Modeling of the Impact on Health Outcomes of the Use of Dabigatran in Patients with Atrial FibrillationChevalier J.a · Giroud M.b · de Pouvourville G.a
aChair of Health Systems, ESSEC Business School, Cergy-Pontoise, and bDijon Stroke Registry, EA4184, Department of Neurology, University Hospital and Medical School of Dijon, University of Burgundy, Dijon, Paris Corresponding Author
Chair of Health Economics and Management
ESSEC Business School, Avenue Bernard Hirsch
PO Box 50105, FR-95021 Cergy-Pontoise Cedex (France)
Background: Atrial fibrillation (AF) is the main direct cause of stroke. Prevention by anticoagulation or antithrombotic treatment is required, vitamin K antagonists (VKAs) and aspirin being the main agents. Dabigatran etexilate is a novel oral direct thrombin inhibitor. The RE-LY study demonstrated that in patients with AF, the rates of stroke and systemic embolism were similar (at a dose of 110 mg) or lower (at a dose of 150 mg) than those observed in patients treated with warfarin, a VKA. The aim of the present study was to estimate, through modeling, the number of severe events avoided with dabigatran at dosages of 110 mg (D110) or 150 mg (D150) twice daily compared to warfarin, when prescribed in the French population for patients with AF who meet the inclusion criteria of the Randomized Evaluation of Long-Term Anticoagulant Therapy (RE-LY) study. Methods: We used a decision tree type model to simulate the outcome at 5 years in a cohort of patients eligible for treatment with dabigatran. We compared 3 hypothetical cohorts: all AF patients are treated with D110, D150 and warfarin. Based on the probabilities of occurrence of the different outcomes observed in the RE-LY study and in the Dijon Stroke Registry, we simulated for each year during 5 years the evolution of the 3 cohorts. The model allows to simulate a prevalent cohort of 461,392 patients at year zero, with a follow-up of 5 years (‘constant prevalence model'), and it can also take into account incident patients during 5 years (‘dynamic prevalence model'). The different events taken into account were: major hemorrhages (excluding hemorrhagic strokes), myocardial infarctions, hemorrhagic strokes, ischemic strokes, recurrence of strokes (without differentiating the mechanism) or deaths. Results: Considering the constant prevalence model, the use of D110 instead of warfarin for the whole target population in France would permit to avoid 10,012 events and to save 18,879 years of life in a period of 5 years. These figures are 13,484 and 27,736 for D150 instead of warfarin. Considering the dynamic prevalence model, the use of D110 for the whole target population in France would permit to avoid 13,620 events and to save 22,674 years of life in a period of 5 years. These figures are 18,186 and 33,091 for D150. Conclusion: The use of dabigatran would lead to a significant reduction of strokes and deaths attributable to AF in France.
© 2013 S. Karger AG, Basel
Atrial fibrillation (AF) is the most common heart rhythm condition in France. The incidence increases with age and 10% of people over 80 years of age are affected .
AF is the main direct cause of stroke. As strokes in AF patients are particularly severe and disabling, its prevention is of great importance in this population [2,3]. Vitamin K antagonists (VKAs) and aspirin are presently the main preventive agents. VKAs have been shown to reduce the risk of stroke by 68% compared to aspirin  but have a narrow therapeutic window of anticoagulation and present many interactions with food or other drugs. If overdosed, VKA administration presents an excess risk of hemorrhages, and protection from stroke is lessened in case of underdosing. Thus, VKAs require regular international normalized ratio (INR) monitoring and dose adjustments to be effective.
Dabigatran etexilate is a novel oral direct thrombin inhibitor which provides stable anticoagulation with a fixed dose and no need of anticoagulation monitoring. It has been marketed since 2008 for the prevention of venous thromboembolic events in patients undergoing scheduled surgery for total knee or hip replacement. The Randomized Evaluation of Long-Term Anticoagulant Therapy (RE-LY) study [5,6] demonstrated that dabigatran is effective in the prevention of thromboembolic events in patients with AF. It compared 2 dosages of dabigatran (110 and 150 mg) twice daily to warfarin (dosage adjusted on INR). It demonstrated that the rates of stroke and systemic embolism were similar (110 mg) or lower (150 mg) than that observed in patients treated with warfarin. The risk of major hemorrhages was similar (150 mg) or lower (110 mg).
