Comparison of differences in medical costs when new oral anticoagulants are used for the treatment of patients with non-valvular atrial fibrillation and venous thromboembolism vs warfarin or placebo in the US

Abstract

Objective:

Medical costs that may be avoided when any of the four new oral anticoagulants (NOACs), dabigatran, rivaroxaban, apixaban, and edoxaban, are used instead of warfarin for the treatment of non-valvular atrial fibrillation (NVAF) were estimated and compared. Additionally, the overall differences in medical costs were estimated for NVAF and venous thromboembolism (VTE) patient populations combined.

Methods:

Medical cost differences associated with NOAC use vs warfarin or placebo among NVAF and VTE patients were estimated based on clinical event rates obtained from the published trial data. The clinical event rates were calculated as the percentage of patients with each of the clinical events during the trial periods. Univariate and multivariate sensitivity analyses were conducted for the medical-cost differences determined for NVAF patients. A hypothetical health plan population of 1 million members was used to estimate and compare the combined medical-cost differences of the NVAF and VTE populations and were projected in the years 2015–2018.

Results:

In a year, the medical-cost differences associated with NOAC use instead of warfarin were estimated at −$204, −$140, −$495, and −$340 per patient for dabigatran, rivaroxaban, apixaban, and edoxaban, respectively. In 2014, among the hypothetical population, the medical-cost differences were −$3.7, −$4.2, −$11.5, and −$6.6 million for NVAF and acute VTE patients treated with dabigatran, rivaroxaban, apixaban, and edoxaban, respectively. In 2014, for the combined NVAF, acute VTE, and extended VTE patient populations, medical-cost differences were −$10.0, −$10.9, −$21.0, and −$21.0 million for dabigatran, rivaroxaban, 2.5 mg apixaban, and 5 mg apixaban, respectively. Medical-cost differences associated with use of NOACs were projected to steadily increase from 2014 to 2018.

Conclusions:

Medical costs are reduced when NOACs are used instead of warfarin/placebo for the treatment of NVAF or VTE, with apixaban being associated with the greatest reduction in medical costs.

Key words::

• Economic model
• Healthcare costs
• New oral anticoagulants
• Non-valvular atrial fibrillation
• Venous thromboembolism

Introduction

Non-valvular atrial fibrillation (NVAF), the most common significant cardiac rhythm disorder, and venous thromboembolism (VTE), including deep vein thrombosis (DVT) and pulmonary embolism (PE), affect several million people in the US1–3. Both diseases are associated with significant morbidity and substantial healthcare resource utilization and costs4–7. NVAF-related strokes account for ≥15% of all strokes (>100,000 strokes annually) in the US4. It was estimated in 2008 that the annual direct medical cost of NVAF was $6 billion for NVAF-related costs only and $26 billion when including other cardiovascular and non-cardiovascular costs8. VTE is the number one cause of unexpected hospital death and was estimated in 2011 to annually cost between $13.5–$27.2 billion in the US9^,10.

NVAF and VTE are commonly treated with anti-coagulation therapy11^,12. The predominately used anticoagulant treatment for NVAF is warfarin and that for VTE is low molecular weight heparin (LMWH) followed by a vitamin K antagonist (e.g., warfarin)11^,12. These anticoagulants have several limitations including that warfarin has a slow onset of action and requires frequent monitoring and dose adjustments and LMWH is administered parenterally, which can compromise adherence and increase its costs of use13. Recently, the new oral anticoagulants (NOACs), dabigatran, rivaroxaban, and apixaban, have been approved by the FDA for treatment of NVAF and VTE. Currently, the NOAC, edoxaban, is undergoing FDA review for the indications of NVAF and VTE.

