on percentages found in studies citing percentages of false negative D-dimers ranging from 10–50 % [12, 15, 18, 20‐22]. However these articles did not mention D-dimer results in non-CVT patients or gave insufficient data on this control group. In our own patient group no false negative D-dimers were found, which was confirmed by three studies, including two large prospective studies [11, 17, 23]. Including these articles in our meta-analysis, the sensitivity of D-dimer in isolated headache patients for detection of CVT is 87 % (95 % CI: 79.0–93.0 %). This sensitivity is lower, but comparable to, sensitivity found in various tests used for D-dimer determination in PE and DVT patients [10]. Unfortunately as no matched isolated headache controls are described, the calculation of specificity and the negative and positive predictive value for this entire group is impossible. As the patients in most of these articles were collected over time in tertiary headache referral centers, they may cause an overestimation of CVT, with large numbers of controls balancing them out.

When evaluating the risk for false negative D-dimer this is higher in patients with a single affected sinus and patients with a longer duration of complaints [14]. Unfortunately information on both these parameters was only available in a small number of patients and conclusions on these factors cannot be drawn from our data. These parameters are an important focus of future prospective research.

In patients with isolated headache D-dimer could become a useful screening tool to exclude CVT and may help reduce unnecessary additional imaging using intra-venous contrast or MRI. In the presence of risk factors for CVT such as pregnancy, puerperium, use of oral contraceptives, neurological symptoms and abnormalities on standard head CT it remains necessary to perform diagnostic follow-up. As is the case in DVT and PE D-dimers cannot be seen a stand-alone test and must be combined with clinical risk factors. The high negative predictive value of 99.8 % may aid the physician in decision making, although a positive result does not prove the presence of CVT due to low positive predictive value. Although sensitivity is lower, it is comparable to sensitivity accepted in excluding PE and DVT. In this series the use of negative D-dimer would have avoided additional neuro-imaging for diagnosing CVT in 502 patients, whereas only 89 patients with false positive D-dimer would have been scanned unnecessarily. One patient would have been missed. Prospective studies confirming our findings concerning low risk CVT patients are needed.

D-dimers have a high negative predictive value in patients with isolated headache for excluding CVT. Sensitivity is lower but comparable to the values accepted in PE and DVT. Low risk patients were defined as headache patients with a normal neurological examination, normal standard head CT and absence of risk factors such as pregnancy or puerperium. Normal D-dimers in these patients may reduce unnecessary imaging, making it a potential valuable marker.