The aim of the present study was to estimate the number of severe events avoided with dabigatran (110 or 150 mg) compared to warfarin, when prescribed in the French population of AF patients who met the inclusion criteria of the RE-LY study. For this purpose, the outcome at 5 years in a cohort of patients eligible for treatment with dabigatran was simulated for 3 hypothetical cohorts where all patients are treated: with dabigatran 110 mg (D110), dabigatran 150 mg (D150) and warfarin.
The 5-year time horizon corresponds to the periodicity of the reappraisal of a medicinal product by the French reimbursement body. The model provides an estimation of the ‘public health impact' of a new treatment, required to document access to reimbursement. Two scenarios were studied: one estimates the outcomes for a cohort of patients put under treatment at year 1 and followed up during 5 years, and one includes annual incident patients.
In 2009, the estimated prevalence of patients with AF in the French population was 569,298 patients, with an annual incidence of 94,530 . Taking into account the inclusion criteria of the RE-LY study, a potential cohort of 461,392 patients was retained (81% of AF patients) with an annual incidence of 72,534 .
Morbidity and mortality data used were derived mainly from the RE-LY study [5,6], as well as from data analysis of the Dijon Stroke Registry. Results of a meta-analysis on treatments used in the prevention of stroke in patients with AF  were also used to estimate the rate of occurrence of events for patients switching to aspirin after a hemorrhagic event.
The population-based Dijon Stroke Registry provides estimations on incidence rates, survival and recurrence after a stroke, according to etiology. It records all cases of stroke which occurred in patients living in the city of Dijon (150,000 inhabitants) and records data on 4,181 patients who experienced a stroke between 1985 and 2009. In the present study, patients aged over 18 years at the time of their stroke were selected. Only patients whose first stroke lasting more than 24 h had occurred before 2004 were kept in order to have a follow-up of at least 5 years. We chose the data interval of the last 20 years to trade off between recent data and a sufficient number of cases. The final number of cases was 2,605 patients (62% of the registry data).
In the registry, etiology is recorded using the following categories: hemorrhagic stroke (14.0%), lacunar stroke (23.3%), cardioembolic stroke (17.6%), other ischemic stroke (41.9%), other not specified.
The AF status of patients experiencing a stroke and registered was not available. Since the origin of strokes for AF patients is cardioembolic, we chose to model survival for AF patients after stroke using data relative to this etiology. Patients for whom the mechanism of stroke was either a cerebral hemorrhage or a subarachnoid hemorrhage were defined, in our study, as patients with hemorrhagic stroke.
We used a decision tree type model, programmed with SAS 9.1. We simulated for each year during 5 years the evolution of a cohort of patients eligible for a treatment with dabigatran. Each year the rate of occurrence of the different events was applied to the cohort of patients, taking into account events which had occurred in previous years. The model simulates a prevalent cohort of 461,392 patients, with a follow-up of 5 years (‘constant prevalence model'), and takes into account incident patients, assuming a constant incidence per year and all incident patients requiring anticoagulation (‘dynamic prevalence model').
The different events taken into account were: major hemorrhages, myocardial infarctions, hemorrhagic strokes, ischemic strokes, recurrence of strokes and deaths.
For each event, the rate of occurrence observed in the RE-LY study was used until a stroke occurred. After the occurrence of a stroke, the probability of recurrence and of death applied was based on the Dijon Stroke Registry's data year by year and differs depending on whether the event was an ischemic or hemorrhagic stroke. Due to the lack of specific data in the French context, we formulated the hypothesis that the occurrence of myocardial infarction did not change the rate of occurrence of consecutive events.
The median duration of follow-up in the RE-LY was 2.0 years; we assumed in the model that the differential efficacy of dabigatran versus warfarin was maintained from year 3 to 5.
The occurrence of the different outcomes was simulated according to 3 courses of treatment: D110, D150, warfarin.
The number of years of life saved was calculated as the difference between the number of years of life observed when all the patients are treated with dabigatran and with warfarin. For example, a patient alive after 5 years will contribute for 5 years in the total number of years of life whereas a patient dying at year 3 will contribute for 3 years.
