Approximately one-third of patients with epilepsy have drug-resistant epilepsy (DRE) [1]. Patients with DRE require more health care resources [e.g., hospitalizations, emergency department (ED) visits] and are more costly to manage than patients with epilepsy treated successfully with anti-epileptic drugs (AEDs) [2]. Treatment options for patients with DRE include ketogenic diet, resective brain surgery to remove the focal area responsible for the epileptic seizures, severing the corpus callosum that connects the two hemispheres of the brain, responsive neurostimulation, and vagus nerve stimulation (VNS; VNSTherapy®; LivaNova; Houston, Texas]) [3, 4]. VNS is reimbursed in over 50 countries including France [5] and the UK [6]. VNS was approved by the Food and Drug Administration (FDA) in 1997 as an adjunctive treatment for patients 12 years or older with partial-onset seizures that are not responsive to AEDs [7]. In 2017, the FDA expanded the indication to include patients as young as 4 years old [8]. Over 100,000 patients with DRE have been managed globally with VNS since 1989 [9]. A recent Cochrane Review found VNS to be “an effective and well-tolerated treatment” (p. 2) based on evidence from five clinical trials [10]. In addition, the American Academy of Neurology found VNS to be a safe and effective treatment for epilepsy [11]. Despite the long-term proven effectiveness of VNS [12], the time to initial implant occurs approximately 20 years after initial diagnosis (Janszky et al. [13]; Cyberonics, data on file, 1997) and after an average of 7 AEDs have been tried [14]. VNS has been shown to reduce health care resource use [15, 16] and costs [16] compared with AEDs alone.

The objective of this research was to estimate, from the perspective of a United States (US) health care payer, the budget impact and population health outcomes that may be expected with expanded use of VNS among patients with DRE aged ≥ 12 years with partial-onset seizures.


An Excel-based model was developed to compare the costs of continued AED treatment with the costs of VNS in addition to AED treatment in patients with DRE. This article is based on previously conducted studies and does not contain any studies with human participants or animals performed by any of the authors.

Patient Population

The number of eligible patients was calculated to be 1,536, based on an assumed 1-million-member health plan, with 84% of the plan population being ≥ 12 years of age [17], 0.94% having epilepsy [18], 57% of those with epilepsy having partial-onset seizures [19] and 36% of those with partial-onset seizures having DRE [1]. In addition, the analysis assumed that 95% of patients ≥ 12 years old with partial-onset seizures who have DRE currently do not have VNS. Although DRE has been defined by the International League Against Epilepsy as “a failure of adequate trials of two tolerated and appropriately chosen and used AED schedules (whether as monotherapy or in combination) to achieve sustained seizure-freedom” [20; p. 1073], the duration of epilepsy for patients who undergo VNS device placement has been reported to be approximately 20 years (Cyberonics, data on file, 1997; Janszky et al. [13]), suggesting that the percentage of patients with DRE without VNS is high.

Cost Inputs

Model inputs include the costs associated with the VNS device, placement, programming, and battery changes; adverse events due to VNS (cough, voice alteration, device removal resulting from surgical site infection); and AEDs (Table 1); and seizure-related costs. All costs are presented in 2016 US dollars.

Table 1 Model inputs: costs (per person) for VNS (device- and adverse event-related) and AEDs, with and without VNS

Seizure-Related Costs

Seizure-related costs for hospitalizations, ED visits, and neurologist visits are calculated by estimating expected levels of seizure frequency for patients without VNS and for patients with VNS using a previously published model structure [21]. The health states, as described in a cost–utility analysis of an AED [22], are based on monthly seizure frequency, and are seizure-free, ≤ 1 seizure per month, > 1 to < 10 seizures per month, and ≥ 10 seizures per month. All patients begin in the health state “≥ 10 seizures per month” based on the reported baseline weighted median number of seizures per month (16.3) in the VNS clinical trial E05 (Cyberonics, data on file, 1997). The allocation of patients by health state at 3 months was based on a post hoc analysis of the clinical trial data [21]. From that point, every 3 months to month 24, VNS patients transition to a health state determined by results from a systematic review of the literature and the VNS Therapy Patient Outcomes Registry [12]. Based on Englot et al. [12], the model assumes that no further changes in seizure frequency occur beyond month 24. Patients in the model that receive AEDs alone do not improve over time. A randomized study of patients receiving AEDs alone or VNS plus AEDs found, at 12 months, no change in seizure frequency for patients that received AEDs alone [23]. The model includes seizure-related resource utilization estimates for hospitalizations, ED visits, and neurologist visits by seizure frequency, which are multiplied by the corresponding unit costs to calculate the annual resource costs by seizure frequency (Table 2). The resulting seizure-related costs without and with VNS are shown in Table 3.

Table 2 Model inputs, seizure-related costs: annual resource utilization and costs (per person), by resource and seizure frequency
Table 3 Calculated model inputs: seizure-related costs (per person) without and with VNS therapy, by resource and year


In a modeled analysis of VNS placed in eligible patients at year 1, VNS resulted in an estimated net cost savings due to expected reductions in seizure frequency beginning the second year after VNS device placement (Fig. 1; Table 4). The initial cost of the VNS device, placement, and programming was offset in 1.7 years after device placement (Fig. 2). On average, VNS resulted in an estimated net cost savings of $77,480 per patient over 5 years, which is a 21.5% reduction in costs compared with AEDs alone. Patients with VNS had an estimated reduction in costs associated with seizure frequency of $127,554 per patient over 5 years compared with patients with AEDs alone. Seizure-related hospitalizations were the main cost driver, resulting in an estimated cost reduction of $118,925 per patient over 5 years for patients with VNS compared with AEDs alone (Fig. 3). A one-way sensitivity analysis was conducted, which varied by ± 20% the eligible patient population, VNS costs, adverse event costs, AED costs, and individual per-person resource costs per year with and without VNS. Results were most sensitive to per-person hospitalization cost per year with and without VNS in the third to fifth years after VNS device placement; however, VNS remained cost saving over 5 years (Fig. 4).