The randomized clinical trials, RE-LY, ROCKET-AF, ARISTOTLE, and ENGAGE-AF-TIMI 48, have demonstrated that the NOACs, dabigatran, rivaroxaban, apixaban, and edoxaban, are efficacious alternatives to warfarin for stroke prophylaxis in NVAF patients14–17. A recent meta-analysis of the four NOAC vs warfarin trials, including 71,863 trial participants, reported that NOAC use was associated with a 19% reduction in stroke and systemic embolic events and a 10% reduction in all-cause mortality compared with warfarin18. The clinical trials, RE-COVER and RE-COVER II, EINSTEIN-DVT and EINSTEIN-PE, AMPLIFY, and the Hokusai-VTE trial, demonstrated that each of the NOACs, dabigatran, rivaroxaban, apixaban, and edoxaban, are non-inferior to warfarin for treatment of acute symptomatic VTE without increasing the risk of major bleeding events19–25. Additionally, the RE-SONATE, EINSTEIN-EXT, and AMPLIFY-EXT trials showed that dabigatran, rivaroxaban, and apixaban significantly reduce the risk for recurrent acute symptomatic VTE vs placebo treatment among VTE patients treated for extended periods26–28. The NOACs have not only been shown to be efficacious for the treatment of NVAF and VTE, but offer pharmacologic advantages over other anti-coagulation therapies in that they are orally administered, have a rapid onset of action, few drug interactions, and predictable pharmacokinetics, thereby eliminating the need for monitoring29.

We have previously published that, based on clinical trial results, medical costs are reduced for patients treated with any of the NOACs vs warfarin for the indications of NVAF and acute VTE and also for VTE patients treated for extended periods with NOACs vs placebo30–32. However, edoxaban was not included as a comparator in the previous study since the edoxaban trial data were not available at the time. Therefore, in the current study we estimated and compared the medical costs from a US payer perspective that may be potentially avoided when any of the four NOACs are used instead of warfarin for the treatment of patients with NVAF. Additionally, we estimated the overall medical costs avoided for NVAF and VTE patient populations combined among a hypothetical cohort of health plan members.

Methods

Estimation of clinical event rates among NVAF patients

The clinical events evaluated among NVAF patients were ischemic or uncertain type of stroke (IS), hemorrhagic stroke (HS), systemic embolism (SE), myocardial infarction (MI), pulmonary embolism (PE), deep vein thrombosis (DVT), major bleedings excluding hemorrhagic stroke (MBEHS), clinically relevant non-major bleedings (CRNMB), and other minor bleeding events. HS was considered both as an efficacy end-point and a safety end-point in the NOAC vs warfarin clinical trials. In order to avoid double counting, HS events in this cost analysis were excluded from major bleeding, but kept as a component of efficacy end-points.

The rates of clinical events among NVAF patients treated with warfarin and NOACs were based on published data obtained from each NOAC vs warfarin comparison clinical trial and calculated as the percentage of patients with each of the clinical events during the trial periods13–17. There were two edoxaban treatment arms in the ENGAGE-AF-TIMI trial, one of NVAF patients treated with 30 mg edoxaban once daily and the other of NVAF patients treated with 60 mg edoxaban once daily17. Since the lower doses of the other NOACs are recommended for special populations of NVAF patients (e.g., renally impaired), for consistency only the edoxaban 60 mg dosing regimen arm was analyzed in the economic model.

Estimation of medical cost differences associated with use of NOACs vs warfarin among NVAF patients

Incremental annual costs, defined as the incremental costs to a US health payer of a NVAF patient experiencing a clinical event during 1 year following the initial event, were obtained from the literature33–35. The incremental annual costs of a CRNMB and other minor bleeding events were estimated using a similar approach as previously published and is based on the cost of office visits obtained from the Medicare Fee Schedule, Payment and Reimbursement Benefit Guideline30–32^,36. All costs were inflation adjusted to 2013 cost levels via the CPI Medical Care Index37. Based on the absolute risks determined for each of the clinical events, the difference in total medical costs from a US payer perspective associated with each NOAC vs warfarin among NVAF patients was determined. Our study focused on the medical cost differences driven by clinical outcomes, with drug costs and other additional monitoring related expenses not included in this analysis.

Sensitivity analyses of medical cost difference estimates among NVAF patients

Univariate (one-way) sensitivity analyses were conducted to determine the effects of varying a single clinical event rate or the corresponding incremental cost on the medical cost differences associated with NOAC vs warfarin use. Clinical event rates were varied between the ranges of their respective 95% confidence intervals and the incremental cost estimates were varied ± 30%. A new medical cost difference estimate from the model was generated when the lower end or upper end estimates of a single variable was used in the economic model. Such univariate sensitivity analysis is used to help evaluate the sensitivity of the overall medical cost difference with the variation of a single model parameter variable.