Results are presented compared to warfarin (D110 vs. warfarin and D150 vs. warfarin). We counted the number of events avoided and the number of years of life saved at 5 years.
Strokes. The yearly rates of hemorrhagic stroke were 0.12, 0.10 and 0.38%, respectively, when patients were treated with D110, D150 and warfarin [5,6]. The relative risk (RR) of a hemorrhagic stroke during treatment with aspirin compared to warfarin is 0.55 . Based on data provided by Boehringer Ingelheim on request from the authors, 51.6% of hemorrhagic strokes were fatal at 90 days, whether patients were treated with D110, D150 or warfarin, and 25.9% in patients treated with aspirin.
The yearly rates of ischemic stroke were 1.34, 0.92 and 1.21%, respectively, when patients were treated with D110, D150 and warfarin [5,6]. The RR of ischemic stroke during treatment with aspirin compared to warfarin was 1.68 . Lastly, based on data provided by Boehringer Ingelheim on request from the authors, 21.1% of ischemic strokes were fatal at 90 days in patients treated with D110, 25.1% with D150, 22.6% with warfarin and 23.9% with aspirin.
The probabilities of survival after a stroke, of recurrence and of survival after recurrence were calculated based on data from the Dijon Stroke Registry.
Recurrence of Strokes. Based on the data from the registry, we considered a single recurrence in the 5-year period. Therefore, the events after a recurrence are a major hemorrhage, a myocardial infarction, or death.
Major Hemorrhages. The yearly rate of major hemorrhage was 2.75% when patients were treated with D110, 3.22% with D150 and 3.19% with warfarin (data provided by Boehringer Ingelheim on request from the authors). Van Walraven et al.  estimated that 18.2 and 15.4% of cases of major hemorrhages were fatal during treatment with warfarin and aspirin, respectively. We assumed that the proportion of cases of fatal major hemorrhages was the same during treatment with dabigatran as with warfarin.
Lastly, we assumed that the occurrence of a major hemorrhage resulted in discontinuation of anticoagulation therapy and its replacement by aspirin. The RR of major hemorrhage during treatment with aspirin (compared to warfarin) was 1.14 .
Myocardial Infarctions. The yearly rate of myocardial infarction was 0.82, 0.81 and 0.64% when patients were treated with D110, D150 and warfarin, respectively [5,6]. The RR of infarction during treatment with aspirin (vs. warfarin) was 1.60 . Although these rates were found to be not statistically different in the RE-LY study, data have been incorporated in our model to perform a probabilistic sensitivity analysis. Cases of infarction were fatal in 1.11% of cases (data from clinical report).
Deaths. The death rate from all causes (including fatal events) was 3.75, 3.64 and 4.13% when patients were treated with D110, D150 and warfarin, respectively. The RR of death during treatment with aspirin compared to warfarin was 1.09 .
All rates of occurence of events are presented in table 1.
We conducted a multivariate sensitivity analysis to estimate the confidence intervals around the difference in frequency of events (D110 vs. warfarin and D150 vs. warfarin).
Occurrence rates of the different events were assumed to follow a normal distribution and we used Wald's method to calculate the confidence interval. According to this method, it is defined as follows:
where p is the observed proportion in the sample, n is the sample size.
For each event, new rates of occurrence were randomly drawn in that confidence interval (CI) and 1,000 new rate sets were created. For each rate set, the difference between the number of events with dabigatran and with warfarin was calculated. We kept the 2.5th and the 97.5th percentiles to obtain a 95% CI for each difference in number of events.
413 (16%) cardioembolic strokes, 364 (14%) cerebral or subarachnoid hemorrhages and 1,828 (70%) strokes with another origin registered in the Dijon Stroke Registry met the inclusion criteria. Patients who had a cardioembolic stroke were older on average than others: the mean age in this group of patients was 80 years (SD = 9) versus 73 years in the other two groups (SD = 14-17; p < 0.0001). Lastly, 47% of the population was male.
Survival after a stroke differed depending on the mechanism of stroke (p < 0.0001). One year after a first stroke, 55% of patients were still alive when the stroke was cardioembolic, 57.7% when it was a hemorrhagic stroke, and 77.9% in other cases of stroke. At 5 years, 31.1% of patients were still alive after a cardioembolic stroke, 42.1% after a hemorrhagic stroke and 53.4% after another type of stroke (table 2).