Fig. 1
figure 1

From Purser et al. [37]

Percentage of patients by number of seizures per month. SF seizure-free, VNS vagus nerve stimulation

Table 4 Results: total organization budget impact, by year
Fig. 2
figure 2

From Purser et al. [37]

Cumulative costs for patients without and with vns with break-even analysis. VNS vagus nerve stimulation

Fig. 3
figure 3

Total plan budget impact by cost category by year. AED anti-epileptic drug, ED emergency department, VNS vagus nerve stimulation

Fig. 4
figure 4

One-way sensitivity analysis tornado diagram. AED anti-epileptic drug, VNS vagus nerve stimulation

The International Society for Pharmacoeconomics and Outcomes Research budget impact analysis good practices task force recommends uncertainty be explored through scenario analyses rather than through standard approaches such as probabilistic sensitivity analysis [24]. Two scenario analyses were conducted. The first scenario increased the number of patients who are seizure-free at 24 months from 8 to 15.4%. This scenario is supported by Elliott et al. [27], who reported that 15.4% of patients were seizure-free for at least 2 years prior to the last follow-up visit in a study of 65 consecutive patients with at least 10 years of follow-up. This scenario resulted in a cost reduction over 5 years of 22.5% with VNS compared with AEDs alone compared with the cost reduction of 21.5% observed in the base case analysis. A second scenario analysis was conducted that assumed the percentage of patients in the ≥ 10 seizures per month category at 3 months who moved to having < 10 seizures per month at 24 months was 42% rather than the base case assumption of 20%. This change results in 63% of patients having < 10 seizures per month at 24 months. This scenario is supported by a study by Englot et al. [12], who reported that 63% of patients with VNS responded to treatment at 24–48 months, with response defined as a ≥ 50% reduction from baseline seizure frequency (which the current model assumes results in < 10 seizures per month). This scenario resulted in a 28.9% cost reduction with VNS over 5 years, with a break-even point of 1.54 years.


The findings from this modeled economic analysis are consistent with previously published analyses. Helmers et al. [16] analyzed medical data from 1,655 patients and estimated total costs [inpatient, outpatient, ED, and drug (AED and non-AED)] for 6 months pre-VNS and up to 3 years post-VNS initiation. The study found that the cost of VNS (device, placement, and outpatient visits) was outweighed by the cost savings at about 1.5 years, which is similar to the finding from the current modeled analysis, which estimated complete cost offset to occur 1.7 years after device placement.

Helmers et al. [16] also reported the incidence for seizure-related hospitalizations before VNS and for the 3 years post-VNS initiation. The incidence for seizure-related hospitalizations was reduced by approximately 50% by the end of the second year post-VNS, and remained at this level for year 3. Bernstein et al. [15] analyzed Kaiser Permanente administrative data to evaluate the resource utilization of patients 1 year prior to VNS and up to 4 years post-VNS initiation. Bernstein et al. [15] also found that, at 4 years post-VNS initiation, hospitalization admissions were reduced by 70%, ED visits were reduced by 99%, and outpatient visits were reduced by 91% compared with the year prior to VNS implantation.

In comparison, the current VNS modeled analysis estimated, by year 5, reductions in hospitalizations by 40%, ED visits by 39%, and neurologist visits by 10% with VNS versus without VNS. The current analysis may be more conservative than previously published analyses because it does not include the potential reductions in resource use that patients with a baseline of 10 or more seizures per month may have with VNS. For example, a study by Boon et al. [25], which considered 15 patients before and after VNS initiation, reported that, even among patients who continued to have 10 or more seizures per month after VNS initiation, there was a decrease in hospital admissions and direct medical costs compared with before VNS initiation.

A recent study of 1,795 patients with epilepsy reported that, if patients were not seizure-free after the first AED, the second and third regimens provided a likelihood of 11.6 and 4.4%, respectively, of providing seizure-freedom [26]. The study noted that, even with introduction over the past 2 decades of more than a dozen new AEDs with various mechanisms of action, the probability of obtaining seizure-freedom is dramatically reduced with each subsequent treatment [26]. However, patients have tried an average of 7 AEDs before VNS [14], suggesting that many patients may be able to benefit from earlier treatment with VNS.

The current model does not consider costs of removing the VNS device for reasons other than infection immediately following implantation. Elliot et al. [27] reviewed 436 consecutive patients that received VNS, and reported that 17% of patients had the device removed for reasons including lack of efficacy, planned magnetic resonance imaging, infection, AED success, and vocal cord paralysis; mean time to removal was 40.4 months from time of initial device placement. In addition, although VNS has been reported to reduce seizure severity [28], the model did not consider a reduction in costs and resource use due to a reduction in seizure severity.

The results of the current analysis, which used literature-based estimates, are consistent with and perhaps conservative compared with other published studies of the cost and resource use reductions that may be expected with VNS.

The US Food and Drug Administration (FDA) recently approved an age expansion to include patients as young as 4 years in the indication for VNS as an adjunctive therapy in reducing the frequency of seizures in patients with partial-onset seizures that are refractory to anti-epileptic medications [8]. VNS in younger patients has been shown to be at least as effective as the currently modeled patient population (≥ 12 years old) [29, 30]; therefore, it would be expected that expanding the population to include patients aged 4 or more years would result in additional cost savings. Additional research is needed to determine the budget impact of expanding the population.


VNS is a proven intervention that offers a long-term solution for patients with DRE by reducing seizure frequency, which leads to lower resource utilization and lower costs.