Anticoagulants for Cerebral Venous Thrombosis
Harmful to Patients?
See related article, p 8.
Evidence relating to efficacy and safety of anticoagulants for CVT is only partially available from studies in the peer reviewed medical literature and from the corresponding authors of those studies. Consequently, this article is an opinion piece rather than a systematic review.
Cerebral venous thrombosis (CVT) accounts for ≈0.5% of strokes and may occur at any age.1 Risk factors include pregnancy, oral contraceptives, head trauma, malignancy, infection, anemia, and dehydration. In older reports, the incidence of CVT was 3 to 4 per 1 million people per year in Europe and the United States.2 However, a survey published in 2012 found an incidence of CVT of 13.2 per 1 million people per year.3 With modern brain imaging technology, more cases are being discovered that probably would have resolved without medical intervention. Relative to pre-1990 clinical outcomes data, the average prognosis has dramatically improved.4,5
A Cochrane review of anticoagulant therapy for CVT4 based the recommendation in favor of using anticoagulation treatment (full-dose unfractionated heparin [UFH] or low-molecular-weight heparin [LMWH] followed by a vitamin K antagonist [VKA]) on 2 small randomized placebo-controlled trials (RCTs, n=79). These trials demonstrated a non-statistically significant reduction in deaths with anticoagulation (risk ratio, 0.33; 95% confidence interval [CI], 0.08–1.28). Subsequently, the European Federation of Neurological Sciences6 and the American Heart Association/American Stroke Association7 issued clinical practice guidelines calling for full-dose heparin acutely followed by VKAs for ≥3 months.
The efficacy and safety of anticoagulants for CVT have not been comprehensively reviewed, with the inclusion of observational studies and the RCTs excluded from the Cochrane review. This review will evaluate published reports from 1990 to 2013 regarding the efficacy and safety of full-dose heparin (UFH or LMWH) during initial hospitalization and subsequent VKA treatment for patients ≥15 years of age with CVT. The objectives of this review were to test the hypotheses that: (1) anticoagulant treatment with full-dose heparin during the initial hospitalization improves chances of survival, (2) posthospitalization VKA treatment improves chances of survival, (3) posthospitalization VKA treatment reduces the rate per month of recurrent venous thromboses (CVT and venous thromboembolism), and (4) anticoagulant-related bleeding risks are small relative to demonstrated benefits in survival and venous thrombosis recurrence reduction with anticoagulants.
A literature review was performed using search terms in PubMed: (cerebral vein or cerebral venous or sinus thrombosis [title/abstract]) and (heparin or anticoagulant or anticoagulation or warfarin or vitamin k [title/abstract]) and (study or randomized [title/abstract]) not (review or case report or children or pediatric [title]) and (1990 [date—publication]: 3000 [date—publication]). Secondary sources of CVT RCTs and studies came from reference lists of articles relating anticoagulant treatment to clinical outcomes.
All RCTs evaluating anticoagulation for CVT treatment encountered were included in the review. Retrospective and prospective observational studies were included if some or all patients with CVT received anticoagulants and if clinical outcomes were available in the article or might have been available from the authors. For each observational study and RCT identified by the PubMed search or reference list, I read the abstract (and if warranted, the full text), to establish whether it met the criteria for inclusion in the review. For RCTs and observational studies meeting the review criteria, I extracted anticoagulation statuses of patients with CVT (ie, full-dose heparin [UFH or LMWH], low-dose heparin, antiplatelet agent, thrombolysis, or no antithrombotic drug) and clinical outcomes (ie, death, major bleeding, heparin-induced thrombocytopenia [HIT], and recurrent venous thrombosis) in accordance with the review objectives. See Appendix I in the online-only Data Supplement for the protocol for data extraction. Where the published report contained missing information regarding anticoagulation statuses and clinical outcomes, I e-mailed the study authors for additional information.
χ2 tests showing the odds ratios (ORs), 95% CIs, and P values were used to compare anticoagulated versus unanticoagulated patients regarding clinical outcomes. SAS 9.2 was used for these analyses.
PubMed returned 95 references that fit the specifications of the search strategy. An additional 56 articles were found from the reference lists from CVT articles. These sources yielded 62 RCTs and observational studies (n=5155) that fit the criteria for the analysis (Figure I in the online-only Data Supplement). See Appendix II (online-only Data Supplement) for the bibliography of included studies. From 53 of these 62 studies, data were collected from the articles regarding the initial hospitalization (Tables 1–3). Data were gathered for the periods of postinitial hospitalization follow-up from 41 studies (Tables 4–6). Studies with no data on either anticoagulation status or clinical outcomes were omitted from the tables. Much of the data on anticoagulation statuses related to clinical outcomes was missing from the reports and was requested from the corresponding authors. The authors of only 7 of 62 studies provided all or part of the requested data.
