Alteplase for treating acute ischaemic stroke

The Appraisal Committee considered evidence submitted by the manufacturer of alteplase and a review of this submission by the Evidence Review Group (ERG). The decision problem addressed by the manufacturer considered whether treatment with alteplase was clinically effective compared with standard medical care (standard medical and supportive management that does not include alteplase) for treating acute ischaemic stroke in adults within 4.5 hours of symptom onset, and whether alteplase treatment was a cost-effective use of NHS resources.

The manufacturer carried out a systematic literature search, which was based on a previously published Cochrane review ('Thrombolysis for acute ischaemic stroke'), but restricted the search to randomised controlled trials of alteplase. For the 0- to 3-hour treatment window, the manufacturer identified no trials other than those included in NICE's previous technology appraisal guidance on alteplase for the treatment of ischaemic stroke from 2007. The trials included NINDS (National Institute of Neurological Disorders and Stroke) 1 and 2, ATLANTIS ('Thrombolysis for acute noninterventional therapy in ischaemic stroke') A and B, and ECASS (the 'European Cooperative Acute Stroke Study') 2. All these trials were multicentre, double-blinded, placebo-controlled randomised controlled trials of alteplase administered at its licensed dose of 0.9 mg/kg. Treatment with alteplase was administered within 3 hours (NINDS 1 and 2), 5 hours (ATLANTIS B) or 6 hours (ATLANTIS A, ECASS 2) of onset of stroke symptoms. Patients were followed up for outcomes for 90 days. The NINDS and ATLANTIS trials were conducted in North America and the ECASS trial in multiple sites in Europe (including the UK), Australia and New Zealand.

The clinical-effectiveness evidence in the manufacturer's submission focused on the extended 3- to 4.5-hour treatment window for which the only directly relevant trial identified by the manufacturer was ECASS 3. Other trials with data indirectly relevant to this treatment window were ATLANTIS A and B and ECASS 2, from which subgroup data specifically for the 3- to 4.5-hour window were used by the manufacturer in sensitivity analyses. The manufacturer noted that this involved stratifying data into subgroups that had not been specified before randomisation. The manufacturer also identified the third International Stroke Trial (IST-3), a randomised open-label blinded endpoint trial in which alteplase was administered within 6 hours of symptom onset. However, the manufacturer commented that this trial was not placebo controlled and therefore did not meet its trial selection criteria, and that no published results were available at the time of submission.

ECASS 3 was a placebo-controlled multicentre trial carried out across 130 sites in 19 European countries. Of the 821 patients 22 were from the UK. Patients were eligible for inclusion if aged between 18 and 80 years, diagnosed with acute ischaemic stroke and able to receive treatment within 3 to 4.5 hours of the onset of stroke symptoms. Before randomisation, brain imaging was used to exclude intracranial haemorrhage. The trial randomly assigned eligible patients to receive 0.9 mg/kg of intravenous alteplase (n=418) or placebo (n=403). Patients were followed up for 90 days for outcomes. Baseline demographic and disease characteristics were similar between participants in the 2 treatment arms, but the initial severity of the stroke (as assessed by the National Institutes of Health stroke scale [NIHSS; a 15-item quantitative measure of stroke-related neurological impairment]) and the proportion of patients with a history of previous stroke were both significantly higher in the placebo arm.

The primary outcome in ECASS 3 was the presence or absence of disability at 90 days as assessed by the modified Rankin scale, which measures the degree of disability or dependence in people who have had a stroke and ranges from 0 (symptom free) to 6 (dead). From intention-to-treat analyses, 52.4% of patients randomised to the alteplase treatment arm had a favourable outcome at 90 days (a score of 0 or 1 [no significant disability]) compared with 45.2% of patients randomised to placebo (odds ratio [OR] 1.34, 95% confidence interval [CI] 1.02 to 1.76, p=0.04). After adjustment for confounding baseline variables (identified as being statistically significant at p<0.10) including treatment arm, NIHSS score, smoking, time from onset of stroke to treatment, and presence or absence of previous hypertension, alteplase remained statistically significantly associated with a favourable outcome (OR 1.42, 95% CI 1.02 to 1.98, p=0.04).

