Page 199 - Haematologica-April 2018

Major bleeding associated with oral anticoagulants
co-morbidities between them. The type of OAC was associated with the site of bleeding (P<0.001); after adjust- ing for age and bleeding provocation using multinomial models, the conditional odds ratios of admission with a subdural/epidural, subarachnoid and intracerebral bleed (versus a lower gastrointestinal bleed) on DOAC compared to warfarin, were 0.28 (95% CI: 0.17-0.47; P<0.001), 0.52 (95% CI: 0.30-0.90; P=0.02) and 0.61 (95% CI: 0.41-0.92; P=0.02), respectively. Mortality rate, the number of com- plications following bleeding, and the duration of hospi- talization were not significantly different between patients treated with warfarin or DOAC.
Table 4 shows the analysis of risk factors for fatality in 2,132 patients with known outcomes. After controlling for age, co-morbidities, bleeding site and provocation, there was no evidence that compared to warfarin, DOAC were associated with different odds of death (adjusted OR=0.96; 95% CI: 0.71-1.28; P=0.77). Intracerebral, sub- arachnoid and subdural/epidural bleeds were associated with 5.8 (95% CI: 4.0-8.2), 3.2 (95% CI: 1.8-5.7) and 2.6 (95% CI: 1.7-3.8) times higher odds of death, respectively, compared with other bleeds. Adjusting for other covari- ates, advanced age, spontaneous bleeding (compared to trauma or falls), liver failure and cancer were significantly associated with increased odds of dying, while indications for OAC were not predictors of an adverse outcome. There was negligible between-hospital variation (intra- class correlation coefficient=0.003). Given the relatively small number of patients for whom outcome was missing (n=60), sensitivity analyses were performed by assuming either all survived or died, and this did not materially alter the significance of the estimates.
Discussion
The main objective of this study was to investigate the aggregate burden of major hemorrhage in patients receiv- ing OAC, across all indications and drug types, in terms of ensuing mortality, morbidity and hospitalization. Our results showed a considerable death rate, with one-fifth of all patients, and one-third of those with intracranial hem- orrhage, dying in hospital within 30 days of a major bleed- ing episode. Of those who were discharged within a month, half had spent 7-30 days in hospital. Our mortality rates, both overall and for the subgroup with intracranial hemorrhage, are comparable with those of other stud- ies.17,18 However, a recent Canadian study that used a sim- ilar definition of major bleeding as ours (including only patients >66 years) showed lower mortality rates than in our study (9.2% for DOAC and 15.2% for warfarin): this difference is likely to be due to the lower rate of intracra- nial hemorrhage (27% overall) seen in their study.19
The proportion of patients with intracranial hemor- rhage (44%) was similar to that in the study by Becattini and colleagues,20 but higher than that in other studies.19,21 The higher rate may be due to: (i) one-third of our cases being from hospitals with specialist neurosurgical units (i.e. endpoints of inter-hospital referrals); (ii) our cohort including patients who were older than those in clinical trials and thus more prone to falls and the development of intracranial hemorrhage; (iii) our definition of major bleed- ing being more severe than those of clinical trials, as we identified only patients who developed major hemorrhage culminating in hospitalization; (iv) cases of intracranial
hemorrhage being more exhaustively documented (e.g. through administration of prothrombin complex concen- trates), compared with gastrointestinal or other bleeds.
The pivotal randomized controlled trials of patients with atrial fibrillation and venous thromboembolism found lower rates of intracranial hemorrhage with all DOAC than with warfarin, and higher rates of gastroin- testinal bleeding with dabigatran and rivaroxaban than with warfarin.9,21 A systematic review of randomized clin- ical trials by Connolly and colleagues22 concluded that “the risks of subdural hematoma were significantly higher with vitamin K antagonists compared with factor-Xa inhibitors and direct thrombin inhibitors” [meta-analysis odds ratio (95% CI): 2.9 (2.1–4.1) and 1.8 (1.2–2.7), respectively]. Our findings appear to be a corollary of the established evi- dence: while our study did not directly assess the inci- dence of bleeding on OAC, we observed increased odds of admission with all types of intracranial hemorrhage versus gastrointestinal bleeding on warfarin compared to DOAC, with the association being strongest for subdural/epidural bleeding. Further investigation into this association between warfarin and subdural hematoma would thus be an important topic in future research.
Analysis of risk factors for death showed that intracra- nial hemorrhage was a strong predictor of death. However, we saw no difference in the proportions of deaths between patients treated with warfarin or DOAC even though the proportion of intracranial hemorrhage was higher for the former. A likely explanation for this is that patients admitted on DOAC were significantly older than those on warfarin, and advanced age was also an independent risk factor for death. The participants in the clinical trials of warfarin versus DOAC were significantly younger than those in our study. However, our results pro- vide reassurance for clinicians who prescribe DOAC ther- apy in that, where major bleeding is concerned, the out- come is no worse than that associated with warfarin despite the lack of antidotes and the advanced age of many patients in this study. Since DOAC antidotes are now becoming available it is important to perform similar studies to assess the effect that these have on outcomes, and thereby improve OAC choices in the future.
A key strength of this study is that the data were prospectively collected by over 30 hospitals across the UK, with information retrieved directly from patients’ case notes and not from large databases in which data are col- lected for different reasons, or missed by administrative coding.23-26 This has enabled us to document for the first time in the UK the characteristics and clinical outcomes of patients who develop major bleeding associated with dif- ferent OAC when prescribed for any clinical indications. Additionally, over three-quarters of cases were fully reported within 90 days of the major bleeding episode, ensuring a high degree of fidelity with actual events and minimizing the chance of recall bias. Furthermore, we were able to collect data on the acute management of major bleeding, whereas this is not usually possible from large data sets. Our analysis showed a very low level of inter-hospital variation, which suggests that our findings should be generalizable to the wider population and potentially further afield. Moreover, the design of the study meant that we did not rely on patients’ consent, a further potential source of selection bias.
Our study also provides data on the outcomes of major bleeding that are more generalizable to contemporary
haematologica | 2018; 103(4)
743