The added value of risk assessment and subsequent targeted treatment for epileptic seizures after stroke: An early-HTA analysis

to the clinical diagnostic process for the prediction of PSE may help to establish targeted and personalized treatment for high-risk patients, which could lead to improved patient outcomes. We assessed the added value of a risk assessment and subsequent targeted treatment by conducting an early Health Technology Assessment. Methods: Interviews were conducted with four relevant stakeholders in the field of PSE to obtain a realistic view of the current healthcare and their opinions on the potential value of a PSE risk assessment and subsequent targeted treatment. The consequences on quality of life and costs of current care of a hypothetical care pathway with perfect risk assessment were modeled based on information from a literature review and the input from the stakeholders. Subsequently, the maximum added value (the headroom) was calculated. Sensitivity analyses were performed to test the robustness of this result to variation in assumed input parameters, i.e. the accuracy of the risk assessment, the efficacy of anti-seizure medication (ASM), and the probability of patients expected to develop PSE. Results: All stakeholders considered the addition of a predictive biomarker for the risk assessment of PSE to be of value. The headroom amounted to € 12,983. The sensitivity analyses demonstrated that the headroom remained beneficial when varying the accuracy of the risk assessment, the ASM efficacy, and the number of patients expected to develop PSE. Discussion: We showed that a risk assessment for PSE development is potentially valuable. This work demonstrates that it is worthwhile to undertake clinical studies to evaluate biomarkers for the prediction of patients at high risk for PSE and to assess the value of targeted prophylactic treatment.


Introduction
Stroke is the most frequent cause of epilepsy in the elderly population, and the incidence of post-stroke epilepsy (PSE) increases annually [1,2].Population-based studies indicate a PSE incidence of approximately 7 % among stroke patients [3][4][5][6].The highest risk for PSE development occurs in the first year post-stroke [4], but also persists increasing over the long term [3,5].PSE relates to a worse functional outcome, a decrease in quality of life (QOL), and a higher mortality rate [7][8][9][10][11].Despite the frequent occurrence of PSE and its serious consequences, it is still largely unknown which underlying mechanisms cause epilepsy after stroke [6].This makes early detection of the disease and adequate treatment difficult.
Currently, risk assessment scores for developing PSE are based on clinical stroke features [12,13].In the case of ischemic stroke, the SeLECT-score takes into account stroke severity, large-artery atherosclerotic etiology, early seizures, cortical involvement, and involvement of the middle cerebral artery territory [12].In the case of hemorrhagic stroke, the CAVE-score takes into account the cortical location, age, lesion volume, and acute symptomatic seizures as high-risk clinical features [13].These models have been validated by looking at how well the discrimination of participants with and without PSE performed by the models matched the observed values in validation cohorts.The area under receiver operating characteristic curve resulted in an overall concordance statistic of 0.77 (0.71-0.82) and 0.69 (0.59-0.78) respectively, which is not yet considered sufficient to implement these prediction rules in the diagnostic process.Another issue preventing clinical implementation is insufficient evidence on the efficacy of prophylactic anti-seizure medication (ASM).Only a few clinical trials examining preventive treatment are conducted, whereby variability in the definitions and methodological designs led to inconsistent and inconclusive findings [14].Therefore, the current guidelines do not advocate for prophylactic treatment [15,16].
It has been shown that not only clinical features themselves but also pre-existing underlying pathophysiological factors might play a role in the development of epilepsy after stroke [17,18].One such biological factor is the dysfunction of the blood-brain barrier (BBB).The BBB separates the circulating blood from the brain parenchyma, protects the brain from neurotoxic substances in the blood, and maintains the microclimate inside the brain tissue [19].Preclinical studies suggest that BBB dysfunction causes an imbalance in neurochemical compounds that can contribute to triggering seizures [20,21].Two common methods to assess BBB dysfunction are MRI [22,23] and the measurement of blood serum S100β levels [24,25].Other studies showed that other factors, such as neuroproteins (e.g., Hsc70) [26], neurotransmitters (e.g.glutamate or GABA) [27], inflammatory molecules (e.g.interleukins or TNFα) [28], or genetic factors (e.g.ALDH2 [29], TRPM6 [30], or CD40 [31]) may also function as potential predictive biomarkers for PSE.Also, realtime brain function measurements with electroencephalography (EEG) might be used as a biomarker for epileptic seizures after stroke [18,32].These biomarkers or (more likely) a combination thereof, added to the clinical information, may improve the predictive potential compared to using merely clinical variables [26].
Obtaining a more accurate prediction of PSE may help to apply prophylactic treatment with ASM for specific patients at risk, which could result in improved patient outcomes.This would not only benefit stroke survivors but also their caregivers and it may reduce the use of healthcare resources and costs.Therefore, in this work, we will conduct an early Health Technology Assessment (early-HTA) to assess whether a more accurate assessment of a patient's risk of developing epilepsy after a stroke, which would enable targeted prophylactic treatment, could be of added value.For this, stakeholders were interviewed, a scoping review was conducted to assess QOL and functional outcomes of PSE patients compared to stroke patients without epilepsy, and a headroom analysis was performed.