Since these variables are often interdependent, Monte Carlo analyses were also conducted as multivariate sensitivity analyses. Gaussian distributions were assumed for the estimates of clinical event rates and the corresponding cost estimates. For each cycle of a Monte Carlo analysis, the values of the clinical event rates were taken randomly from a Gaussian distribution of the mean and standard deviation of the corresponding variables. Ten thousand such Monte Carlo iterations were conducted for each NOAC vs warfarin comparison. Descriptive statistics of the total medical cost differences were measured from the results of such 10,000 random Monte Carlo cycles. The 95% confidence intervals of the mean medical cost differences were evaluated as the range between the 2.5–97.5 percentile of medical costs evaluated from the 10,000 random cycles of Monte Carlo simulation for each NOAC vs warfarin comparison.

Estimation of medical cost differences among combined population of NVAF and VTE patients

Medical cost differences associated with NOAC use for patients treated for acute VTE and patients treated for VTE for extended periods were obtained from previous publications and were based on results of randomized phase III clinical trials31^,32. The treatment arms of the acute VTE trials were dabigatran vs warfarin, both with initial treatment with a parenteral anticoagulant, rivaroxaban vs enoxaparin with warfarin/acenocoumarol, apixaban vs enoxaparin followed by warfarin, and edoxaban vs warfarin, both with initial treatment with enoxaparin or unfractionated heparin19–25. The control arms of the extended VTE trials were all placebo treatment26–28. To estimate the medical costs differences of the NOACs in previous publications, the rates of clinical events among VTE patients treated with warfarin, placebo, and NOACs were calculated as the percentage of patients with each of the clinical events during the trial periods and based on published trial data20^,23–28. A hypothetical health plan population in the US with 1 million members was used to estimate and compare the NVAF and VTE combined medical cost differences for patients treated with NOACs vs warfarin/placebo. Prevalence rates of NVAF and VTE were derived from published literature38–40. The same usage rate (i.e., 10% for NVAF and acute VTE, 5% for VTE-extended) for each NOAC was assumed in order to facilitate the comparison of the combined medical cost differences among NOACs. The usage rates represented approximate proportions of patients who currently receive NOACs for anticoagulation therapy. Based on an assumed population growth rate of 2% for the hypothetical cohort, the medical cost differences of NOACs vs warfarin/placebo were also projected and compared in the years 2015–2018.

Results

Based on the published results of clinical trials, among NVAF patients the estimated clinical event rates per year (95% confidence interval) for IS were 0.91% (0.72%, 1.18%) for dabigatran, 1.43% (1.14%, 1.78%) for rivaroxaban, 0.97% (0.78%, 1.19%) for apixaban, and 1.25% (1.04%, 1.49%) for edoxaban. Among NVAF patients MBEHS rates were estimated at 3.22% (2.84%, 3.63%) for dabigatran, 3.28% (2.90%, 3.67%) for rivaroxaban, 1.89% (1.69%, 2.12%) for apixaban, and 2.49% (2.26%, 2.76%) for edoxaban. The rates of all evaluated clinical events for the NOACs vs warfarin are reported in Table 1.

Table 1. Estimated clinical event rates (per patient-year) associated with use of new oral anticoagulants (NOACs) vs warfarin among non-valvular atrial fibrillation patients.