ECASS 3 also reported the composite outcome of death or dependence (defined as a score of 3 to 6 on the modified Rankin scale) at 90 days. There were no statistically significant differences in the number of patients who were dead or dependent between the alteplase and placebo treatment arms (33.5% compared with 38.5%, relative risk [RR] 0.87, 95% CI 0.73 to 1.05).

The ECASS 3 trial also reported as a secondary endpoint a global outcome score at 90 days, which combined a score of 0 to 1 on the modified Rankin scale, a score of 1 on the Glasgow outcome scale (a 5-point measure of brain injury, with 1 indicating independence and 5 death), a score of 95 to 100 on the Barthel index (a 10-item measure of a person's daily function, with higher scores reflecting higher function), and a score of 0 to 1 on the NIHSS. Randomisation to alteplase was associated with a statistically significantly higher probability of achieving a favourable global outcome score (OR 1.28, 95% CI 1.00 to 1.65, p=0.05).

The manufacturer presented a summary of adverse reactions based on intention-to-treat analyses in the ECASS 3 trial at 90 days. Fatal adverse reactions were reported for 7.7% of patients in the alteplase arm and 8.4% of those in the placebo arm. A statistically significantly higher proportion of patients in the alteplase arm had an intracranial haemorrhage (27.0% compared with 17.6%, p=0.001) or a symptomatic intracranial haemorrhage (2.4% compared with 0.3%, p=0.008) compared with the placebo arm. Three patients (0.7%) randomised to the alteplase arm had a fatal intracranial haemorrhage. Investigator-defined drug-related adverse reactions were reported for 23.9% of patients in the alteplase treatment arm and 6.9% of patients in the placebo arm. Other serious adverse reactions were reported for 25.1% of patients in the alteplase arm and 24.6% of those in the placebo arm.

The manufacturer conducted meta-analyses to calculate the relative risks associated with alteplase for all-cause mortality within 90 days, death or dependence within 90 days, and symptomatic intracranial haemorrhage within 10 days for each of the 3 treatment windows (0 to 3 hours, 3 to 4.5 hours, and 0 to 4.5 hours). The manufacturer presented results from both fixed and random-effects models. For the 0- to 3-hour window, the manufacturer used data from ECASS 2 and NINDS and used the relative risks from ECASS 3 for the 3- to 4.5-hour window. For the 0- to 4.5-hour window, the manufacturer used data from ECASS 2 (0 to 3 hours), ECASS 3 (3 to 4.5 hours) and NINDS (0 to 3 hours). Heterogeneity between the 3 studies was low for the outcomes of all-cause mortality and death or dependence, but moderate for the outcome of symptomatic intracranial haemorrhage.

The manufacturer compared all-cause mortality at 90 days for the 2 treatment arms. No statistically significant difference was observed between alteplase and placebo for the 3- to 4.5-hour (RR 0.82, 95% CI 0.50 to 1.33, p=0.42), the 0- to 3-hour (RR 1.05, 95% CI 0.55 to 2.03, p=0.88) or the 0- to 4.5-hour (RR 0.89, 95% CI 0.67 to 1.18, p=0.41) treatment windows. For the outcome of death or dependence at 90 days, the manufacturer reported no statistically significant difference between alteplase and placebo for the 3- to 4.5-hour window (RR 0.87, 95% CI 0.73 to 1.05, p=0.14). However, a statistically significant difference in favour of alteplase was reported for both the 0- to 3-hour window (RR 0.81, 95% CI 0.72 to 0.92, p=0.002) and the 0- to 4.5-hour window (RR 0.83, 95% CI 0.75 to 0.92, p<0.001). For the outcome of symptomatic intracranial haemorrhage occurring within 10 days, patients randomised to receive alteplase within the 3- to 4.5-hour window had a statistically significant higher risk (RR 4.82, 95% CI 1.06 to 21.87, p=0.04), with similar results for the 0- to 4.5-hour treatment window (RR 4.18, 95% CI 1.39 to 12.53, p=0.01). For the 0- to 3-hour window, the manufacturer reported a higher risk of symptomatic intracranial haemorrhage among patients randomised to alteplase that was not statistically significant (RR 3.94, 95% CI 0.61 to 25.47, p=0.15).

