Cost-Effectiveness of Intra-Arterial Treatment as an Adjunct to Intravenous Tissue-Type Plasminogen Activator for Acute Ischemic Stroke
Background and Purpose—The objective of this study was to determine the cost-effectiveness of intra-arterial treatment within the 0- to 6-hour window after intravenous tissue-type plasminogen activator within 0- to 4.5-hour compared with intravenous tissue-type plasminogen activator alone, in the US setting and from a social perspective.
Methods—A decision analytic model estimated the lifetime costs and outcomes associated with the additional benefit of intra-arterial therapy compared with standard treatment with intravenous tissue-type plasminogen activator alone. Model inputs were obtained from published literature, the Multicenter Randomized Clinical Trial of Endovascular Therapy for Acute Ischemic Stroke in the Netherlands (MR CLEAN) study, and claims databases in the United States. Health outcomes were measured in quality-adjusted life years (QALYs). Treatment benefit was assessed by calculating the cost per QALY gained. One-way and probabilistic sensitivity analyses were performed to estimate the overall uncertainty of model results.
Results—The addition of intra-arterial therapy compared with standard treatment alone yielded a lifetime gain of 0.7 QALY for an additional cost of $9911, which resulted in a cost of $14 137 per QALY. Multivariable sensitivity analysis predicted cost-effectiveness (≤$50 000 per QALY) in 97.6% of simulation runs.
Conclusions—Intra-arterial treatment after intravenous tissue-type plasminogen activator for patients with anterior circulation strokes within the 6-hour window is likely cost-effective. From a societal perspective, increased investment in access to intra-arterial treatment for acute stroke may be justified.
Until recently, intravenous tissue-type plasminogen activator (tPA) administered within 4.5 hours of symptom onset was the only therapy supported by randomized clinical trials to reduce long-term disability in acute ischemic stroke (AIS).1 Intra-arterial therapy (IAT), either through local delivery of thrombolytic agents or mechanical thrombectomy, has been tested in various capacities during the past 2 decades and resulted in considerable controversy.2,3 Completed after the emergence of more effective thrombectomy devices, the recent Multicenter Randomized Clinical Trial of Endovascular Therapy for Acute Ischemic Stroke in the Netherlands (MR CLEAN) trial provides credible evidence for the clinical effectiveness of IAT for the treatment of proximal arterial occlusion in the anterior circulation within 6 hours after symptom onset, largely in patients who have already received intravenous tPA.4–6
Although the results of MR CLEAN demonstrated the effectiveness of IAT, IAT is a costly therapy requiring expensive devices and medications, highly trained proceduralists, periprocedural support from interventional teams including anesthesiology and close postprocedure monitoring. Consequently, assuming IAT is effective, the question remains whether it is cost-effective. Prior attempts to address this question, because of the lack of efficacy data, were based on rates of revascularization to predict clinical outcome at 90 days.7 This model attempts to determine whether these cost-effectiveness estimates are robust by analyzing the economic value of IAT based on the clinical outcomes directly seen in MR CLEAN.
A decision analytic model was programmed in Microsoft Excel to analyze the cost-effectiveness of using IAT (ie, arterial catheterization with delivery of a thrombolytic agent, mechanical thrombectomy, or both) in addition to standard care (ie, treatment with intravenous tPA within 4.5 hours) as compared with standard care alone. Eligible patients had an AIS caused by a proximal intracranial occlusion of the anterior circulation within 6 hours after symptom onset, after the inclusion criteria of MR CLEAN.8 A Markov state transition model was developed to represent the clinical outcomes over time in patients with acute stroke and simulated a cohort of patients presenting to the emergency department with AIS and meeting the criteria for IAT (Figure 1). Finally, lifetime cost (broken down into acute care and long-term care cost), quality-adjusted life years (QALY), and life years for each treatment group from the time of stroke until death was calculated.
Patients entered the model at stroke onset, at which point they either received IAT with standard care or standard care alone. All patients were then subsequently assigned an initial modified Rankin Score (mRS): mRS 0 with no symptoms, mRS 1 with no significant disability, mRS 2 with minor disability, mRS 3 with moderate disability, mRS 4 with moderated to severe disability, mRS 5 with severe disability, and mRS 6 or death. The cycle length was 1 year with time horizon of 30 years. At the end of each cycle patients remained in their current health state, died, or had a recurrent stroke after which they were reallocated to a lower health state with equal probability of recurrent stroke death. Of note, patients were not further stratified by the presence of symptomatic intracranial hemorrhage in the IAT or standard care group as there was no statistical difference in the rates of hemorrhage between the 2 arms in the MR CLEAN trial.8 Death whether by stroke or other causes was the only absorbing state, after which patients were excluded from the model.