Drug name	Dabigatran-150 mg vs warfarin (% per patient year [CI])	Rivaroxaban vs warfarin (% per patient year [CI])	Apixaban vs warfarin (% per patient year [CI])	Edoxaban 60 mg vs warfarin (% per patient year [CI])
Trial name	RE-LY	ROCKET-AF	ARISTOTLE	ENGAGE-AF
Primary efficacy end-points
Ischemic or uncertain type of stroke	0.91 [0.72, 1.18] vs 1.20 [1.20, 1.20]	1.43 [1.14, 1.78] vs 1.52 [1.52, 1.52]	0.97 [0.78, 1.19] vs 1.05 [1.05, 1.05]	1.25 [1.04, 1.49] vs 1.25 [1.25, 1.25]
Hemorrhagic stroke	0.10 [0.05, 0.19] vs 0.38 [0.38, 0.38]	0.26 [0.16, 0.41] vs 0.44 [0.44, 0.44]	0.24 [0.16, 0.35] vs 0.47 [0.47, 0.47]	0.25 [0.18, 0.36] vs 0.47 [0.47, 0.47]
Systemic embolism	0.08 [0.06, 0.10] vs 0.12 [0.12, 0.12]	0.04 [0.02, 0.12] vs 0.19 [0.19, 0.19]	0.09 [0.04, 0.18] vs 0.10 [0.10, 0.10]	0.08 [0.04, 0.15] vs 0.12 [0.12, 0.12]
Secondary efficacy end-points
Myocardial infarction	0.81 [0.60, 1.09] vs 0.64 [0.64, 0.64]	0.91 [0.71, 1.19] vs 1.12 [1.12, 1.12]	0.54 [0.40, 0.71] vs 0.61 [0.61, 0.61]	0.71 [0.56, 0.89] vs 0.75 [0.75, 0.75]
Pulmonary embolism or deep-vein thrombosis	0.14 [0.07, 0.31] vs 0.09 [0.09, 0.09]	0.04 [0.04, 0.05] vs 0.05 [0.05, 0.05]	0.04 [0.01, 0.11] vs 0.05 [0.05, 0.05]	0.14 [0.08, 0.24] vs 0.14 [0.14, 0.14]
Safety end-points
Major bleedings-excluding hemorrhagic stroke	3.22 [2.84, 3.63] vs 3.19 [3.19, 3.19]	3.28 [2.90, 3.67] vs 2.96 [2.96, 2.96]	1.89 [1.69, 2.12] vs 2.62 [2.62, 2.62]	2.49 [2.26, 2.76] vs 3.43 [3.43, 3.43]
Clinically relevant non-major bleedings	0 vs 0	11.86 [10.94, 12.88] vs 11.40 [11.39, 11.41]	1.99 [1.78, 2.19] vs 2.92 [2.92, 2.92]	8.73 [8.02, 9.44] vs 10.15 [10.14, 10.16]
Other minor bleedings	14.90 [13.91, 15.88] vs 16.37 [16.36, 16.38]	2.35 [1.97, 4.85] vs 2.03 [2.03, 2.03]	14.05 [13.46, 14.84] vs 19.8 [19.8, 19.8]	4.11 [3.72, 4.60] vs 4.89 [4.89, 4.90]

CI, Confidence interval.

Medical cost differences among NVAF patients treated with NOACs vs warfarin

Based on previously published literature, the incremental 1-year medical costs of NVAF patients with clinical end-points relative to patients who did not have a clinical event in 2013 dollars were estimated as the following: IS = $43,89033, HS = $57,38433, SE = $21,94533, MI = $41,27734, PE/DVT = $19,53235, MBEHS = $38,45433, CRNMB (based on estimated four claims of CPT code 99215: level 5 established office patient visit) = $547, and other minor bleeding (based on CPT code 99214: level 4 established office patient visit) = $102. In a year, the total medical cost differences associated with NOAC use relative to warfarin among NVAF patients were estimated at −$204, −$140, −$495, and −$340 per patient for dabigatran, rivaroxaban, apixaban, and edoxaban, respectively (Table 2).

Table 2. Differences in medical costs among non-valvular atrial fibrillation (NVAF) patients treated with new oral anticoagulants (NOACs) vs warfarin.

Outcomes	Dabigatran-150 mg vs warfarin ($/patient-year)	Rivaroxaban vs warfarin ($/patient-year)	Apixaban vs warfarin ($/patient-year)	Edoxaban-60 mg vs warfarin ($/patient-year)
Primary efficacy end-points
Ischemic or uncertain type of stroke	−$126	−$40	−$37	$0
Hemorrhagic stroke	−$161	−$104	−$132	−$124
Systemic embolism	−$9	−$32	−$3	−$9
Secondary efficacy end-points
Myocardial infarction	$71	−$88	−$30	−$19
Pulmonary embolism or deep-vein thrombosis	$11	−$1	−$2	$1
Safety end-points
Major bleedings—excluding hemorrhagic stroke	$12	$122	−$280	−$181
Clinically relevant non-major bleedings	$0	$2	−$5	−$8
Other minor bleedings	−$1	$0	−$6	−$1
Total medical cost difference	−$204	−$140	−$495	−$340

Negative cost difference values indicate a cost reduction in comparison to warfarin.