Of 3854 patients acutely hospitalized for CVT, authors reported antithrombotic treatment status for 3438 patients (full-dose heparin [UFH or LMWH]: 2695; low-dose heparin: 138; antiplatelet agent: 60; VKA only: 28; thrombolysis only: 13; or no antithrombotic drug: 504). Authors reported anticoagulation status (ie, full-dose heparin [UFH or LMWH] versus not fully anticoagulated) related to survival of the initial hospitalization for 2768 patients (72%).
With these incomplete data, Table 1 shows the death rate in patients with CVT given full-dose heparin versus those not fully anticoagulated in hospital (ie, patients with CVT who received no antithrombotic drug treatment, low-dose heparin, or a platelet antagonist, excluding the few patients receiving thrombolysis alone or VKAs alone). The overall rate of deaths with full-dose heparin was significantly less than without anticoagulation—full-dose heparin: 9.1% (202/2211) versus no full-dose heparin, thrombolysis, or VKA: 14.0% (78/557), OR=0.62; 95% CI, 0.47 to 0.82; P=0.0007. However, for studies published after 1999, the risk of death was nonsignificantly greater with anticoagulation—full-dose heparin: 9.7% (192/1980) versus no full-dose heparin, thrombolysis, or VKA: 8.8% (9/102), OR=1.11; 95% CI, 0.55 to 2.24; P=0.77.
Regarding patients who received no full-dose heparin and died and had information reported on the cause of death and the presenting status (16/78 of such patients), Table 2 shows that they all either died of causes other than CVT, presented with poor prognostic signs, or received a weaker antithrombotic drug (ie, low-dose heparin or a platelet antagonist).
The Cerebral Venous Thrombosis Portuguese Collaborative Study Group (VENOPORT)8 was the only study to statistically analyze initial therapy selection bias: 31 of 49 (63%) patients with intracranial hemorrhage (ICH) were anticoagulated as opposed to 81 of 93 (87%) with no hemorrhage (χ2=10.9; P=0.0009). Despite the significant selection bias favoring full-dose heparin for relatively good prognosis patients over poor prognosis patients, anticoagulation was not statistically significantly correlated to the clinical outcome of death/dependency.
Major Bleeding During Full-Dose Heparin Treatment
Table 3 shows RCT and observational study authors’ reports of the incidence of any major bleeding (ie, new ICH or increased volume of ICH or extracranial bleeding) or HIT. Only 27% of patients with CVT who received full-dose heparin had major bleeding and HIT outcomes data reported (717/2695). A new ICH or an increased volume of ICH occurred in 2.6% of patients with CVT (17/643). Major systemic bleeding during full-dose heparin treatment occurred in 2.3% of patients with CVT with data available (14/598).
As shown in Table 3, CVT study authors specifically addressed HIT or commented on all heparin-related complications, thereby including HIT, in only 13.8% of the patients who received full-dose heparin (372/2695). For this relatively small proportion of patients with CVT for whom the incidence of HIT was mentioned or available from the authors, 1.3% developed HIT (5/372). In 1 case, the HIT was fatal.9
Posthospitalization Follow-up Deaths by Anticoagulation Status
More than 99% of anticoagulation given during follow-up to prevent thrombosis recurrences in these studies consisted of VKAs, primarily warfarin. Less than 1% of posthospital anticoagulation was LMWH or non-VKA oral anticoagulants. Consequently, VKAs will include these other agents in this analysis.
Table 4 shows the study data of deaths by VKA status during posthospitalization follow-up. Of 3085 patients with CVT enrolled in trials and studies with follow-up data, 79% of patients had survival information (2433 patients). Of these patients, 72 died during follow-up (2.9%). Authors published or made available to me by e-mail survival data by anticoagulation status on only 24% of patients (733/3085). Although the low percentage of patients with survival outcome data by anticoagulation status limits the value of the comparison, patients receiving posthospitalization VKAs had significantly lower death rates than unanticoagulated patients (VKAs: 10/618 [1.6%] versus no anticoagulation: 8/115 [7.0%]; OR=0.22; 95% CI, 0.08–0.57; P=0.0007).