Methods
With the combination of stakeholder interviews, a literature review on functional outcome and QOL, and a headroom analysis, we used an evidence-informed approach to perform an early-HTA to obtain a deliberate understanding of the potential value of more accurate PSE risk prediction and targeted prophylactic treatment in stroke patients.

Stakeholder interviews
Four relevant stakeholders in the field of PSE were contacted: a neurologist, a nurse specialist, a patient with PSE, and a patient representative with epilepsy.The stakeholders were invited for an interview by email in which the study methods and aim of the early-HTA were explained briefly, and prior to the interview permission was obtained to make audio recordings.A semi-structured interview was performed for all stakeholders separately, in which questions were slightly adjusted to match the stakeholder's expertise.The interview consisted of a summary of current healthcare given by the researcher (EE), after which stakeholders were asked about their experience with current practice and what aspects of improvement could be of value.After the researcher explained the early-HTA, the stakeholders were asked whether they found the biomarker research of value and what they thought might be needed to implement the new method in the clinic.The interview script can be found in Supplementary 1.
The audio recordings of the interviews were subsequently transcribed and the results were grouped by the researcher (EE) into the following categories: experiences with care-as-usual (CAU) and possible points of improvement in CAU, and expectations and requirements for the implementation of an additional biomarker assessment.The results were discussed with the other authors.

Functional outcome and quality of life
A scoping review was performed to obtain information on functional outcome and QOL of patients with PSE, as compared to stroke patients who do not suffer from seizures.The modified Raskin Scale (mRS) is used as a measure of functional outcome, which has been shown to be associated with stroke severity [33].With an mRS-score higher than 2 patients are no longer able to live (completely) independently, and with an mRS-score higher than 4 patients will likely end up in a nursing home.To obtain information regarding mRS after stroke and PSE we searched the literature using the following search string in PubMed: ("poststroke epilepsy" OR "post-stroke epilepsy") AND ("mRS" OR "modified Rankin Scale").Additionally, studies were included based on cross-referencing.Only the studies including patient numbers of mRS > 2 in stroke as well as PSE were included.We also searched for research articles that reported QOL scores in stroke patients with and without PSE, for which the following search string was used in PubMed: ("poststroke epilepsy" OR "post-stroke epilepsy") AND ("quality of life").We extracted the mRS and QOL scores from the identified studies.

Headroom analysis
To examine the potential maximum added value of more accurate risk assessment for the prediction of PSE and subsequent prophylactic treatment we applied the headroom method [34,35] For this, we compared two strategies (Fig. 1): The first strategy reflects care-as-usual (CAU), where the stroke patients do not receive an additional risk assessment and no patient receives prophylactic treatment.In the second (new) strategy, the use of a hypothetical perfect risk assessment and subsequent targeted prophylactic treatment to prevent seizures in highrisk patients was added to CAU.In both strategies, we considered two health states: stroke with and without PSE.Below we describe the resource use and costs used in the two strategies in detail.