Primary model input parameters were drawn from published literature and costs from the National Inpatient Sample9 (Tables 1 and 2). Whenever possible, this study relied on assumptions used in prior peer-reviewed cost-effectiveness analyses to enhance the validity of these findings as well to maximize their comparability to other stroke treatments.
All patients were assumed to be 65 years old at the time of index stroke. This was the mean age enrolled in the MR CLEAN trial.8 Mortality rates from the index stroke and functional outcomes (mRS 0–5) after 90 days were also based on the results of that trial. Stroke recurrence rate was estimated at 5.1% a year associated with recurrence stroke mortality of 19% according to the published literature.13
Patients after the index stroke were assumed to have an all-cause mortality higher than the average population.12,13 Age-specific mortality rates were obtained from life tables from the US National Vital Statistics report.14 Probability of death from all causes was obtained by multiplying the mortality rate by the relative death hazard ratios reported by Samsa et al12 according to the mRS at 90 days. Life years were discounted at 3%/y.15
Quality-of-life estimates for stroke survivors were based on published utility values stratified by mRS and summed over the lifetime of the patient to obtain QALY estimates.12
Cost of the index stroke were derived from previous literature using US nationwide estimates of Medicare Costs.16 These costs were broken down by dispositions after the index stroke, including dead at discharge, discharge to skilled nursing facility, or discharge to home or home health services. Based on the definition of the mRS, these dispositions were assigned mRS 6 for dead at discharge, mRS 4 to 5 for discharge to skilled nursing facility, and mRS 0 to 3 for discharge to home or home health services.10 The additional cost of administering tPA was estimated by the difference between the cost of AIS without tPA from the cost of AIS with tPA, in the absence of complications, using diagnosis-related groups 63 and 66, respectively.17 A similar method was used to calculate the additional cost of IAT. It was the difference between the cost of IAT and the cost of AIS with tPA, both in the absence of complications using diagnosis-related groups 27 and 66, respectively. Recurrent stroke hospitalization costs were derived from previous models obtained from an economic study that assessed these costs at 5 major academic centers.18 Annual posthospitalization costs were obtained from Earnshaw et al.10 All costs are discounted by 3%/y.15 All costs before 2012 were inflated to 2012 dollars according to the medical care component of the Consumer Price Index.19
Utility for each mRS ranged from 0 to 1. Whereas a utility value of 1 represents perfect health or no loss of utility and 0 is death or no utility whatsoever. Respective utility values were multiplied to the life years spent in that health state and summed over the lifetime of the patient to obtain QALYs. Utility values were obtained from published literature, based partly on patient surveys.12
The incremental cost and QALY between IAT and standard therapy were assessed. The incremental cost-effectiveness ratio (ICER) was obtained by dividing the cost difference by the difference in QALY. Interventions were considered cost-effective if the ICER was less than or equal to the conventional threshold of $50 000.18 Interventions were considered borderline if the ICER was between $50 000 and $100 000.
To estimate the overall uncertainty of model results, deterministic 1-way sensitivity analysis was performed. Parameters analyzed included clinical efficacy (percentage of patients reporting each mRS at 90 days for each treatment arm), acute stroke mortality rate, acute stroke care cost, tPA treatment cost, IAT cost, annual posthospitalization cost, recurrent stroke cost, mortality of recurrent stroke, utilities, and death hazard ratios (Figure 2). Values derived from the MR CLEAN trial, outcome utilities, recurrent stroke probability, and death with recurrent stroke probability were varied by their 95% confidence intervals, death hazard ratios were varied based on consensus of an expert panel, and all cost variables were varied by 50% in either direction.
In addition to 1-way sensitivity analysis, we wanted to determine how much worse IAT could have performed but still produced a cost-effective ICER. We constructed 3 hypothetical worse outcomes representing least unfavorable, unfavorable, and most unfavorable scenarios for IAT performance by assuming worse outcomes in none to mild disability states at 90 days. This was accomplished by setting each mRS 0 to 2 at its 10th, fifth, and first percentile based on its β distribution, representing least unfavorable, unfavorable, and most unfavorable outcomes, respectively (Figure I in the online-only Data Supplement). Then the total difference was redistributed evenly among mRS 3 to 6, so that the sum of all outcomes still added to 1 (Table 3). Finally, the ICER in each instance was calculated by comparing with the base case for standard therapy alone.
Monte Carlo simulation was also performed in which all variables in the 1-way sensitivity analysis were varied simultaneously. We assumed clinical 90-day mRS outcomes followed a Dirichlet distribution, which allows for the multivariate generalization of the β distribution. Proportion of patients receiving tPA, incidence of recurrent stroke, utility, and recurrent stroke mortality followed a β distribution. All costs were varied using a normal distribution (assuming 50% in each direction as the 95% confidence interval). Relative mortality risk for each mRS after the index stroke was varied using the log normal scale. The analysis was run 10 000 times to capture stability in the results. Uncertainty was represented using a scatter plot (Figure 3).