Posthospitalization Follow-up Venous Thrombosis Recurrences by Anticoagulation Status
Given the variability in follow-up durations (ie, mean follow-up ≤3 months to ≥3 years), the most meaningful measure of venous thrombosis recurrences cerebral venous thrombosis + venous thromboembolism (CVT+VTE) was the rate of patients with venous thrombosis recurrence per patient-month (ie, counting patients with recurrences rather than total number of venous thrombosis recurrences). Calculating the rates per month of patients with venous thrombosis recurrences in studies with no published mean duration of VKA treatment required estimating the mean duration. For studies that did not report the mean duration of VKA treatment, 9 months of VKA treatment on average was imputed based on the International Study on Cerebral Venous Thrombosis (ISCVT) (follow-up phase; n=589), which reported a median VKA duration of 7.7 months, median duration of overall follow-up of 16 months, and mean duration of follow-up of 18.6 months (18.6/16×7.7=8.95).
As shown in Table 5, CVT authors reported the data on recurrence of venous thrombosis (CVT or VTE) in only 18 of 41 long-term studies. The overall rate of venous recurrence during follow-up was available for 85% of the total patient-months of follow-up (78 760/92 847 patient-months). The overall recurrence rate was 0.22% patient recurrences per patient-month (172 patients with venous recurrences/78 760 patient-months).
For the patients with data on venous thrombosis recurrences by anticoagulation status (ie, while on VKAs versus while not on VKAs), the patient recurrence rate while on VKAs significantly exceeded the rate while not taking VKAs (on VKAs: 0.33%/mo [35/10 761] versus not on VKAs: 0.20%/mo [63/30 963]; OR=1.60; 95% CI, 1.06–2.42; P=0.0246).
Table 6 shows the data on major bleeding while patients were taking VKAs during follow-up (major bleeding: 0.21%/patient-month [10 major bleeds/4813 patient-months], nonfatal brain bleeds: 0.04%/patient-month [2/4813], and fatal bleeds/mo: 0.06%/patient-month [3/4813]). Although the ISCVT did not report the overall bleeding rates, it included bleeding rates in the subgroup of patients ≥65 years of age: 0.60%/patient-month (fatal bleeds: 2/333 patient-months [2/37 patients on VKAs]).
Selective reporting was suggested by the limited availability of data on the relationship of anticoagulation to clinical outcomes and major variation in outcomes between studies. This makes the statistical finding that anticoagulation was associated with a reduced chance of death acutely highly doubtful. In addition, the overall significant reduction of initial hospitalization deaths with full-dose heparin (Table 1) may have also reflected that in older studies, physicians tended to avoid anticoagulation in patients presenting with poor prognostic signs. For example, the VENOPORT study from Portugal (n=142; data accumulated 1980–1998), which did not detail its hospital deaths by anticoagulation status, reported that patients with ICHs on presentation (a poor prognostic sign in many studies) were significantly less likely to receive anticoagulation treatment (P=0.0009).8 Full-dose heparin has been used in 95% of patients with CVT reported since 1999 (Table 1), and the studies published since 1999 have shown a weak trend toward a higher death rate with full-dose heparin (OR=1.11; 95% CI, 0.55–2.24; P=0.77).
Serious complications of heparin (Table 3) may well have also been underestimated by the selective reporting (only 24% of heparinized patients with bleeding data: 643/2695). These complications cannot be justified because the evidence since 1999 does not support that the overall risk of death is reduced.
Likewise, the meager data available relating VKAs to fewer deaths in follow-up studies (OR=0.22; 95% CI, 0.08–0.57; P=0.0007) cannot be used as an endorsement of VKA treatment. It more likely represents selective reporting. The authors reported the anticoagulation statuses of only 18 of 72 patients who died during follow-up (Table 4). The Mayo Clinic CVT follow-up study of Gosk-Bierska et al10 accounted for 13 of 18 of the reported deaths. In that study, 12 of 13 deaths were from causes other than recurrent venous thrombosis, and the death rate was almost equal between groups (VKA treated: 6/77 versus no anticoagulation: 7/77).
The available data show that venous thrombosis recurrences occurred more frequently while patients were taking VKAs compared with when they were not (OR=1.60; 95% CI, 1.06–2.42; P=0.0246). Again, anticoagulation status of only 57% of patients with venous thrombosis recurrences were published (98/172). The chances are slim that a full accounting of the 172 patients with CVT who developed recurrent venous thromboses, if available from the study authors, would show that the overall risk of recurrence was less while taking VKAs than while off of VKAs.