The manufacturer conducted a systematic review of published cost-effectiveness analyses but did not identify any studies that evaluated the cost effectiveness of alteplase for the treatment of acute ischaemic stroke within 3 to 4.5 hours of onset of symptoms. Instead, the manufacturer adapted a published cost-effectiveness analysis (Sandercock et al. 2002) relevant to the decision problem from the previous guidance on alteplase in acute ischaemic stroke (NICE technology appraisal guidance 122) and used this as part of its submission.

The manufacturer developed a Markov model simulating patients with acute ischaemic stroke who do or do not receive alteplase within 4.5 hours of onset of symptoms. Patients were modelled through 3 possible health states: independent, dependent and dead. The independent state was defined by a modified Rankin scale score of 0 to 2 and the dependent state by a modified Rankin scale score of 3 to 5. The model had 3 time phases: from 0 to 6 months when the model assumed the treatment effect of alteplase was complete at 90 days and maintained at 6 months; from 6 to 12 months when the model assumed no further treatment effect; and beyond 12 months when the model also assumed no further treatment effect from alteplase. However, beyond 12 months the model assumed that people in the dependent or independent states could have a recurrent stroke. The model also assumed that people in the dependent state at 12 months and beyond do not move to an independent state, and that people in the independent state at 12 months and beyond do not move to a dependent state unless they survive a recurrent stroke. The model assumed a lifetime horizon with a cycle length of 6 months for the first 12 months, followed by cycles of 12 months thereafter.

The manufacturer chose a population for the model based on SITS-MOST ('Safe implementation of thrombolysis in stroke-monitoring study'), a European observational study of patients receiving alteplase. The manufacturer considered that this study population represented the mean age (68 years) and gender distribution (39.8% female) of patients who would receive alteplase in clinical practice in England and Wales.

For the first phase (0 to 6 months) of the manufacturer's economic model, the size of the effect of treatment with alteplase was informed by the manufacturer's meta-analyses for the 3 treatment windows as described in section 3.8. For the standard treatment arm, the proportion of people in each health state (39.53% independent, 32.56% dependent and 27.91% dead) was informed by the Lothian stroke registry, a registry in Edinburgh, Scotland, of 1,779 inpatients with suspected or confirmed stroke from 1989 to 2000. The manufacturer also provided an alternative distribution of the proportion of people who received standard treatment in each health state from the placebo arm of the ECASS 3 trial at 90 days. This alternative distribution (61.54% independent, 30.27% dependent and 8.19% dead) was used by the ERG in exploratory sensitivity analyses. The manufacturer then used the relative risks of death and death or dependence to calculate the distribution of people in the alteplase arm across the independent, dependent and dead states at the end of the first phase (6 months). The manufacturer used the relative risk of death to estimate the proportion of people who would die and therefore enter the dead state during the first phase. The proportion of people in the dependent state at 6 months was calculated as the difference between the estimated proportion of people who were dead or dependent, and the estimated proportion who were dead. The manufacturer assumed in the model that a symptomatic intracranial haemorrhage had a cost impact (because it required a further diagnostic computed tomography [CT] scan), with the health consequences being captured in the outcome of death or dependence from the clinical trials.