Sensitivity analyses of medical cost difference estimates among NVAF patients

Univariate and multivariate sensitivity analyses demonstrated consistent medical cost differences associated with NOAC use relative to warfarin among NVAF patients. Figure 1 indicates how variations in clinical event rates influenced the estimated cost differences of NOACs. The clinical event rates with major impact on the estimated medical cost differences associated with dabigatran use instead of warfarin were MBEHS, MI, and IS, with the corresponding medical cost difference ranges of (−$352, −$46), (−$292, −$88), and (−$289, −$89), with negative numbers indicating a cost reduction. The clinical event rates with major impact on the estimated cost differences associated with rivaroxaban use instead of warfarin were MBEHS, IS, MI, and HS, with the corresponding medical cost difference ranges of (−$286, +$11), (−$267, +$13), (−$224, −$25), and (−$196, −$55). The clinical event rates with major impact on the estimated cost differences associated with apixaban use instead of warfarin were IS, MBEHS, MI, and HS, with the corresponding medical cost difference ranges of (−$578, −$398), (−$573, −$408), (−$550, −$422), and (−$538, −$430). The clinical event rates with a major impact on the estimated cost differences associated with edoxaban use instead of warfarin were IS, MBEHS, and MI, with the corresponding medical cost difference ranges of (−$433, −$236), (−$430, −$237), and (−$402, −$263).

Figure 1. Univariate sensitivity analyses examining the influence of variations in clinical event rates on the medical cost differences per patient associated with new oral anticoagulant (NOAC) use relative to warfarin.

Figure 2 shows how variations in the incremental medical costs of NVAF patients with events influenced the estimated medical cost differences associated with NOACs. The incremental medical costs with major impact on the estimated cost differences associated with dabigatran use instead of warfarin were HS (−$253, −$156), IS (−$242, −$166), and MI (−$226, −$183). The incremental medical costs with major impacts on the estimated cost differences associated with rivaroxaban use instead of warfarin were MBEHS (−$177, −$104), HS (−$171, −$109) and MI (−$167, −$114). The incremental medical costs with major impacts on the estimated cost differences associated with apixaban use instead of warfarin were MBEHS (−$579, −$411) and HS (−$535, −$455). The incremental medical costs with major impacts on the estimated cost differences associated with edoxaban use instead of warfarin were MBEHS (−$394, −$286) and HS (−$377, −$303).

Figure 2. Univariate sensitivity analyses examining the influence of variations in incremental costs of clinical events on the medical cost differences per patient associated with new oral anticoagulant (NOAC) use relative to warfarin.

The Monte Carlo multivariate analysis, in which each variable of the univariate analysis was allowed to vary simultaneously for 10,000 cycles, was used to further test the consistency of the medical cost difference estimates for NOACs in comparison to warfarin (Figure 3). The means of medical cost differences (with 95% confidence interval) associated with NOACs were as follows: dabigatran: −$204 (95% CI: +$51, −$463), rivaroxaban: −$139 (95% CI: +$137, −$411), apixaban: −$496 (95% CI: −$268, −$750), and edoxaban: −$340 (95% CI: −$140, −$563). Of the 10,000 random Monte Carlo simulation cycles, 94.2%, 85.1%, 100.0%, and 99.9% had a cost reduction <$0 for dabigatran, rivaroxaban, apixaban, and edoxaban, respectively.

Figure 3. Distribution of total medical cost differences per patient from 10,000 cycles of Monte Carlo simulations.