Data were not available to ascertain whether there was a significant increase in VTE recurrences in the first 2 months after stopping of anticoagulant drugs attributable to rebound hypercoagulation after stopping VKAs, as has been shown in patients with VTE treated with oral anticoagulants.11
The data from this review are inadequate to fully assess the risk of major bleeding with VKAs (0.21%/mo) with patients with CVT (Table 6). For comparison with a common indication for VKAs, the mean rate of major bleeding found in a meta-analysis of observational studies of warfarin for atrial fibrillation (n=484 241 patient-years) was 0.37% per patient-month.12
Quality of Trials Used to Support Anticoagulants for CVT
The European and American guidelines supporting anticoagulant drugs for CVT rest on only 2 placebo-controlled RCTs.13,14 De Bruijn et al14 conducted an RCT in the Netherlands and the United Kingdom in the early 1990s (n=59). At the end of 3 weeks, the proportion of patients with CVT dead or disabled in the treatment group was virtually the same as the control group (anticoagulated: 6/30 versus control: 7/29). Consequently, the case for full-dose heparin rests entirely on an RCT conducted by Einhäupl et al13 from 1981 to 1984 (n=20).
The premature stoppage of the Einhaupl trial because of more deaths in the control group (3/10 deaths versus 0/10 deaths anticoagulated group; P=nonsignificant) meant that there was an early termination bias. In this trial, the long delay before the treatment was started (mean delays: 33 and 25 days for the heparin and placebo groups, respectively) makes it highly dubious that anticoagulation accounted for the difference in mortality. The Einhaupl article reports that 3 of 10 control arm patients developed new intracranial hemorrhages (ICHs) after the initiation of placebo treatment. However, the development of new ICHs weeks after CVT diagnosis is inconsistent with the natural history of CVT and needs to be explained. Major bleeding was significant: complications in the heparin group included hematuria in 1 patient, a hematoma after puncture of the femoral artery in another patient, and a stillbirth of a 38-week-old fetus in a pregnant woman.13 The death of the fetus at 38 weeks of the woman in the anticoagulated group was not tallied as a death in the experimental treatment arm of the trial as it arguably should have been.
One of the 3 patients in the control group who died was treated with full-dose heparin before death and still counted as a control group death: In the control group, 1 patient had severe pulmonary infarction. This patient was subsequently given heparin but died 2 hours later of cardiac arrest.13 In this case, the diagnosis of pulmonary infarction was a clinical diagnosis. Nor was there any confirmation at autopsy that the diagnosis was indeed pulmonary infarction. The results of the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) study, showing that 75% of patients suspected of having pulmonary emboli do not have them, came out in 1990,15 well after this RCT was conducted (1981–1984). This means that pulmonary emboli diagnoses from earlier studies, such as this Einhaupl RCT using only clinical diagnoses of pulmonary emboli, need to be reconsidered.
Selective reporting is likely. The available evidence does not support the use of full-dose heparin during the initial hospitalization. Indeed, available data from studies since 1999 suggest no mortality benefit or even a possible increase in mortality because of anticoagulants. Too little study data on VKAs given or not given during posthospitalization follow-up have been divulged to conclude that VKAs prevented deaths in patients with CVT. In addition, the risk of venous thrombosis recurrence was significantly higher while patients took VKAs compared with while not taking VKAs. The risk of harm from anticoagulants, particularly from bleeding, is considerable. Anticoagulation for CVT is not evidence based to be effective or safe.
The opinions expressed in the article are not necessarily those of the editors or of the American Heart Association.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.113.003519/-/DC1.
- Received September 11, 2013.
- Accepted October 3, 2013.
- © 2013 American Heart Association, Inc.
- Coutinho JM,
- Zuurbier SM,
- Aramideh M,
- Stam J
- Coutinho J,
- de Bruijn SFTM,
- deVeber G,
- Stam J
- Ferro JM,
- Canhao P,
- Stam J,
- Bousser MG,
- Barinagarrementeria F
- Saposnik G,
- Barinagarrementeria F,
- Brown RD Jr.,
- Bushnell CD,
- Cucchiara B,
- Cushman M,
- et al
- Roskell NS,
- Samuel M,
- Noack H,
- Monz BU
- de Bruijn SF,
- Stam J