430 cases of stroke recurrences were recorded in the Dijon Stroke Registry. Table 3 lists survival probabilities without recurrence. Thus, 1 year after a first stroke, 53.5% of patients who had a cardioembolic stroke were still alive without recurrence versus 55.8% of patients who had a cerebral or subarachnoid hemorrhage and 75% of patients in whom the mechanism of the first stroke was of another origin.
Constant Prevalence Model
Results (including CIs) are presented in table 4. Prevention with D150 would allow to avoid 10,841 (95% CI: 2,868-19,323) strokes (not counting recurrences). Although the model predicts also fewer strokes for D110 (2,443; 95% CI: -5, 852 to 10,895), this difference was not statistically significant, the CI including zero.
The model predicts an excess of 1,332 major hemorrhages when patients are treated with D150 (as D110 avoids 8,395 major hemorrhages) and an excess of 3,728 and 3,601 myocardial infarctions with D110 and D150, respectively. The number of deaths avoided is 6,770 and 10,565 with D110 and D150, and the number of years of life saved is 18,879 (95% CI: -18, 886 to 55,224) and 27,236 (95% CI: -2, 551 to 67,005).
Overall, the cumulative number of morbid events, a net unweighted measure of the clinical benefit of treatment, reaches 7,136 for D110 in a period of 5 years and 6,062 in the same period for D150 (for results with CI, see table 4). When deaths are included, the results favor D150 (10,012 vs. 13,484).
Dynamic Prevalence Model
Considering incidence during 5 years increases the number of events. The number of strokes avoided becomes respectively 3,299 and 14,607 with D110 and D150. This benefit is significantly positive with D150. The model predicts an excess of 1,651 major hemorrhages when patients are treated with D150 (as D110 avoids 11,307 major hemorrhages) and an excess of 4,963 and 4,781 myocardial infarctions, respectively, with D110 and D150. The number of deaths avoided is 9,191 and 14,085 with D110 and D150, and the number of years of life saved is 22,674 (95% CI: -22,693 to 66,201) and 33,091 (95% CI: -3,227 to 80,125).
Overall, the cumulative number of morbid events reaches 9,674 for D110 in a period of 5 years and 8,360 in the same period for D150 (for results with CI, see table 4). When deaths are included, the results favor D150 (13,620 vs. 18,186).
Our model predicts that, relative to each dose, dabigatran has the potential to avoid between 13,620 and 18,186 events over a 5-year period. A gross extrapolation of such numbers to the European Union population, assuming similar prevalence of AF in all countries, leads to a range of 146,000-179,000 morbid and fatal events avoided.
The unweighted sum of morbid events alone can be considered as a crude measure of the net clinical benefit of dabigatran versus warfarin. It is more extensive than the published calculations of the net clinical benefits of novel anticoagulants [11,12], which only consider the balance between ischemic strokes and intracranial hemorrhages, whereas we include major hemorrhages and myocardial infarction. When fatality is not taken into account, D110 appears to be superior to D150, due to a lower number of major hemorrhages than warfarin and D150. If mortality rates are considered, D150 is dominant. This result is strengthened by the results of the RE-LY study: D150 demonstrated a significant reduction for vascular death and a nearly significant difference for overall deaths. Estimation of the number of deaths must nevertheless be taken cautiously, considering the large CIs around the point estimates.
Our choice of a 5-year period did not allow us to compare our predictions with other published models, which all use a lifetime horizon. As such, it does not allow to compute a cost-effectiveness ratio. Nevertheless, it is plausible that results in France would be close to results published in countries with comparable daily cost of treatment of dabigatran, which is the case for Canada and the UK. All things equal elsewhere, the price of dabigatran is an important driver of the cost-effectiveness ratio. In France, dabigatran has obtained a flat daily price for both dose of EUR 2.53, whereas the price reported in cost-effectiveness studies in the UK [13,14] was EUR 3.15 (GBP 2.52) and EUR 2.52 (CAD 3.20) in Canada . In the three published cost-effectiveness studies of dabigatran in the UK and Canada, the point estimate incremental cost-effectiveness ratios were under the usually accepted thresholds, with the exception of the subgroup of patients with a well-controlled INR under warfarin in Pink et al. .