Care-as-usual (CAU) strategy
The acute phase of stroke can be defined as the period between stroke onset and the moment the patient is medically stable.During this phase, the patient is hospitalized for an average duration of 6 days (own institute numbers), with daily costs of €482 [36].Accordingly, investigations are conducted to determine the patient's exact condition.In the case of ischemic stroke, patients are often acutely treated (i.e.intravenous thrombolysis (IVT) [37] or intra-arterial thrombectomy (IAT) [38]).This treatment comes with a cost of €389 (own institute numbers).Other than monitoring the blood pressure with antihypertensive drugs, there is no acute treatment available for patients with cerebral hemorrhage.
When a post-stroke seizure occurs within 7 days after stroke onset this is seen as an acute symptomatic seizure, also described as an early seizure by the International League Against Epilepsy (ILAE) [39].Although this type of seizure increases the risk for PSE development, the relatively lower risk of seizure recurrence means that it is not yet classified as PSE at this stage.
After about one week, approximately 30 % of the patients are transferred to a rehabilitation center or a nursing home (own institute numbers), where sufficient rehabilitation facilities are available.The duration and intensity of these therapies are adjusted to the functional status of the patient and stroke severity.In this study, an average duration of admission was assumed to be 90 days (RR), with 3 h of rehabilitation treatment per day (RR).
From seven days after stroke onset, the seizure is seen as an unprovoked seizure, also described as a late seizure by the ILAE [39], and since the risk for recurrent seizures exceeds 60 %, it is classified as PSE at this stage.The likelihood of developing PSE is highest during the first months after the stroke and is assumed to concern about 15 % of the patients based on literature [13,40].These patients who developed PSE will be re-admitted to the nursing ward for approximately six days (own institute numbers).Also, a larger population of ~ 10 % of these patients are expected to end up in a rehabilitation clinic, as they are often more severely impaired compared to stroke patients who did not develop seizures (RR).Furthermore, these patients will be given ASM for approximately 6 months starting from the moment of their first unprovoked seizure, costing €4 per day (own institute numbers), assuming an ASM efficacy of 70 % (RR).
After approximately three months, therapists will help the patient at home to deal with their long-term deficits in the most optimal way.Only the recurring rehabilitation, of which the expectations are that 50 % (RR) of the patients still require physiotherapy (€41 [36]), occupational therapy (€40 [36]), and speech therapy (€37 [36]) on a biweekly basis (RR), were taken into account for the headroom calculation after the first year.

New strategy
Adding a risk assessment to the current clinical diagnostic process will include extra costs, which were assumed to be an additional MRI exam (€251[41]).As we conducted a headroom analysis, we assumed perfect accuracy of the risk assessment.For the patients that were identified with high risk for PSE development (estimated to be 15 % of the patients [13,40]) we included daily prophylactic ASM usage (€4 per day for a lifetime, own institute numbers), and assumed optimal effectiveness of ASM.As a result, in the new strategy no patient developed PSE.

Analysis
Input values for the unit prices, care usage, and care volume were based on the costing manual of the Institute for Medical Technology Assessment [36], or, if not available, on information from our institution, or expert's opinion (RR).All unit prices were converted to the year using the Dutch Consumer Price Index (CPI)[41] (with a factor CPI / CPI 2014 = 121.43/ 99.40 = 1.22).
The time horizon was set to 10 years, based on the expected lifespan of our study population [42].All costs for the first year can be found in Supplementary 2A.After the first year, only extra rehabilitation and ASM costs were considered to be made after the development of PSE, as the stroke rehabilitation costs were expected to be the same for both strategies and therefore wouldn't affect the results (Supplementary 2B).
An amount of €50,000 was chosen as the willingness to pay (WTP) per QALY, as determined by the disease burden tool of the Institute for Medical Technology Assessment [43].
The headroom was then calculated according to Equation 1: Sensitivity analyses were performed to address the uncertainties of input values used in our model.We varied three different relevant input parameters while keeping all other input parameters constant: 1) the accuracy of the biomarker (40 %-100 %), 2) the ASM efficacy (50 %-100 %), and 3) the number of patients that are expected to develop PSE (0 %-25 % in both strategies).Also, we examined the robustness of the headroom for changes in the biomarker accuracy (40 %-100 %) and the ASM efficacy (50 %-100 %) simultaneously.

Stakeholder interviews
From the interview with a stroke patient we learned that the communication of physicians about the possibility of having epilepsy after a stroke is missing in current practice: "We were never told that there was a chance of epilepsy after the stroke at all.When this happened, people around me didn't know what was happening and panicked.We would have liked to know in advance what to expect."This was confirmed by the clinicians who argued that the probability of developing PSE is low and would unnecessarily worry patients.The application of a risk assessment that can predict the chance of developing epilepsy after stroke was therefore seen as of additional value.Besides, both the neurologist mentioned that "a biomarker would be valuable to include in consideration of treatment options, in addition to the factors we are currently looking at."This view was shared by the nurse specialist.On the other hand, they argued that the current consensus is that prophylactic treatment is not useful and that more research is needed to optimize this specific mode of treatment.Also, the patient representative was unsure about undergoing prophylactic treatment because the adverse effects of ASM can have a major impact, but "With the knowledge I have now, after epilepsy, I would do everything I could to prevent it from happening."Additional examinations necessary to measure a certain biomarker for predicting the risk of PSE development were not necessarily experienced as burdensome by the patient, certainly not if this could improve the treatment plan.

