Introduction

According to the World Health Organization (WHO) reports, epilepsy is one of the three most prevalent neurological diseases internationally, affecting 1% of the world’s population [1]. Thirty percent of people suffering from epilepsy do not achieve significant improvement with the current antiepileptic drugs, not even through their combinations, and require other complementary treatments [1, 2]. This disease variant is known as drug resistant epilepsy or refractory epilepsy. A small percentage of these patients reach a better condition through surgery, but they still require complementary treatments and have to meet certain requirements to undergo surgery.

Vagus nerve stimulation (VNS) is one of the most successful complementary therapeutic methods for the treatment of refractory epilepsy today [3–5]. There are two approaches for this therapy, one based on active implanted devices, which was the first used stably, known as VNS based on active implanted devices (iVNS), and another based on external devices, called transcutaneous electrical VNS (tVNS), which popularity has increased in recent years [5]. The automatism of the iVNS devices and their proven effectiveness are important advantages, but their weaknesses are associated with the risks of the required surgery and their manufacturing and implanting costs. The tVNS-based approach excludes the surgery risks because it is an external device and the cost drops down because no special materials or high-specialized technologies are required for manufacturing. At the same time, several researching teams have reported similarity in therapy effectiveness for both approaches [5, 6].

Therapy based on tVNS has had multiple applications in recent years. Kim et al. [7] report twelve investigations about tVNS applied as complementary therapy in subjects suffering from refractory epilepsy. All these studies have had satisfactory results, with sensitivity greater than 60%, but a well-defined standard is not been established for the value of the stimulation parameters.

For the tVNS application in a public health scheme, providing therapeutic services to large populations, it is necessary to minimize costs without giving up the effectiveness of the treatment, so iVNS is not appropriate for this purpose [8]. A better solution is the tVNS because devices are cheaper, surgery is not required and the therapeutic effects have been demonstrated. Thus, the use of iVNS devices is reserved for those patients with greater economic resources [7, 9].

While other authors have been focused on the evaluation of therapeutic devices, in this document two topics are presented. Firstly, a platform to provide a therapeutic service to people suffering from refractory epilepsy is described. It is composed by a novel low-cost tVNS device and two complementary software subsystems, all developed by the authors. Artificial intelligence techniques, such as neural networks and statistical inference, will be applied to predict the treatment effectiveness for each patient. Secondly, the pilot trial carried out to evaluate a therapeutic device, developed by the authors, is explained and the outcomes are discussed. Currently, a first version of the proposal is running at the institute of neurology and neurosurgery of Havana and its growth and expansion to other hospital centers is expected.

Materials and methods

A proposed solution

The proposed solution is a three-layer system as shown in Figure 1. The low-level layer, or patient layer, is composed by the therapeutic devices assigned to each patient. These devices are for personal use and each patient will apply the therapy at home by himself, following the medical indications. Previously, the patient and his companion were trained to use the device and be able to fill out the crisis and adverse effects diary, supported by International League Against Epilepsy (ILAE). This document is, a follow-up tool to quantify the therapy impact on the patient’s evolution. Furthermore, this document has been used in previous reported researches about the same theme, so an outcome comparison can be set with the described proposal.

Display full size

Figure 1. 

Representation of the proposed system

The middle layer is a Windows application developed by the authors, named hospital data collector (HDC). Its main objectives are the following:

• 1)

  Enrollment of patients: the general data of each patient and the main information of the clinical record are input and stored in a local database.

• 2)

  Setup of therapeutic devices: each device is connected to this Windows application through a universal serial bus (USB) port to set the therapy parameters, that is, the duration and daily rate of the therapeutic sessions as well as the width, polarity, amplitude and frequency of the therapeutic pulses.

• 3)

  Recording of events associated with therapeutic sessions: the developed device stores in an internal memory the start time of each therapeutic session and its duration. This information is uploaded to the HDC and the specialist can check if the patient is following the medical instructions to apply the therapy.

• 4)

  Data collection from the crisis and adverse events diary: the patient and his companion are trained to fill out this document and this information is stored because it is basic to calculate the monthly seizure rate (MSR), the best variable to quantify whether the treatment is being effective.