Our model presents the following limitations. Our estimations are based on the efficacy results of the RE-LY, which may not be reproducible in real life. As stressed by Pengo et al. , the absence of testing for anticoagulation, which alleviates the burden of monitoring for patients, may also lead to lower compliance and persistence. Moreover, poor compliance may put patients at risk because of the rapid reversibility of anticoagulation versus VKAs. Conversely, our model assumes that patients under warfarin would have the same quality of control as in the trial, which is not the case in France: according to a transversal survey performed by the French Drug Agency on the observed results of INR testing, only 46.1% were in the range from 2 to 3 .
Mortality rates were not adjusted for age. Indeed, mortality, from all causes combined, increased with age but cases of stroke were also more frequent in older subjects. We applied the mortality rate observed in the RE-LY study (mean patient age: 71.6) and in the stroke registry (mean age of cardiocerebral and hemorrhagic stroke patients: 77) to older patients. The mean age of the dabigatran target population, in the French cohort, was 77.9 years. The published clinical trial data did not permit taking this difference into account, since results according to age were not reported.
Patients in the dabigatran target cohort were at higher risk than patients recorded in the Dijon Stroke Registry, since this registry did not identify patients with AF. Several studies have demonstrated that patients with AF have a higher risk of stroke than the general population. In the Framingham Heart Study, patients with AF had a 5 times greater risk of stroke than other patients . Furthermore, strokes were severer in patients with AF . They were fatal in almost twice as many cases, and morbidity was also higher. Lastly, severity of stroke increased with age: in patients over 75 years old, half of the stroke cases were severe or fatal. After 1 year of follow-up, 63% of patients with AF who had a stroke died, versus 34% in other patients, and 23% had a recurrence of stroke versus 8%. In the FINMONICA Study  which included 7,000 patients who had a first ischemic stroke, patients with AF had a risk of death 1.26 times greater than other patients. Similarly, in the European Stroke Project , mortality 28 days after a first acute stroke was higher in patients with AF (19.1 vs. 12%). Patients with AF also had a higher risk of recurrence . The rate of recurrence and death after a stroke in the model is probably underestimated.
On the other hand, assuming that the occurrence of myocardial infarction did not increase the risk of subsequent events overestimates the benefits of dabigatran, since the simulation resulted in an excess of events as compared to warfarin, although this difference was not statistically significant.
The objective of our study was to provide French decision-makers with an estimation of the number of events avoided in patients treated with dabigatran versus warfarin over a fixed period of time, in order to provide them with expected outcomes. Our results show that for both the 110- and 150-mg doses, the cumulative number of morbid events avoided versus warfarin is positive (but not significant) and the lower limit of the CI around the number of stroke avoided versus warfarin is positive, thus providing treated patients with a clinical benefit.
The study was supported by a grant from Boehringer Ingelheim.
Chair of Health Economics and Management
ESSEC Business School, Avenue Bernard Hirsch
PO Box 50105, FR-95021 Cergy-Pontoise Cedex (France)
Copyright: All rights reserved. No part of this publication may be translated into other languages, reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording, microcopying, or by any information storage and retrieval system, without permission in writing from the publisher or, in the case of photocopying, direct payment of a specified fee to the Copyright Clearance Center.
Drug Dosage: The authors and the publisher have exerted every effort to ensure that drug selection and dosage set forth in this text are in accord with current recommendations and practice at the time of publication. However, in view of ongoing research, changes in government regulations, and the constant flow of information relating to drug therapy and drug reactions, the reader is urged to check the package insert for each drug for any changes in indications and dosage and for added warnings and precautions. This is particularly important when the recommended agent is a new and/or infrequently employed drug.
Disclaimer: The statements, opinions and data contained in this publication are solely those of the individual authors and contributors and not of the publishers and the editor(s). The appearance of advertisements or/and product references in the publication is not a warranty, endorsement, or approval of the products or services advertised or of their effectiveness, quality or safety. The publisher and the editor(s) disclaim responsibility for any injury to persons or property resulting from any ideas, methods, instructions or products referred to in the content or advertisements.