Endovascular Thrombectomy with or without Bridging Thrombolysis in Acute Ischemic Stroke: A Cost-Effectiveness Analysis

Intravenous thrombolysis (IVT) with recombinant tissue plasminogen activator has been a long-standing, evidence-based treatment approach for acute ischemic stroke [1, 2]. However, IVT is typically delivered within 4.5 h, has minimal efficacy in proximal large vessel occlusion (LVO) [3, 4], and may even increase the risk of intracranial hemorrhage [5]. Since 2015, several randomized clinical trials (RCTs) have shown robust evidence supporting the use of endovascular therapy in patients who present with an acute ischemic stroke secondary to a LVO [6‒15]. Meanwhile, the 2019 European Stroke Organisation (ESO) – European Society for Minimally Invasive Neurological Therapy (ESMINT) guidelines found high-quality evidence recommending the use of endovascular thrombectomy (EVT) and best medical management within 6 h after symptom onset and moderate-quality evidence for use of EVT and best medical management in patients presenting within the 6 h–24 h window [16]. The ESO-ESMINT guidelines published an expedited recommendation strongly recommending EVT plus IVT over EVT alone [17]. Similarly, the 2019 American Heart Association guidelines and the Society of Vascular and Interventional Neurology Guidelines and Practice Standards (GAPS) Committee provided a Class I recommendation for the use of IVT prior to EVT in IVT-eligible patients [18, 19].

While there remains limited evidence supporting added clinical benefit of IVT when added to EVT, which remains corroborated by meta-analyses of randomized trials [20], there is also the concern of cost of IVT [21]. The economic value of functional gains with the addition of IVT is unclear and could shape decision making in clinical practice and health policy. To this end, we performed a model-based economic evaluation to assess the short-term and long-term effectiveness of the two strategies based on results from our systematic review and meta-analysis of RCTs [20].

This analysis was a model-based cost-utility analysis. We combined a decision tree for short-term and a Markov model for long-term costs and outcomes. The detailed protocol of the systematic review and this cost-utility analysis are described elsewhere [22].

With this cost-utility analysis, we estimated total and incremental cost and quality-adjusted life years (QALYs) and incremental cost-effectiveness ratio, expressed as an incremental cost per QALY gained of IVT in addition to EVT compared with EVT alone.

The target population is adult patients with acute ischemic stroke secondary to anterior circulation LVO. Of the various populations included in this analysis [23‒28], the mean age ranged from 68.7 to 74.0 years old (median 69.3), with 48.8–62.8% (median 56.4%) being male. In this analysis, a hypothetical of 1,000 patients (500 male and 500 female), aged 69, entered the model.

We compared the costs and outcomes of two strategies: EVT plus IVT versus EVT alone. For EVT-only strategy, patients received one or more of the following techniques for thrombectomy: stent retrievers, aspiration catheters, or both. For EVT plus IVT strategy, patients received IVT with a fibrinolytic agent (i.e., alteplase or tenecteplase) before the thrombectomy procedure as bridging therapy [22]. There was no consideration of time-dependent treatment effects. We presumed all patients were treated within 4.5 h at a time where the average beneficial effect of IVT would occur.

This analysis predicted the long-term costs and outcomes over a 20-year time horizon as the reference case analysis. Considering the life expectancy of people aged 69, a 20-year time horizon could be considered as lifetime horizon [29]. Due to uncertainties of long-term outcomes for both treatment strategies, we also conducted the scenario analysis for outcomes and costs over a 90-day, 1-year, and 5-year time horizon. An annual discount rate of 3% was applied to both costs and QALYs for reference case and the scenario analysis with a 5-year time horizon.

We conducted the analysis from a healthcare perspective to consider the direct medical costs. From the healthcare perspective, we considered both medical costs (i.e., costs related to patient care) and non-medical costs (i.e., overhead costs such as finance, human resources, administration) that are incurred during treatment [30].

The following main assumptions for this analysis were included:

• The treatment strategies only impacted the risk of sICH in the short term. We only considered the impact in the first 3 months after the index stroke incident [23‒28].

• At 3 months after the first stroke incident, a patient would be either functionally independent, disabled, or dead [31]. The long-term health outcomes would be conditional on their health states at 3 months (i.e., functional independence or disability).

• The functional independence or disability was defined by the modified Rankin Scale [32].

• Regardless of EVT only or EVT plus IVT for the first stroke incident, there was no difference in treatment decision for recurrent strokes. Patients can experience one recurrent stroke (with a maximum of two strokes) within the lifetime horizon.

• Disability was associated with increased risk of mortality and reduced health-related quality of life.