The third layer is called “system cloud application (SCA)”. This application stores all the information collected by the HDCs of the medical centers integrated into the system. In this way, a knowledge base is created as the starting point to predict the therapy effectiveness for each new patient enrolled into the system. The use of the devices can also be optimized taking in count these predictions. The knowledge base will integrate general data, clinical data and data from the therapy itself, mainly stimulation parameters and changes in MSR.

Results

Eighteen patients were studied in a single blind randomized pilot trial to demonstrate the effectiveness of a proposed therapeutic device. The average age was 29.94 years and they average suffering epilepsy was 19.44 years. The sex representation was almost similar, despite the random nature of the studied sample. General data are shown in Table 1.

The graph in Figure 2 illustrates the decrease in MSR for each patient.

Display full size

Figure 2. 

MSR reduction per patient at the end of the pilot trial

Other results of the pilot trial are the following:

• 1)

  The monthly seizure-free days increased in all patients classified as SRT, which represents an increase in their quality of life.

• 2)

  Comparing the initial video-EEG test with the same at the pilot trial end, the authors observed a decrease of the epileptiform discharges frequency in all patients who reacted as “responders to the therapy”.

• 3)

  The neuropsychological tests applied conclude that patients classified as SRT demonstrate a greater willingness to social integration and face epilepsy more optimistic, they feel protected with the tVNS.

• 4)

  Clinical analyzes remained without significant changes at the end of the pilot trial.

All data associated with the test is stored in a properly secured repository. The original of manually completed documents is preserved, for example crisis and adverse event diaries. The information captured by the proposed system or digitally imported is stored in a Structured Query Language (SQL) database designed by the authors.

Discussion

The authors classify the pilot trial outcomes as preliminary due to the sample size, although it is comparable with that of other previously published trials [10, 11]. However, it should be noted that the reached sensitivity confirms the hypothesis that the tested device is a promising tool for refractory epilepsy therapy based on tVNS.

The size of the pilot trial sample was computed using the equation 1:

Display full size

Here, N is the size of the trial sample, ρ is tabulated coefficient associated with the confidence level; the value 1.96 is assumed, representing a two-tailed test with α = 0.05, S is the standard deviation, e is the maximum allowable error.

The standard deviation of the time suffering from epilepsy of the patients studied was 11.41, and a maximum permissible error of four was assumed. The N value obtained in this way was 13.89 and was extended to 18 for practical considerations. Besides, when reviewing similar trials published from 2012 to date, several reports are based on N values between 12 and 20 patients [12–18].

The MSR behavior throughout the pilot trial is shown in Table 2. In this way, the reader can see that the first six patients were studied for nine months, while the remaining twelve were studied for six months as explained above; that is the meaning of the blank cells.

The pilot trial outcomes are promising. All the recruited patients completed the planned pilot trial time without changes in their antiepileptic drug treatment. A MSR reduction was observed in 17 of them, only a patient kept its MSR almost unchanged. However, two patients with MSR reduction did not classify as SRT because they did not meet the criteria; their MSR did not decreased by more than 50%. No patients worsened their condition after participating in the pilot trial. The graph in Figure 2 illustrates the decrease in MSR for each patient.

Subject P-01 is the only one who practically remains unchanged at the end of the pilot trial, it can be observed in Figure 2. All the others, 17 in total, had a MSR decrease by more than 35%, although the threshold to classify as SRT was set at 50%.

The reported adverse effects did not cause permanent harm or cause any patient to abandon the trial. Five patients reported adverse effects, three presented redness in the stimulation, it was eliminated by slightly moving the placement of the electrodes in the left ear concha. Two other patients reported isolated headache events that disappeared with mild analgesics and did not prevent the treatment application. Similar adverse effects were reported previously [12, 14, 15].

Several studies have been published in the last two decades on the use of tVNS in different pathologies, including refractory epilepsy [12–18]. In the majority of clinical reports or trials, the effectiveness, tolerability and safety of tVNS is determined from the use of different values for the stimulation parameters, which makes it difficult to establish comparisons [8, 19]. However, scientific community agrees that there is a decrease in seizure rate in patients with drug-resistant epilepsy.