For the base case, standard therapy yielded 3.10 QALYs at a total lifetime cost of $130 144. In comparison, IAT in addition to standard therapy yielded 3.80 QALYs at a total lifetime cost of $140 055. Therefore, IAT resulted in a lifetime gain of 0.7 QALYs at an additional cost of $9911, which represented an additional cost of $14 137 for each QALY gained (Table I in the online-only Data Supplement).
One-Way Sensitivity Analysis
The tornado diagram illustrates the effects of varying input parameters on the ICER or cost per QALY (Figure 2). Overall, the 1-way sensitivity analysis showed that variation of each assumption from their minimum to maximum values still resulted in QALYs gained and cost savings. Across all parameter ranges, ICERs were within the acceptable limits of cost-effectiveness, defined as willingness to pay (WTP) of <$50 000/QALY. Of all parameters, the variable with the largest impact on the ICER across the range of values explored in sensitivity analysis was the probability of standard treatment producing an outcome of mRS 4 (30.2%). The upper bound probability (35.8%) produced an ICER of $157 and the lower bound probability (25%) of $22 467. The ICER was most affected by the outcomes of each treatment arm producing mRS 4, 3, and 5 at 90 days in addition to the cost of IAT and the annual posthosptialization cost for mRS 4 to 5.
When compared with the standard care arm, the least unfavorable, unfavorable, and most unfavorable scenarios produced ICERs of $41 816, $56 146, and $132 128, respectively. In each hypothetical worse case scenario, IAT continued to produce a benefit in QALYs when compared with standard therapy alone: 0.28 for least unfavorable, 0.21 for unfavorable, and 0.09 for most unfavorable. In the least unfavorable scenario IAT continued to be a fully cost-effective therapy. However, in the most unfavorable scenario, IAT only produces a marginal benefit over standard therapy.
Multiway Probabilistic Sensitivity Analysis
The results of the multiway probabilistic sensitivity analysis are shown in Figure 3. In 19% of the simulation runs, IAT was the dominant strategy with a lower cost and higher QALYs compared with standard treatment alone. In 79% of the simulation runs, IAT was more costly but also more effective than standard treatment alone. Overall, in 97.6% of simulation runs, IAT was recommended, meaning it was either the dominant strategy or was preferred over standard therapy alone at a WTP threshold of $50 000/QALY. In only 1.4% of the runs was standard therapy more effective than the addition of IAT. The cost-effectiveness acceptability curve in Figure 4 shows the cost-effectiveness of IAT as a function of the WTP threshold.
The addition of IAT surpassed conventionally accepted cost-effectiveness thresholds, leading to better outcomes at a cost of $14 137 for each QALY gained in the base case. This conclusion held up to significant variation in modeling assumptions. When pessimistic functional outcome distributions for IAT were compared with standard therapy, IAT remained cost-effective in the least unfavorable and unfavorable scenarios. IAT remained cost-effective at a WTP threshold of $50 000/QALY even when the number needed to treat to produce a mRS of 0 to 2 was increased to 20 in the least unfavorable outcome and at a WTP threshold of $100 000/QALY when the number needed to treat was increased to 26 in the unfavorable outcome, compared with the number needed to treat of 7 in MR CLEAN itself. Given that the recent Endovascular Treatment for Small Core and Proximal Occlusion Ischemic Stroke (ESCAPE) and Extending the Time for Thrombolysis in Emergency Neurological Deficits–Intra-Arterial (EXTEND-IA) trials had number needed to treat of 4 and 3, respectively, these assumptions are particularly pessimistic and support the idea that IAT is likely cost effective.20,21
There have been previous models that predicted the cost-effectiveness of IAT.7 Because of the previous lack of positive randomized clinical trials, past studies used results from the single arm trials such as Mechanical Embolus Removal in Cerebral Ischemia (MERCI) to predict clinical outcomes based on the rates of recanalization.3,22 This study had the benefit of a randomized clinical trial (MR CLEAN) with clinical outcomes as a primary end point. Hence, direct measurements of outcomes were used and not extrapolated from clinical outcomes of reperfusion rates. In spite of these differences in methodology, the results were largely concordant with prior IAT ICERs: $12 00023 for mechanical thrombectomy within 8 hours compared with no treatment and $16 000 for mechanical thrombectomy compared with patients who received intravenous tPA within 3 hours.24
If the efficacy of IAT demonstrated in MR CLEAN is supported by future trials, the data suggest that there is a strong societal rationale for investment in IAT systems of care. Not only does IAT results in improvement in clinical outcomes for a relatively large subset of patients at high risk for disability, but the net costs of IAT seem to be within the bounds of conventional cost-effectiveness thresholds, even under pessimistic assumptions. Thus, even if costs-per-acute treatment increased substantially, it is likely that IAT would still be a relatively good investment. This suggests that even costly investments in the capital needed to support IAT systems of care (eg, telemedicine, transport networks) may also be good investments from a societal perspective. Similarly, initiatives to improve patient recognition of stroke symptoms and self-efficacy to seek emergent medical care as well as physician educational initiatives may also prove to be wise investments. Finally, future research that results in more expensive IAT protocols (eg, more costly devices or imaging protocols) or which targets patient populations resulting in moderately less efficacy than was seen in MR CLEAN may also be cost-effective and should be explored.