We combined a decision tree with a 90-day time horizon and a Markov decision-analytic model with a lifetime horizon to predict the costs and outcomes after EVT plus IVT and EVT-only strategies. The decision tree (Fig. 1) examined the short-term costs and outcomes in 90 days after the occurrence of index stroke. We considered the impact of intracranial hemorrhage but assumed that after 90 days of the index stroke event, a patient is either functionally independent, disabled, or dead.

We then used a Markov model (Fig. 2) to assess the costs and outcomes over 1-year, 5-year, and 20-year time horizons. The Markov model simulates the long-term costs and outcomes after 3 months, including four mutually exclusive health states: functional independence, disability, recurrent stroke, and death. Our analysis had a cycle length of 3 months. Patients could transfer between health states or stay in the same health at the end of each cycle. For example, patients who are functionally independent can stay functionally independent, develop recurrent stroke or disability, or die. However, we assumed that patients who have experienced recurrent stroke could transfer to functional independence, recurrent stroke, or death, but they cannot experience another stroke recurrence within their lifetime. Death is the absorbing health state, the state that cannot be left once entered.

Table 1 summarizes the parameters used in this model. Additional probability parameters are summarized in Table S1 (for all online suppl. material, see https://doi.org/10.1159/000535796). The clinical outcome inputs on the treatment effects of EVT alone and EVT plus IVT in the short run (0–90 days after the first stroke incident) were derived from the results of our systematic review [20]. We assumed that the functional independence, disability, and death were exclusive health states at 90 days, and the sum of their probabilities was 1. Meanwhile, the intracerebral hemorrhage was not exclusive to other types of events. The probabilities of intracerebral hemorrhage, functional independence, disability, and death were based on the median proportions of these outcomes following EVT alone in the included studies [23‒28]. Briefly, the median incidence of intracerebral hemorrhage was 4.3%, reported by Yang and colleagues [24], and in a separate trial, functional independence at 3 months was 54.3%, disability was 28.5%, and mortality was 17.2% [23]. We estimated the probabilities in patients receiving EVT plus IVT treatment based on relative risks and assumed control group (EVT alone group) probabilities. Our meta-analysis reported the relative risks were 0.83 (95% confidence interval (CI) 0.59–1.18), 0.96 (95% CI: 0.88–1.04), and 1.06 (95% CI: 0.88–1.28) for sICH, functional independence, and death [20]. We estimated the probability of disability by subtracting the probabilities of functional independence and death from 1.

After 90 days, the transition probabilities between health states were based on medical literature [31, 33, 38]. We used the Copenhagen Stroke Study as our basis to estimate the transition probability to recurrent stroke from either functional independence or disability [33, 34]. Furthermore, Hankey and colleagues reported an increased risk of death and disability for people who experienced a recurrent stroke compared to the first stroke [38]. The other long-term transition probabilities from functional independence and disability to other states were based on a model-based cost-utility analysis by Xie and colleagues, which examined the cost-effectiveness of EVT versus IVT [31].

We assumed the treatment costs excluding medication cost for IVT and intracerebral hemorrhage were equivalent for those who receive EVT and IVT and those who receive EVT only. The cost estimate came from the administrative data from our institution. The cost for IVT agent was estimated to be $8,088 (for alteplase 100 mg). Other cost estimates were based on literature, including treatment costs for intracerebral hemorrhage [35], annual costs of functional independence, disability, and recurrent stroke per cycle (3 months) [36]. We also applied an end-of-life care cost for one cycle for those who died in the model, based on a cost report for patients dying in the intensive care unit [39].

We conducted a targeted literature search on utility values for functional independence and disability after stroke and recurrent stroke. The baseline utility was 0.32 for those who enter the model (after the index stroke) and those who experience recurrent stroke, based on a study on utility values of health states after stroke [37]. Broderick and colleagues summarized the utility values by modified Rankin Scale [32]. We used 0.76, the median value for a mRS score of 2, as the utility for people who regained functional independence, and 0.33, the median value for a mRS score of 4, for those who experienced disability [32]. The utility for patients with a major intracranial bleeding event was estimated to be 0.15 [40].

Using our decision analytic model, we estimated total and incremental cost and QALYs for each strategy, and incremental cost-effectiveness ratio was expressed as an incremental cost per QALY gained of IVT in addition to EVT compared with EVT only. We ran Monte Carlo simulations 10,000 times to capture the parameter uncertainties and used the mean values to represent the reference case. Table 1 summarizes types of distributions assigned to each input parameter used in the probabilistic analysis. The results were presented in a scatter plot on the cost-effectiveness plane or a cost-effectiveness acceptability curve. We presented uncertainty quantitatively as the probability that an intervention is cost-effective at specific willingness-to-pay values.

We examined the robustness of our results on reference case cost-effectiveness through one-way sensitivity analyses on key model inputs and assessed the impact of structural uncertainty through the following scenario analyses: short-time horizons (90-day, 1-year, or 5-year), no intracerebral hemorrhage in the model, and no recurrent stroke.