For the second phase of the model (6 to 12 months), the manufacturer assumed that people could move from the independent or dependent state to any other health state with equal probabilities for both treatment arms. These transition probabilities were based on the Lothian stroke registry. For the third phase of the model (beyond 12 months), the annual risk of a recurrent stroke (0.05), and the associated risk of mortality (0.25), were taken from the Lothian stroke registry. To estimate the mortality risk for people who did not have another stroke, the manufacturer took data from the Office for National Statistics life tables for England and Wales and adjusted them upward by a factor of 2.3 (taken from the Perth Community Stroke Study) to reflect the higher mortality rates among people who have had a stroke compared with the general UK population.

The manufacturer conducted a literature review to identify appropriate utility values for the independent and dependent states in the model. The manufacturer did not identify any relevant utility values additional to those used in NICE technology appraisal guidance 122 on alteplase in acute ischaemic stroke within the first 3 hours after symptom onset. The manufacturer's submission for this appraisal identified 1 trial (Dorman et al. 1997) that collected EQ-5D utility values in a sample of 147 patients from the Lothian stroke registry. This trial provided utility values of 0.74 (95% CI 0.69 to 0.79) for the independent state and 0.38 (95% CI 0.29 to 0.47) for the dependent state. The model assumed that these utility values remained fixed over time unless a person had a recurrent stroke which resulted in dependence, and thus a move from the independent to the dependent state.

The model included drug acquisition and administration costs as well as the costs of acute and long-term stroke care. The cost of alteplase was based on the mean body weight (76 kg) of patients in the 3- to 4.5-hour cohort from the SITS-MOST trial. Based on the recommended dose of 0.9 mg/kg, the average dose was 68.4 mg, resulting in a total estimated cost of £480 (£300 per 50-mg pack and £180 per 20-mg pack). Administration costs associated with alteplase of £1,316 per patient were based on estimates of extra staff time in the trial by Sandercock et al. (2002). For people in either treatment arm who had a symptomatic intracranial haemorrhage, the model included a one-off cost of £100 for an additional CT scan. For all health states in the model, the manufacturer applied annual costs specific to the state (adjusted for inflation to 2012 to 2013 prices) adapted from a study by Youman et al. (2003), which calculated the costs of acute events and long-term stroke care.

The manufacturer's base-case deterministic cost-effectiveness analysis for the 0- to 4.5-hour window estimated an incremental cost-effectiveness ratio (ICER) of £2,441 per quality-adjusted life year (QALY) gained for alteplase compared with standard care (incremental costs £811; incremental QALYs 0.333). The probabilistic cost-effectiveness analysis resulted in an ICER of £2,296 per QALY gained.

The manufacturer conducted a number of one-way sensitivity analyses on various parameters in the model, including the relative risks associated with alteplase of death or of death or dependence, the risk of recurrent stroke and mortality irrespective of previous treatment, the dose of alteplase treatment, the annual costs of care in the dependent and independent states, the cost of a fatal stroke, and the utility values. The results of these one-way sensitivity analyses indicated that the ICERs were robust to changes in most input parameters, except for the relative risks of death and death or dependence for treatment with alteplase applied in the first phase of the model. When the manufacturer used the upper limit of the 95% confidence interval for the relative risks of death (1.18) and death or dependence (0.92), alteplase treatment led to a very small loss in QALYs at a decreased cost compared with standard care, resulting in £44,342 saved per QALY lost. The manufacturer also conducted additional deterministic sensitivity analyses, which included different scenarios using additional clinical efficacy data from unplanned subgroup analyses of the ATLANTIS A and B and ECASS 2 trials for 3 to 4.5 hours, and weighted the results based on the assumption that in UK clinical practice a higher proportion of patients are treated during the 0- to 3-hour window (76%) than during the 3- to 4.5-hour window (24%). Using these analyses, alteplase either dominated placebo (that is, was both less costly and more effective) or had an ICER below £2,000 per QALY gained. Results of the probabilistic sensitivity analysis showed that alteplase had a high probability (above 90%) of being cost effective at a level of £20,000 to £30,000 per QALY gained.