The real-world cost implications of IAT remain to be seen. This study does not account for some important costs such as infrastructure, service redesign, or specialist training. Although the overall cost per QALY is reasonable for IAT, inefficient system changes may entirely erode this advantage. Conversely, it may be the case that the real-world cost implications of IAT may be even greater than seen here—overall hospitalization costs and rehabilitation costs may be lower in IAT patients because of their improved functional outcomes. Moreover, it is conceivable that the infrastructure benefits used to deliver IAT may have broader benefits leading to more general improvement in AIS care.
There are some limitations to this study. The acute treatment costs for IAT and intravenous tPA were updated using the National Inpatient Sample data to reflect the current market. However, there were not more updated studies for costs reflecting the index stroke or recurrent stroke. Hence, these estimates were inflated from previous literature and may not as closely estimate current market values. Updated studies estimating costs for long-term stroke care were also unavailable. Previous neurological cost-effective models (ie, Earnshaw et al) have based estimates on figures by Caro et al,25 which estimated lifetime cost of stroke care using a Stroke Treatment Economic Model using unit costs from the United Kingdom in 1996. Lifetime cost of stroke estimates is also used in cardiology cost-effectiveness models, which have looked at medications such as Ximelgatran or Dabigatran for preventing strokes in atrial fibrillation.26,27 These estimates were 5-fold higher and derived from looking at the cost of post stroke care in the Rochester Stroke Registry from 1937 to 1989.28 Other lifetime stroke cost estimates from Taylor et al29 were in the same range as the neurology models but were derived from Medicare claims data in 1987 using the National Medical Expenditure Survey. In lieu of more updated figures, estimates from Earnshaw et al were used and inflated to 2012 dollars. The data suggest that regardless of which of these costs is most accurate, IAT is likely cost-effective, but a better understanding of the costs of care is likely necessary to intelligently evaluate the cost-effectiveness of more marginally cost-effective treatments.
In conclusion, IAT is likely cost effective for patients with anterior circulation strokes and proximal occlusion within the 6-hour window in addition to standard medical therapy. From a societal perspective, we will have to start changing the mindset behind stroke care. Providers can no longer stop at administering intravenous tPA, but patients with suspected anterior infarcts must also have vessel imaging to see whether they would be a candidate for IAT. Furthermore, it will involve widening the network of comprehensive stroke centers so that patients meeting the criteria can have prompt access to IAT.
Sources of Funding
This work was supported by the National Institute of Health to Dr Burke.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.115.009779/-/DC1.
- Received April 15, 2015.
- Revision received April 15, 2015.
- Accepted April 24, 2015.
- © 2015 American Heart Association, Inc.
- Jauch EC,
- Saver JL,
- Adams HP Jr,
- Bruno A,
- Connors JJ,
- Demaerschalk BM,
- et al
- Smith WS,
- Sung G,
- Saver J,
- Budzik R,
- Duckwiler G,
- Liebeskind DS,
- et al
- 9.↵HCUPnet. Healthcare Cost and Utilization Project. Agency for Healthcare Research and Quality; 2012. http://hcupnet.ahrq.gov/. Accessed January 7, 2015.
- Earnshaw SR,
- Jackson D,
- Farkouh R,
- Schwamm L.
- Samsa GP,
- Reutter RA,
- Parmigiani G,
- Ancukiewicz M,
- Abrahamse P,
- Lipscomb J,
- et al
- Sacco RL,
- Wolf PA,
- Kannel WB,
- McNamara PM.
- Boudreau DM,
- Guzauskas GF,
- Chen E,
- Lalla D,
- Tayama D,
- Fagan SC,
- et al
- Tung CE,
- Win SS,
- Lansberg MG.
- 19.↵Consumer Price Index. 2014. http://www.bls.gov/cpi/. Accessed August 9, 2014.
- Smith WS,
- Sung G,
- Starkman S,
- Saver JL,
- Kidwell CS,
- Gobin YP,
- et al
- Kim AS,
- Nguyen-Huynh M,
- Johnston SC.
- Caro JJ,
- Huybrechts KF.
- Leibson CL,
- Hu T,
- Brown RD,
- Hass SL,
- O’Fallon WM,
- Whisnant JP.
- Taylor TN,
- Davis PH,
- Torner JC,
- Holmes J,
- Meyer JW,
- Jacobson MF.