Quality of life and functional outcome
The literature search to assess functional outcome identified 14 papers.After applying the in-and exclusion criteria we included 6 studies [44][45][46][47][48][49] (Fig. 2).A higher percentage of patients with an mRS-score > 2 was found in the PSE group compared to the group with stroke only, meaning that more patients with severe stroke developed seizures and became partially or completely dependent on daily care.For all studies, the difference in the percentage of patients who had a mRS > 2 between stroke patients and PSE patients was approximately 30 %.
The literature search on QOL also identified 14 articles, of which one study reported results in stroke patients with and without PSE.This study by Winter et al. (2018) [50], used the EuroQol 5 dimensions (EQ-5D), which is a widely used generic instrument of generic QOL, also in patients who have suffered a stroke [51].The EQ-5D includes five dimensions of generic health-related QOL to calculate a utility score, and the visual analogue scale (EQ-VAS).A significantly lower QOL was observed in patients with PSE compared to stroke-only patients 24 months after stroke onset (EQ-5D utility score 0.52 ± 0.31 vs. 0.66 ± 0.24 (p = 0.02) in an acute ischemic stroke population (n = 374) [50].

Headroom analysis
An EQ-5D utility score of 0.66 was assumed for patients without PSE and an EQ-5D utility score of 0.52 for patients with PSE [50], which led to 6.39 QALYs in CAU and 6.60 QALYs in the new strategy.A total cost of €40,178 was found for the CAU strategy and a total cost of €37,695 was found for the new strategy.Thus, adding a perfect risk assessment for early identification of patients at risk for developing PSE and subsequent effective targeted treatment to prevent seizures in high-risk patients leads to a reduction of €2,483 in healthcare expenditure and a QALY gain of 0.21, resulting in a headroom of €12,983 (Table 1).
The sensitivity analyses (Fig. 3 and Table 1) show that health improvements, cost savings, and the headroom remained substantial when lowering the ASM efficacy, as well as the number of patients that are expected to develop PSE.The addition of a risk assessment will remain beneficial in costs up to a biomarker accuracy of 85 %, while the gain in QALYs leads to the headroom being beneficial from a biomarker accuracy of 50 % (Headroom = €6,395, Fig. 3, Table 1).When we simultaneously vary biomarker accuracy and ASM efficacy, we see that the biomarker accuracy can even be lower than 50 % in case ASM effectiveness increases (Fig. 4).For example, the new proposed strategy would already be beneficial when the biomarker is only 40 % accurate in the case that ASM efficacy is 73 %.Assuming a hypothetical perfect biomarker accuracy an ASM efficacy of 53 % would already be sufficient.

Fig. 2.
An overview of studies that included mRS-scores for both stroke and post-stroke epilepsy patient groups and performed a non-longitudinal group-wise comparison.The percentage of the patient population that became partially or completely dependent on daily care (mRS > 2) was higher in the PSE group compared to the group with stroke-only in all studies.Abbreviations: mRS, modified Rankin Scale; E.P. Elschot et al.

Table 1
Costs, QALYs, and the associated headroom of the first 10 years after stroke for each strategy.The sensitivity analyses show that the new application will remain beneficial when varying the input parameters.A more detailed overview of the costs can be found in Supplementary 2. *Abbreviations: CAU, care-as-usual; QALYs, Quality adjusted life years; ASM, Anti-seizure medication;