The manufacturer's base-case deterministic cost-effectiveness analysis for the 3- to 4.5-hour window resulted in an ICER of £6,272 per QALY gained for alteplase compared with standard care (incremental costs £2,068; incremental QALYs 0.33). The probabilistic cost-effectiveness analysis resulted in an ICER of £6,169 per QALY gained. The results of the one-way sensitivity analyses were similar to those for the 0- to 4.5-hour window, indicating that the ICERs were robust to changes in most input parameters, except for changes in the relative risks of death and death or dependence for treatment with alteplase. In an additional sensitivity analysis, the manufacturer pooled 3- to 4.5-hour efficacy data from the ECASS 2 and ATLANTIS trials with the ECASS 3 data. This sensitivity analysis resulted in an ICER of £5,631 per QALY gained for alteplase compared with standard care.

For the 0- to 3-hour treatment window, the manufacturer presented cost-effectiveness results similar to those presented in the previous guidance on alteplase in acute ischaemic stroke (NICE technology appraisal guidance 122); that is, alteplase dominated standard care, resulting in lower costs and more QALYs for both the deterministic and the probabilistic analyses.

The ERG considered that the clinical-effectiveness evidence submitted by the manufacturer was of good quality. The ERG noted that the patients randomised to the alteplase arm of both the NINDS trial (which provided clinical evidence for the 0- to 3-hour window) and the ECASS 3 trial had strokes that were less severe on average, which in turn may have biased the results in favour of alteplase. However, the manufacturer also presented adjusted analyses for the primary outcome (disability at 90 days) from the ECASS 3 trial. The ERG considered that the meta-analytical approach by the manufacturer was appropriate. The ERG agreed with the manufacturer that data derived from unplanned subgroup analyses from the ATLANTIS A and B and ECASS 2 trials, in which treatment with alteplase was administered up to 6 hours from onset of symptoms, should not be included in the base-case meta-analyses. The ERG also noted that heterogeneity between the studies included in the base-case meta-analyses for the 0- to 4.5-hour window for the outcomes of death and death or dependence was low and not statistically significant.

The ERG commented that the manufacturer submitted an economic model that was in line with the decision problem defined in the scope and closely adhered to the NICE reference case requirements for economic analysis. The ERG commented that the manufacturer provided a reasonable strategy for searching the literature for existing cost-effectiveness studies, although it did not explicitly state its exclusion criteria. The ERG stated that it was appropriate for the manufacturer to conduct separate analyses for patients eligible for treatment within the 0- to 3-hour window and the 3- to 4.5-hour window. The ERG noted that the utility values for the dependent and independent states did not allow for any decreases in health-related quality of life over time, which may have overestimated the lifetime QALYs accrued in the independent state. The ERG stated that, although this may have biased the QALY gains in favour of alteplase, the manufacturer's economic model was not sensitive to changes in the utility values, and so the effect of adjusting these values over time in the model was likely to be small. The ERG noted that in the probabilistic sensitivity analysis, the manufacturer sampled independently the relative risks for death and death or dependence associated with treatment with alteplase, which had a marked impact on the ICERs in the one-way sensitivity analyses. The ERG noted that this did not take into account the likely correlation between the outcomes of death and death or dependence, and that the probabilistic sensitivity analysis might not provide an accurate description of the uncertainty around the mean costs and QALYs, although the ERG did not expect it to have a large impact on the ICER.

The ERG conducted an exploratory sensitivity analysis by replacing the proportion of people in the 3 health states (dependent, independent and dead) at the end of the first phase (0 to 6 months) taken from the Lothian stroke registry with those observed in the ECASS 3 trial population. Because mortality rates were lower in the ECASS 3 trial, a higher proportion of patients were in the independent state (61.54%) and a lower proportion in the dead state (8.19%). This sensitivity analysis resulted in an ICER of £4,451 per QALY gained for alteplase compared with standard care for the 0- to 4.5-hour window (incremental costs £698; incremental QALYs 0.157).

Full details of all the evidence are in the manufacturer's submission and the ERG report.