The set of a standard for the tVNS treatment is a target to reach, but therapeutic outcomes have demonstrated its utility [19]. The therapeutic sessions and their duration vary, the duration between 20 min and 60 min and the number of sessions between three and five. The frequency of the therapeutic pulses in all the trials reviewed is 20 Hz or 25 Hz, while the pulse width varies between 150 µs and 250 µs. The same current intensity has been applied to all studied patients in some trials, but others have changed this parameter depending of the patient. All studies agree that the therapeutic effects stabilize after the third month, defining by the sixth month whether patients will respond positively to the treatment or not [14–18].

Barbella et al. [14], used very low stimulation currents, between 0.6 mA and 0.8 mA, and a third of their patients decreased their MSR by more than 30% at the end of six months of the trial. This result is very poor compared to those reported by other groups of researchers, so the authors of this paper consider that this outcome is due to the low applied current and propose that this parameter never be less than 1.6 mA.

The 4 mA current applied by Liu et al. [18] seems high for comfortable stimulation sessions, but the duration of these is 20 min, the shortest reported. The authors of this paper consider that patient comfort improves substantially when the applied current is lower. In this way, the therapy is prevented from being stressful for the subject.

Several comparisons are made between tVNS and iVNS. Recently, several scientific teams reported that both approaches involved the same anatomical pathways. Although the mode of action of is unknown, it is thought that it could inhibit the synchronization of the different nuclei that must be discharged in unison for normal electrical activity to become epileptogenic activity [20, 21].

Based on stated above, the selection of stimulation parameters made by the authors seems to be a valid proposal. More than 80% of patients respond to the therapy and only 27.78% of the subjects reported mild adverse effects, without having to abandon the test or causing permanent damage or consequences.

Patients who responded to treatment continue with the tVNS therapy and their MSR are computed every two months. It has been observed a slower MSR decrease, demonstrating a trend to keep stable. Authors considered that patients who did not meet the SRT criteria and continue decreasing their MSR could be classified as “slowly responder to tVNS”.

The outcomes of this research with the corresponding to other previously published is compared in Table 3. The main variable to compare is the “responders” because they reflect the therapeutic effect of the tested device.

Song’s [17] outcomes demonstrates that one month of therapy is insufficient to see positive results. Reviewing the rest of the trials, six months seem an acceptable time to determine if a patient will respond to tVNS-based therapy. Another aspect to highlight is that tVNS-based therapy only manages to eradicate all seizures as an exception; its main effect is a considerable decrease in the MSR, taking into account that all patients classified as responders decreased their MSR value by more than 50%.

When seeing the results obtained by the authors in Table 3, the sensitivity value stands out, is the highest among the compared trials. This is a partial result because the studied sample is small to set definitive conclusions. The authors will increase the size with new trials under the same conditions.

Outcomes on treatment result predictions are not presented yet because the prediction model need a bigger knowledge base that the available nowadays. The proposed model is based on the multivariate regression standard; it combines variables from the clinical records with age, sex, years suffering epilepsy and the MSR at the beginning of the treatment. The expectation is to get reliable predictions when knowledge base includes information about several hundred of patients with a continuous improvement.

The outcomes demonstrate that the proposed approach is effective. The developed device got the sanitary approval for its clinical use issued by a regulatory body that has approval with those of other countries. A comfortable and effective device was developed and tested successfully; patients learn its use quickly and feel good using this device.

The proposed platform has not been fully evaluated, but the design is flexible and can be adjusted to different scenarios. In fact, in this initial phase the upper layer only stores information to create the knowledge base and does not issue predictions of therapeutic effectiveness for new patients enrolled. The two lower layers have been the basis for the evaluation of the proposed device and have worked satisfactorily, without errors occurring that prevented the development of the pilot test.

The papers reviewed by the authors focus on the testing of a device while the present document discuss a working platform and the evaluation of the therapeutic device are presented.

The proposed system is a promising option for the implementation of a refractory epilepsy therapy service based on tVNS. Nevertheless, new trials are being preparing to reinforce the outcomes with a greater solid statistic.