Medical cost differences among NVAF and VTE patients treated with NOACs vs warfarin/placebo

The estimated medical costs differences of NVAF and VTE patient populations among the hypothetical cohort of 1 million insured lives are shown in Table 3. In 2014, among the hypothetical population of one million insured lives, the medical costs were projected to be reduced by $3.0, $2.1, $7.3, and $5.0 million for NVAF patients treated with dabigatran, rivaroxaban, apixaban, and edoxaban, respectively; by $0.7, $2.2, $4.1, and $1.6 million for patients treated for acute VTE with dabigatran, rivaroxaban, apixaban, and edoxaban, respectively; and by $6.3, $6.6, $9.6, and $9.5 million for VTE patients treated for extended periods with dabigatran, rivaroxaban, apixaban 2.5 mg, and apixaban 5 mg, respectively. In 2014, for the combined NVAF and acute VTE patient population within the hypothetical population of one million insured lives, medical costs were projected to be reduced by $3.7, $4.2, $11.5, and $6.6 million treated with dabigatran, rivaroxaban, apixaban, and edoxaban, respectively. In 2014, for the combined NVAF, acute VTE, and extended VTE patient population within the hypothetical population of one million insured lives, medical costs were projected to be reduced by $10.0, $10.9, $21.0, and $21.0 million for dabigatran, rivaroxaban, apixaban 2.5 mg, and apixaban 5 mg, respectively. In the model, the differences in medical costs associated with use of the NOACs were projected to steadily increase in the years 2015–2018 (Figure 4).

Figure 4. Projected medical cost differences in the years 2014–2018 among non-valvular atrial fibrillation (NVAF) patients and patients treated for acute venous thromboembolism (A) and for NVAF patients, VTE patients treated for acute VTE, and VTE patients treated for extended periods (B) of a hypothetical cohort of 1 million insured lives.

Table 3. Medical cost differences among non-valvular atrial fibrillation (NVAF) patients and patients treated for acute venous thromboembolism and for VTE patients treated for extended periods of a hypothetical cohort of 1 million insured lives (2014–2018).

	Year 2014	Year 2015	Year 2016	Year 2017	Year 2018
Non-valvular atrial fibrillation (NVAF) patient population
Dabigatran	−$3,025,075	−$3,085,577	−$3,147,288	−$3,210,234	−$3,274,439
Rivaroxaban	−$2,077,411	−$2,118,959	−$2,161,338	−$2,204,565	−$2,248,656
Apixaban	−$7,325,621	−$7,472,133	−$7,621,576	−$7,774,008	−$7,929,488
Edoxaban	−$5,033,264	−$5,133,930	−$5,236,608	−$5,341,340	−$5,448,167
Venous thromboembolism (VTE) patient population—acute VTE only
Dabigatran	−$656,304	−$669,430	−$682,819	−$696,475	−$710,405
Rivaroxaban	−$2,169,134	−$2,212,516	−$2,256,767	−$2,301,902	−$2,347,940
Apixaban	−$4,131,354	−$4,213,981	−$4,298,261	−$4,384,226	−$4,471,911
Edoxaban	−$1,550,065	−$1,581,066	−$1,612,687	−$1,644,941	−$1,677,840
VTE patient population—Extended VTE treatment only
Dabigatran	−$6,286,174	−$6,411,897	−$6,540,135	−$6,670,938	−$6,804,357
Rivaroxaban	−$6,632,220	−$6,764,865	−$6,900,162	−$7,038,165	−$7,178,929
Apixaban-2.5 mg	−$9,561,024	−$9,752,244	−$9,947,289	−$10,146,235	−$10,349,159
Apixaban-5 mg	−$9,548,002	−$9,738,962	−$9,933,741	−$10,132,416	−$10,335,064
NVAF and acute VTE combined population
Dabigatran	−$3,681,379	−$3,755,007	−$3,830,107	−$3,906,709	−$3,984,843
Rivaroxaban	−$4,246,545	−$4,331,476	−$4,418,105	−$4,506,467	−$4,596,596
Apixaban	−$11,456,975	−$11,686,115	−$11,919,837	−$12,158,234	−$12,401,398
Edoxaban	−$6,583,329	−$6,714,996	−$6,849,296	−$6,986,281	−$7,126,007
NVAF, acute VTE, and extended VTE combined population
Dabigatran	−$9,967,553	−$10,166,904	−$10,370,242	−$10,577,647	−$10,789,200
Rivaroxaban	−$10,878,765	−$11,096,340	−$11,318,267	−$11,544,632	−$11,775,525
Apixaban-2.5 mg	−$21,017,999	−$21,438,359	−$21,867,126	−$22,304,468	−$22,750,558
Apixaban-5 mg	−$21,004,977	−$21,425,077	−$21,853,578	−$22,290,650	−$22,736,463