Our results found that the additional IVT did not lead to significant QALY gain but increased the costs. The total costs and QALYs were estimated to be $85,711 and 4.289 QALYs for the EVT plus IVT strategy over a 20-year time horizon, and $76,561 and 4.158 QALYs for the EVT-only strategy (Table 2). The probability of being cost-effective was 24.2% for the EVT plus IVT strategy, compared with 75.8% for the EVT-only strategy over a 20-year time horizon at a willingness-to-pay threshold of $50,000 per QALY gained (Fig. 3).

Scenario analyses with shorter time horizons showed similar results: the average costs per patient were estimated to be $55,482, $57,751, $68,314, and $85,611 for EVT plus IVT and $47,304, $49,510, and $59,770 for EVT-only strategy, over 90 days, 1 year, 5 years, and 20 years, respectively. For all scenarios, the cost saving of EVT-only strategy was driven by the avoided medication costs of IVT (ranging from $8,178 to $9,050). Both strategies had similar outcomes in scenario analyses. The additional IVT led to slight decrease in QALY estimate during the 90-day time horizon (loss of 0.002 QALY), but a small gain over 1-year and 5-year time horizons (0.011 and 0.0636 QALY). As the small QALY gain accumulated a long time horizon, analyses with longer time horizon show lower probabilities of EVT-only strategy being cost-effective. At a willingness-to-pay threshold of $50,000 per QALY gained, the probabilities of EVT only being cost-effective were 100%, 100%, and 99.3% over 90-day, 1-year, and 5-year time horizons. At the willingness to pay thresholds of $100,000 per QALY, the probabilities of EVT-only strategy being cost-effective were 100%, 100%, 74.8%, and 28.9% over 90-day, 1-year, 5-year, and 20-year time horizons. Our scenario analyses, assuming there was no recurrent stroke or intracranial hemorrhage, showed similar cost-effectiveness results with reference case (see online suppl. Table S2; Fig. S1). We conducted a one-way sensitivity analysis, which showed that the cost-effectiveness results were sensitive to the relative risks of outcomes for EVT plus IVT versus EVT only; however, the results were robust when we varied other cost and probability parameters (see online suppl. Table S3).

Our model-based analysis found that EVT-only treatment strategy was less costly compared with EVT plus IVT strategy over all time horizons. The cost-saving was mainly due to the saved cost of IVT medicine [41]. Over the lifetime horizon, our analysis suggested that there was an insignificant loss in QALY for EVT-only strategy. Because of the similar short- and long-term clinical outcomes of EVT with and without IVT and the cost-saving of EVT-only strategy, our analysis suggested that EVT-only strategy was cost-effective for the typical patient treated with tissue plasminogen activator within a 4.5-h window. In contrast, the probability of EVT plus IVT strategy being cost-effective increased as time horizon prolonged. This was because of the small increase in probability of functional independence and decrease in the risk of death, which would further translate to increased QALYs over a longer time horizon. The interpretation of cost-effectiveness results should be cautious because only short-term evidence was derived from RCTs included in our systematic review to support the model, but there were more uncertainties around the estimates for clinical outcomes, especially on the long-term mortality and disability, and their impact [20].

The main strength of our study was that our analysis was based on clinically relevant parameters from a high-quality systematic review and meta-analysis of eligible RCTs. In contrast to a previously published cost-effectiveness study which used data from 4 RCTs to estimate costs over a 10-year horizon [42], we used data from 6 RCTs, allowing for better representation of clinical data related to this context over a 20-year horizon. Our results validate the findings from this cost-effectiveness study by Qureshi and colleagues [42]. We used the pooled point estimates and confidence intervals with a probabilistic analysis approach to examine the impact of parameter uncertainties. Second, our model structure, combining a decision tree over a short time horizon for the index stroke treatment and a Markov model for the lifetime outcomes, provided disaggregated cost-effectiveness results over various time horizons. In the short term, EVT-only strategy led to a small QALY gain and was cost-effective, when accounting for the lower risk of intracranial hemorrhage compared to that of the EVT plus IVT group. Although the QALY results flipped over a longer time horizon, attributed to small QALY gains for EVT plus IVT strategy, the EVT-only strategy was still cost-effective under conventional willingness to pay thresholds (e.g., $50,000 per QALY gained). Our analysis can support decision making by informing patients and healthcare providers of the relative merits of different strategies.

