All of the studies included in this review employed a cost utility analysis (CUA). Utility analysis is viewed as a particularly useful technique because it allows for given treatment outcomes to be adjusted for the health-related quality of life, while simultaneously providing a generic outcome measure for comparison of costs and outcomes in different programs. The generic outcome, which is usually expressed in quality-adjusted life-years (QALYs), is determined by adjusting the length of time affected by the health outcome according to the utility value of the resulting level of health status. Other generic outcome measures, such as disability-adjusted life-years (DALYs) and healthy years equivalent, have been proposed as alternatives to the QALY.

Considering the importance of the perspective of an evaluation study, it was remarkable that seven of the eight studies explicitly mentioned the perspective used. Two studies were performed from the societal perspective (i.e., all costs and effects were included regardless of who incurred the costs and who obtained the effects),16,18 and four studies (and one other study that did not explicitly mention the perspective17) were performed from the healthcare-system perspective.12,13,15,19 The studies assessed only the direct medical costs of rtPA treatment for stroke. One study was performed from both the healthcare-system and societal perspectives.14

The studies provided data regarding the cost-effectiveness of thrombolysis treatment for stroke for six countries: USA (n=2), UK (n=2), Canada (n=1), Australia (n=1), Spain (n=1), and Denmark (n=1). The national healthcare systems in different countries appear diverse, and so the components of cost evaluated by these studies differ markedly.

This systematic review concentrates on model-based economic evaluation studies. A model-based approach is necessary to estimate the long-term cost impact and consequences of interventions as well as the impact on the economic results of varying assumptions about risks of events, effectiveness of therapy, the cost of the intervention itself, and patient care.

Five studies used a Markov modeling approach to evaluate the cost-effectiveness of rtPA treatment for stroke. Another three studies used a discrete-event simulation model, the Model of Resource Utilization, Costs, and Outcomes for Stroke (MORUCOS), and the Stroke Outcome Model (SOM), but both of these simulation models use a semi-Markov modeling approach. Markov models are based on a series of "states" that a patient can occupy at a given point in time. Time elapses explicitly in a Markov model, with the probability of a patient occupying a given state assessed over a series of discrete time periods, called cycles. The length of these cycles vary with the disease and intervention being evaluated, but can be a certain number of months or even years. Each state in the model has a cost associated with it and, for CUA, a utility value. The duration of time during which the average patient occupies the various states in the model will-when weighted according to the relevant cost or utility-be used to calculate expected costs and outcomes. The rate with which patients move between the states in the model is determined by a set of transition probabilities.

Fagan et al.13 and Sinclair et al.14 developed a model to describe the short- and long-term outcomes of hospitalization. During hospitalization, the risk of symptomatic intracranial hemorrhage or death varied according to whether rtPA therapy was chosen. The 30-year time horizon was chosen in both of those studies. Sandercock et al.16 used a Markov modeling approach to predict the health and economic outcome of rtPA and standard care groups. The Markov model used age-specific mortality, risk of recurrent stroke, and stroke-specific case-fatality to estimate the probabilities of being dead, dependent, and independent at the beginning of each year. The Markov process was run repeatedly in 1-year cycles until the end of the cohort lifetime, and totals were computed for the accumulated health outcomes and costs. Mar et al.19 employed a cycle length of 1-year and the whole life as the time horizon. The probabilities of transition to other states changed depending on the type of stroke. The states of stroke were transitory, since at the end of the 1-year cycle all patients had moved to another state (i.e., death, autonomy, or disability). Ehlers et al.20 designed a decision tree with Markov modeling of the long-term consequences, which calculated the marginal cost-effectiveness ratio for time spans of 1, 2, 3, and 30-years. The model assumes that the patient can receive either thrombolysis or conservative treatment. Depending on the treatment instituted, the patient is exposed to a different risk of intracranial hemorrhage.

Other studies adopted modeling approaches other than the simple Markov modeling approach. Chambers et al.15 designed the SOM. The SOM comprises two modules: acute care and long-term care/prevention of recurrence among stroke survivors. A prototype primary prevention module has also been developed. The model was constructed so that results from the long-term care/prevention of recurrence module could be used as a payoff in the acute care module. This design allowed both the long-term consequences of acute events and interventions during acute care to be considered. Stahl et al.17 conducted a discrete-event simulation model of the process of stroke care from symptom onset through administration of rtPA. A literature review was performed to determine the process ties, performance of CT, health outcomes, and cost estimates. The model assumed that the patient received either a National Institute of Neurological Disorders and Stroke (NINDS)-compliant strategy (i.e., evaluation by an emergency physician within 10-minutes, interpretation of CT scans within 45-minutes, and administration of rtPA within 1-hour of presentation) or current practice. Moodie et al.18 designed the MORUCOS system, which is a detailed model consisting of four modules (natural history, costs, outcomes, and interventions) that was developed through a series of linked spreadsheets. In this model, interventions were assumed to be operating in a steady state (i.e., fully implemented and operating in accordance with efficacy potential), and they were applied to all eligible patients who presented during a 1-year period. Interventions were applied for a duration that realistically reflected their real-world use. The time horizon for tracking-associated costs and consequences extended over the lifetime of the target population (Table 1).

Fagan et al.,13 Sinclair et al.,14 Stahl et al.,17 Mar et al.19 estimated the outcomes (with or without rtPA treatment) based on the NINDS trial, which was a randomized, double blind, placebo-controlled trial evaluating the short- and long-term outcomes of tissue plasminogen activator (tPA) in 624 patients with acute ischemic stroke. Moodie et al.18 used stroke incidence, mortality, and service utilization data from the North East Melbourne Stroke Incidence Study (NEMESIS), a community-based stroke incidence study that provides the most realistic picture of current-practice stroke care, including post acute care in Australia.