Discussion

Based on our economic analysis using clinical trial data, annual medical costs were estimated to be lower for NVAF patients taking any of the NOACs vs warfarin. Among NVAF patients, apixaban was estimated to be associated with the greatest reduction in medical costs, which is driven by medical cost reductions associated with both efficacy and safety end-points. In our economic analysis, clinical event rates reported in the original trials were used, regardless of a superiority or non-inferiority test. This is commonly done in the economic models of clinical trial findings, since clinical trials are generally not powered for the economic evaluations of trial results.

For the combined population of NVAF patients and acute VTE patients among a hypothetical cohort of 1 million insured lives medical costs are projected to be reduced in 2014 by $3.7, $4.2, $11.5, and $6.6 million for patients treated with dabigatran, rivaroxaban, apixaban, and edoxaban, respectively. For the combined population of NVAF patients, acute VTE patients, and VTE patients with extended treatment among a hypothetical cohort of 1 million insured lives, medical costs are projected to be reduced in 2014 by $10.0, $10.9, $21.0, and $21.0 million for patients treated with dabigatran, rivaroxaban, 2.5 mg apixaban, and 5.0 mg apixaban, respectively. Additionally, the reductions in medical costs associated with use of the NOACs were projected to steadily increase in the years 2014–2018. Our economic analysis was based on clinical trial data and, therefore, the direct application of the results to the real-world setting, where many other factors, including local healthcare cost, drug adherence, population risk, and health behaviors may vary, will require further assessment. However, prior to the broad exposure of NOACs to the real world, clinical trial results are considered the best basis for conducting this type of research41^,42.

In the real-world setting, stroke incidence has been observed to be higher among NVAF patients than reported in clinical trials. A real-world analysis conducted in the US on 119,764 Medicare patients with NVAF (mean age = 79.3 years) reported an annual incidence of ischemic stroke of 3.6% among patients taking warfarin and that warfarin use was only associated with a 27% risk reduction for ischemic stroke44. The results of the study by Mercaldi et al.44 suggest the incidence of ischemic stroke (3.6%) may be greater among NVAF patients treated with warfarin in the real world than observed among clinical trial NVAF patients using warfarin (1.2–1.5%)14–17^,44. Additionally, clinical event rates reported in the NOAC vs warfarin trials are only representative of first-event occurrences in 1 year and do not measure multiple events occurring in a 1-year time period.

The efficacy of warfarin for stroke prophylaxis in the real world, which is about half that observed in clinical trials, is likely the result of several factors including a wide range of drug–drug and drug–food interactions, low adherence to treatment guidelines, low adherence to warfarin anti-coagulation therapy, and fluctuation of time in therapeutic range44. In fact, Casciano et al.45 conducted a retrospective database study on 13,289 patients with NVAF (CHADS₂ score ≥2) and reported that 47% did not receive warfarin and of those treated with warfarin only 40% were adherent to therapy. A second larger retrospective database study of 171,393 NVAF patients observed similarly that ∼42% of NVAF patients with high stroke risk received warfarin and the percentage of patients that received warfarin did not widely differ among patients at high, moderate, and lower stroke risk as assessed by CHADS₂ score46. Furthermore, this study reported that only about a third of NVAF patients at any stroke risk received uninterrupted warfarin therapy for 6 months46. Since routine anti-coagulation management and low adherence levels to warfarin were not taken into account in our study, the medical cost differences associated with NOAC use relative to warfarin may be much more substantial when examined in the real world, although again further study is required in this regard.