Nevertheless, our model-based analysis had several limitations. First, our analysis assumed that the major difference between EVT-only strategy and EVT plus IVT strategy was their impact on functional dependence and disability, while the clinical pathway, comorbidities, and outcomes of stroke may be far more complicated [43]. However, available studies supported this assumption and found similar outcomes following the two strategies except for intracranial hemorrhage, functional independence, and death [23‒28]. Second, parameter uncertainties existed in our analysis. Our systematic review found low-to-moderate-certainty evidence on the outcomes over a 90-day time horizon. We conducted probabilistic analysis and sensitivity analyses to examine their impact. The cost-effectiveness results were sensitive to the relative risks of outcomes for EVT plus IVT versus EVT. However, the relative risk estimates used in this analysis were based on our comprehensive systematic review and meta-analysis [20]. Long-term uncertainty could stem from employing the mRS for defining functional independence, whereas domain-specific outcome measures may better capture the rate and extent of stroke recovery [44]. Our parameters also assumed the stability of the mRS beyond the 90-day estimate, potentially impacting the results. Third, our cost parameters were based on patient charges from a single center, which may limit the generalizability of our results. Fourth, the time-dependent benefit of IVT when added to EVT remains unclear, but it is possible – perhaps even with tenecteplase – that for patients for whom EVT is not readily available, IVT is cost-effective. It is important to note that many systems utilize a hub-and-spoke model [45]; these results perhaps should not be generalized to all scenarios as our models do not account for time from IVT – for example, administered at spoke center – to EVT at hub. Additionally, patients who underwent IVT and had already recanalized prior to EVT were not sub-analyzed based on the study-level results from our systematic review, which could significantly decrease costs for these patients. Finally, our study did not consider the potential advantage of tenecteplase over alteplase for LVO [46] or any potential functional benefit of IVT in anterior LVO when EVT may be delayed (e.g., due to inter-hospital transfer). That being said, many patients included in the RCTs comparing IVT and EVT versus EVT alone from our meta-analysis were transferred from non-thrombectomy-capable centers. Therefore, we anticipate the effect of delay on cost-effectiveness of bridging therapy with alteplase to be negligible. Another limitation of this analysis is that it is insufficient to examine the impact of age on cost-effectiveness. Empirical evidence suggests patients older than 80 years of age may have a worse prognosis [47, 48]. There were insufficient data to support an analysis focusing on this subgroup population. For example, our previous systematic review and meta-analysis did not identify sufficient information to conduct a subgroup analysis for elderly patients [20]. Nevertheless, we were able to conduct one-way sensitivity analyses of probabilities of outcomes, such as functional independence and death, within the 90-day time horizon, and the cost-effectiveness results were similar to the reference case, even assuming patients would have a worse prognosis.

Our analysis results suggested that from a cost-effectiveness perspective, the value of IVT in addition to EVT may be limited. This study could also inform the relative merits of EVT-only and EVT-plus-IVT strategies in short- and long-time horizons and provide support in clarification of patient preference and decision making. However, it may be challenging to discard thrombolytics because thrombolytics have long been considered standard of care [18, 49]. Our results are mostly applicable to the US setting, given the sources of cost data, but the two driving factors of our results, similar outcomes between strategies and cost saving due to IVT medication cost, may also exist in other settings [20]. Our results are also probably applicable in other settings, but they are not generalizable to other geographic regions.

The transitions between health states after stroke are naturally complicated. Notably, this analysis is based on study-level data over a short time horizon. However, individual-level patient data over a longer time horizon may better represent the transitions between health states. Further evidence may shed light on these issues and reduce uncertainties. Moreover, the long-term effects of additional IVT may impact cost-effectiveness. It means that for different IVT agents, the cost-effectiveness may differ too. IVT agents, such as tenecteplase, have only been introduced recently [50], and further evidence on the effects and safety of different IVT agents can improve our knowledge on their cost-effectiveness.

Our model-based cost-effectiveness suggested that EVT-only treatment may be cost-effective for patients with acute ischemic stroke secondary to LVO. In the short term, the cost-effectiveness of EVT-only strategy was driven by fewer intracranial hemorrhage events. Over the lifetime horizon, EVT-only strategy was less costly and associated with an insignificant loss in QALYs compared with EVT-plus-IVT strategy.

Professor Issam Awad, Director of Neurovascular Surgery at University of Chicago, read and provided feedback on the manuscript.

This study did not involve human participants, and therefore, informed consent was not obtained. Approval by the Institutional Review Board was not applicable.

The authors declare no competing interests.

This study was not funded.

Conception and design: R.Z.M., Y.Z., and T.K.-H. Acquisition, analysis, and interpretation of the data: R.Z.M., Y.Z., M.Z., S.X., J.C.-P., H.D., E.T., S.A.K., A.T., R.B., and S.T. Drafting of the manuscript: R.Z.M. and Y.Z. Critical revision of manuscript: R.Z.M., Y.Z., M.Z., and S.X. Statistical analysis: Y.Z., M.Z., S.X., A.K., and A.J.D. Supervision: T.K. All authors reviewed and approved the final version of the manuscript.

Rami Z. Morsi and Yuan Zhang contributed equally to this work.

Data related to the findings from this study can be requested from the corresponding author.