More efficacy data are available nowadays, such as that from the study by Chambers et al.,15 who used clinical data including trials, meta-analysis, and prospective cohort studies such as the Oxfordshire Community Stroke Project and the Northern Manhattan Stroke Study. Sandercock et al.16 estimated efficacy based on a systematic review of all relevant randomized trials supplemented by data from a local stroke registry, the Lothian Stroke Register (LSR). The LSR data items analyzed included the length of hospital stay, the score on the modified Rankin Scale (mRS), the occurrence of recurrent stroke, death from recurrent stroke, and death from all causes up to 12-months after the index stroke. Ehlers et al.20 extracted efficacy data from a meta-analysis of six large-scale, randomized, and placebo-controlled studies of thrombolytic therapy (Table 1).20

As described above, all of the studies reviewed were undertaken from different perspectives (i.e., societal, healthcare system, or mixed). Many experts consider a societal perspective to be highly appropriate for carrying out economic evaluations of healthcare interventions. However, only two of the studies were performed from the societal perspective and included direct and indirect costs.17,19 To measure direct costs, the studies used published literature and hospital records. Indirect costs were measured based on the average daily wages reported by the United States Bureau of Labor Statistics, the Sakontzen questionnaire, and the matrix of societal needs defined by a committee of social service experts from the province of Gipuzkoa, Spain.19 Other studies included only direct costs based on published literature, local surveys, expert opinions, and hospital records. Since the healthcare system in each country is different, the components of direct cost evaluated by these studies differ markedly. In our review, two studies used a microcosting approach and listed key unit costs and quantities of resources base on hospital records and literature.18,20 The precision in costing has also varied. The least-precise estimates are likely to be based on average per diems (or daily costs), while the most-precise estimates are likely to be based on microcosting, in which each component of the resource used (e.g., laboratory tests or drugs) is estimated and a unit cost is derived for each.

Seven of the studies reviewed herein used QALYs as the health outcome summary measure to compare health outcomes of different types. A QALY is a year lived whilst weighted by the quality of life during that time. A patient's quality of life depends on his or her current health state.13-17,19,20 Fagan et al.,13 Sinclair et al.,14 Chamber et al.,15 Stahl et al.,17 Sandercock et al.,16 and Ehlers et al.20 categorized the health state based on the mRS as follows: no symptoms (R0), no significant disability (R1), minimal disability (R2), moderate disability (R3), moderate to severe disability (R4), severe disability (R5), and dead. Mar et al.19 used the European Quality of Life Questionnaire (EuroQol) with a random sampling of patients to determine values for the utilities of a health state. Moodie et al.18 used DALYs to estimate the health gain attributable to each intervention (Table 1).

All of the studies reviewed herein concluded that thrombolysis treatment can be a cost-effective strategy, especially in the long term. Fagan et al.13 estimated an increase in hospitalization costs of US$ 1.7-million and decreases in rehabilitation and nursing home costs of US$ 1.4-million and US$ 4.8-million, respectively, per 1,000 eligible treatment patients, for a healthcare system that includes care facilities from acute through to long term. Sinclair et al.14 estimated a lifetime cost difference of 3.8-million Canadian dollars (Can$) in favor of tPA versus no tPA (Can$-3,800/patient). In the hypothetical cohort, tPA treatment resulted in 13,130-QALYs, versus 9,670-QALYs with no tPA treatment. This translated into a net benefit of 3,460 additional QALYs per 1,000 patients (i.e., 3.46-QALYs/patient). Chamber et al.15 demonstrated that rtPA in acute ischemic stroke administered within 3-hours of symptom onset could result in 155 additional QALYs per 1,000 patients, at a cost saving of €2.33-million, suggesting that in treated patients, the savings related to disability and long-term care considerably outweigh any potential extra costs of acute therapy, given a broad cost perspective and a time horizon of 5-years. Stahl et al.17 found that the NINDS-compliant strategy including t-PA treatment resulted in an average QALYs value of 3.64, versus 3.63 QALYs for the base case, at an approximate cost of US$ 434/patient.

Sandercock et al.16 showed that over the cohort lifetime, applying rtPA becomes the dominant strategy. Treatment with rtPA was more effective (gain in QALYs of 3.63 per 100 patients treated), less expensive than standard treatment (cost savings of €350,532), and resulted in a reduced cost of €96,565/QALY. Moodie et al.18 estimated that aspirin intervention was effective for reducing stroke mortality, but required a higher overall cost because of increased stroke survival (US$ 1.7-million), whereas rtPA effectively reduced post-stroke disability and led to long-term cost savings (US$ 0.4-million). Mar et al.19 calculated that the ICER obtained with thrombolytic treatment had values of €3,841/QALY and €2,733/QALY for men and women, respectively, without consideration of societal cost, but produced a net saving of €14,913/QALY for men and €16,021/QALY for women when the societal costs were incorporated in the model. Ehlers et al.20 calculated the ICER for 1, 2, 3, and 30-years. In that study, short-term thrombolysis (first year) increased the net health costs, but the picture changed after 2-years when, given the assumptions about effect and costs, thrombolysis became cost-effective. In the long term (30-years), IV thrombolysis with rtPA was the dominant strategy compared with conservative treatment (Table 2).

The overall results are summarized in Table 3. IV rtPA was associated with an acceptable increase in short-term cost (ranges: US$ 36-236/patient, US$ 29,148-55,591/QALY), and a net long-term cost saving that was higher from a societal perspective (ranges: -US$ 12,043 to -US$ 630/patient, -US$ 207,253 to -US$ 21,938/QALY) than from a healthcare-system perspective (ranges: -US$ 5,811 to US$ 5,415/patient, -US$ 41,137 to US$ 4,662/QALY).