The fact that warfarin use increases the risk for major bleeding events is of great concern, especially among NVAF patients with a high risk of bleeding. The average absolute risk for major bleeding events in clinical trial patients using warfarin ranged between 2.6–3.4%14–17. In the real world major bleeding events can be higher, and exceptionally so in patients of greater age, and with greater comorbidities, and in those not receiving treatment through anti-coagulation clinics44^,47. Major bleeding event rates for the 70,057 Medicare patients taking warfarin in the Mercaldi et al.44 study were 7.6%. Annual rates of major bleedings have been reported to be substantially greater in routine medical clinics, averaging 10.9% in comparison to anti-coagulation clinics, which average 2.8%47. The probable increased risk for major bleedings in the real world may imply that NOACs with lower bleeding risk may fare even better in the real world, where warfarin’s time in therapeutic range is frequently lower than in clinical trials48.

Similar to NVAF patients, rates of clinical events related to VTE may also be higher among VTE patients in real-world settings. In the acute VTE NOAC clinical trials, the incidences of recurrent VTE and MB were <3% and <2%, respectively19–25. The Worcester VTE study, which had a mean follow-up period of 1216 days of 1567 persons with first-time VTE reported cumulative rates of ∼16% for recurrent VTE and 12% for MB49. The Worcester VTE study also noted that the cumulative incidence of MB was 8% in the month following the initial VTE event49. Studies with longer follow-up periods have reported 5-year cumulative incidence rates of recurrent VTE ranging between 21.5–29.1%50^,51. The potentially greater event rates of recurrent VTE and MB in real-world settings implies the differences in medical costs between VTE patients who use NOACs vs warfarin/placebo estimated in this study, which were based on clinical trial data, are likely under-estimated, thus making the impact of the cost differences even greater in favor of the NOACs.

Dabigatran, rivaroxaban, and apixaban are priced similarly with a wholesale acquisition cost in the US of $8.35 per day52. Edoxaban is currently not marketed in the US yet, but may likely be priced similar to the other NOACs. Thus, the drug prices of NOACs are not likely to affect the relative cost differences among the four NOACs. Warfarin for NVAF and VTE treatment in comparison to NOACs is less expensive when only drug costs are taken into account. However, the avoidance of NVAF- and VTE-related clinical events and, hence, lower healthcare costs should assist clinicians and health policy-makers in determining the most cost-effective pharmacotherapies. Having the medical costs for clinical events avoided as a component of overall drug cost is highly relevant given the fact that hospitalizations and physician/clinical services account for 51% of all US health expenditures53. Moreover, it is important to balance optimizing the quality of care and, hence, patient outcomes with reducing pharmacotherapy costs as this is a central strategy of healthcare reform for improving healthcare quality and driving clinical practice innovation.

Limitations

This economic analysis has limitations in that monitoring related expenses, as well as the long-term burden of clinical events, indirect costs, and quality-of-life, all of which may be impacted by more efficacious pharmacotherapy for NVAF and VTE were not taken into account. Further studies incorporating all of these costs, as well as drug costs, will provide valuable information for healthcare providers, patients, and health policy-makers and will likely be best accomplished using a real-world observational study design. The incremental annual medical costs for patients with clinical events were based upon those obtained from published studies, which estimated costs from different US health insurers. Therefore, the medical costs are generalized and may not apply to specific payers of US health plans and also healthcare systems outside of the US. Since rates of stroke, VTE recurrence, and MB may be greater in real-world settings43^,44^,49–51, the estimated medical cost differences associated with NOAC use relative to warfarin/placebo may be under-estimated. Thus, the direct application of the results to routine clinical practice, where many factors including population risk may vary, will require further assessment.

Conclusions

When any of the NOACs are used instead of warfarin/placebo for NVAF and VTE treatment medical costs are reduced, with apixaban being associated with the greatest reduction in medical costs. The estimated medical cost reductions associated with NOAC use, relative to warfarin/placebo, may be helpful in determining the overall cost impacts of the use of NOACs among NVAF and VTE patients in the US, although further evaluation may be needed to validate these results in the real-world setting.

Transparency

Declaration of funding

Sponsorship for this study was provided by Bristol-Myers Squibb and Pfizer.

Declaration of financial/other relationships

AA is a consultant for Novosys Health, which received funding from Bristol-Myers Squibb and Pfizer in connection with conducting this study and development of this manuscript. AB is an employee of Bristol-Myers Squibb and owns stock in the company. JT is an employee of Pfizer and owns stock in the company. JL and MLS are also employees of Novosys Health. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.