Discussion
We performed an early-HTA analysis to explore the value of a risk assessment for developing epileptic seizures after a stroke and subsequent targeted prophylactic treatment.From the point of view of all interviewed stakeholders, the application of a suitable risk assessment would be of value.We showed this is beneficial both in terms of financial costs (€2,483 cost savings) and health benefits (0.21 QALY gain) compared to CAU, leading to a headroom of €12,983.The sensitivity analyses showed that it will remain beneficial when lowering the biomarker accuracy up to 50 %, when lowering ASM efficacy up to 30 %, or when lowering the population that is expected to develop PSE up to 1 %.The two-way sensitivity analysis taking into account changes in both biomarker accuracy and ASM efficacy shows that the headroom will even remain substantial for a biomarker accuracy lower than 50 % in case the ASM efficacy is higher than 89 %, or for an ASM efficacy lower than 30 % in case the biomarker accuracy is higher than 80 %.Now that this study shows us the potential added value of a PSE risk assessment and targeted treatment new research topics arise.First, a suitable biomarker sensitive to PSE development needs to be studied.Although in the current study it was assumed that a risk assessment for PSE prediction would be based on biomarkers obtained from an additional MRI exam, other factors, such as blood biomarkers or genetic factors might be used for risk prediction.Nevertheless, the headroom analysis shows that as long as the costs for implementing such an application, in addition to the MRI, are less than €12,983, there are still merits to this approach.
Second, before accurate risk prediction will actually lead to the prevention of PSE, the availability of effective prophylactic treatment approaches is necessary.This was also mentioned by the stakeholders, who argued that there is currently no consensus for the prophylactic treatment for seizure development in stroke patients, and more research is required before it can be applied in the clinic.Our model showed the potential of ASM prophylaxis, even at low efficacy.However, it should be taken into account that potential side effects of the prophylactic ASM, including potential overtreatment of patients at risk who will never develop PSE, were not considered in our model.To our knowledge, the effect of prophylactic ASM usage on the QOL in patients without one or more epileptic seizures is not investigated yet, which makes it complicated to include in the model.Nevertheless, we expect only a minimal decrease in QOL due to prophylactic ASM usage, assuming only minimal to absent side effects, which is most likely.The differences in QOL due to ASM side effects are relatively small compared to the decrease in QOL caused by seizures.Furthermore, ASM is already widely used in the general epilepsy population [52] and has been shown to lead to an improvement in QOL despite the potentially adverse side effects, where even a twofold increase in the probability of improved QOL can be reached when treatment is adjusted after patient's complaints [53].An optimal risk-to-benefit balance has to be defined by obtaining a better understanding of the patient's individual epilepsy characteristics, such as the nature of epilepsy and response to certain ASMs, thereby preventing or correcting overtreatment [54].Also, the use of ASM after PSE diagnosis was shown to be well tolerated and provided better seizure control [55].However, these randomized clinical trial studies of prophylactic PSE treatment have been performed only in small populations with insufficient quality to recommend clinical use.The results of these studies however hint toward a positive effect of short-term treatment only after early seizures have occurred, because of the increased risk for recurrent seizures [56][57][58][59].The evidence also suggests that the type and the duration of ASM should be determined for each patient specifically, as the severity of the patient's condition and underlying diseases must be considered and overtreatment should be avoided [54,60,61].
For this study, we faced the general challenges of performing an early-HTA related to diagnostic applications in medical healthcare [62].A standardized approach for conducting an early-HTA, and a headroom analysis specifically, is missing [63].We combined the quantitative headroom analysis with qualitative stakeholder interviews and a scoping literature study.We considered this a key advantage of this study, because the proposed application of a specific biomarker for predicting disease outcomes does currently not exist, and the future scenario had to be based on assumptions and values from comparable studies.Despite the limited number of stakeholders that we were able to interview, we did obtain a more comprehensive insight into the needs, expectations, concerns, and aspirations of the stakeholders.Supplemented by a general overview of the existing knowledge obtained from the scoping literature review, this was taken into account in the model to perform a more comprehensive and accurate analysis.We noticed that the generalizability of the input values obtained from the literature is challenging because of the variability between study methods, mainly varying in outcome parameters and the moment of the examination.For example, there is a variation in including patients with early seizures, late seizures, or both.Furthermore, the moment of examination varied over the studies from days to weeks, to even years after stroke.Also, cost-effectiveness studies are available both for stroke as well as epilepsy, but economic studies on PSE are currently missing.For that reason also, we conducted several sensitivity analyses to demonstrate the robustness of our results to variations in input values and assumptions.

Conclusions
A risk assessment for early identification of PSE development and subsequent targeted treatment in stroke patients has substantial potential benefit.The results show that it is valuable to conduct clinical studies researching the development of biomarkers for the prediction of patients at risk for PSE to guide the effective use of ASM in high-risk patients.Prognostic biomarkers might be incorporated into the clinical routine, which can lead to personalized and optimized treatment plans that can also be of benefit to the patient's caregivers and care providers.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Decision tree showing CAU and a risk assessment additional to the CAU for stroke patients.For the headroom calculation, it was assumed that no patient would develop PSE after CAU + Risk assessment and targeted treatment (boxes with red text).Abbreviations: CAU, Care-as-usual; PSE, Post-stroke epilepsy; ASM, Anti-seizure medication.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3 .
Fig. 3. Results of the sensitivity analyses showing the robustness of our model to uncertainty in biomarker accuracy (left column), ASM efficacy (middle column), and the number of patients that are expected to develop PSE (right column) in terms of ΔCosts (top row), ΔQALYs (middle row), and Headroom (bottom row).The input values used for the base case headroom analysis are indicated by the blue dots.*Abbreviations: ASM, Anti-seizure medication; PSE, post-stroke epilepsy; QALYs, Quality adjusted life years.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)