Dasatinib, high-dose imatinib and nilotinib for the treatment of imatinib-resistant chronic myeloid leukaemia: a systematic review and economic evaluation

Population

People with imatinib-resistant CML in the chronic, accelerated or blast-crisis phases were eligible for inclusion.

Interventions

Studies of dasatinib, nilotinib and high-dose imatinib were considered for inclusion.

Comparators

Potential comparators were dasatinib, nilotinib and high-dose imatinib, hydroxycarbamide (hydroxycarbamide), interferon alfa, acute leukaemia-style chemotherapy, allo-stem cell transplantation, standard-dose imatinib and best supportive care, depending on the phase of CML.

Outcomes

Studies reporting one or more of the following outcome measures were eligible for inclusion: treatment response rates [including molecular, cytogenetic and haematological responses (HRs)]; time to, and duration of, response; overall survival; event-free survival; progression-free survival; adverse effects of treatment; HRQoL; time to treatment failure; costs and cost-effectiveness.

Study design

The hierarchy of evidence was used to determine the inclusion of trials and studies into the review. Randomised controlled trials (RCTs) or prospective non-randomised comparative studies, where adequate matching was considered to have been achieved, were eligible for inclusion. Where no such evidence existed, single-arm cohort studies were included.

Studies published as abstracts or conference presentations were eligible to be included only if sufficient details were presented to allow an appraisal of the methodology and the assessment of results to be undertaken.

For the systematic review of cost-effectiveness, studies were eligible for inclusion if they reported the results of full economic evaluations, i.e. cost-effectiveness analyses, cost–utility analyses or cost–benefit analyses.

Studies were excluded if participants were aged < 18 years, did not have CML or were imatinib naive or imatinib intolerant. Studies of high-dose imatinib [> 400 mg b.i.d. (twice daily) in chronic phase] as first-line treatment were also excluded.

Design and characteristics of included studies

One published update of a RCT and three new single-arm cohort studies met the inclusion criteria (Figure 1). Data extraction forms for these studies can be seen in Appendix 4. The RCT (Kantarjian and colleagues6) compared high-dose imatinib (600 or 800 mg/day) with dasatinib (140 or 180 mg/day) and was reported in detail in the context of its dasatinib intervention in the PenTAG AR2 (see p. 57, p. 59 and pp. 79–89). The update of this RCT was published in 2009,7 and longer follow-up from the dasatinib arm of this study, as well as data from the high-dose imatinib arm, are included in the present review. However, methodological flaws associated with this RCT render it of limited value as a comparative study (see PenTAG AR,2 section 3.2.4, p. 90), and the PenTAG AR2 presented the dasatinib arm as non-comparative evidence. In line with this, data from the dasatinib and high-dose imatinib arms of the RCT are presented separately and are not compared in the present systematic review (see Chapter 3, Critical appraisal of included evidence)

The single-arm cohort studies each had a single high-dose imatinib arm. In one study, by Rajappa and colleagues,8 all participants received imatinib at 800 mg/day, whereas in the remaining studies the imatinib dose varied from 600 to 800 mg/day according to whether individual participants met criteria for dose escalation or reduction. The interventions in the RCT and observational studies are summarised, respectively, in Tables 2 and 3, and can be viewed in detail in Appendix 4. None of the studies reported whether or not participants received any treatment concurrent with imatinib.

The designs of the RCT and single-arm cohort studies are summarised, respectively, in Tables 4 and 5. The RCT was conducted in 58 centres in 23 countries, including the UK, Europe, the Russian Federation and Asia. The three single-arm cohort studies were conducted in single countries: Republic of Korea, Italy and India. Apart from the Korean study, which involved 19 centres, the number of centres was small or unclear (Table 5). The studies included only participants with chronic-phase CML, except for the single-arm cohort study by Koh and colleagues,10 which also included very small numbers of participants with accelerated phase and blast-crisis phase (Table 6). Inclusion and exclusion criteria were reported in detail for the RCT (Table 4), but only briefly for the single-arm cohort studies (see Table 5). The RCT required participants to be at least 18 years of age and have ‘adequate hepatic and renal function’, and excluded those with BCR–ABL (oncogene fusion protein consisting of BCR and ABL genes) mutations known to be particularly resistant to imatinib. The single-arm cohort study by Koh and colleagues10 required participants to be aged 15–75 years with ‘adequate organ function’. All other inclusion and exclusion criteria reported in the RCT and single-arm cohort studies were based on cytogenetic or molecular aspects of CML or imatinib dosing.

Failure on standard-dose imatinib was defined in terms of resistance and suboptimal cytogenetic, haematological and molecular response. None of the studies defined imatinib failure as intolerance (Table 6). The criteria used to define imatinib failure were slightly different in each of the four studies (Table 7).

Baseline characteristics of the participants in the RCT and cohort studies are summarised in Table 6. For high-dose imatinib, the proportion of male participants in the RCT (45%) was lower than in the three single-arm cohort studies (70–71%). Across the four studies,7–10 the participants ranged in age from 18 to 85 years. The cohort study by Rajappa and colleagues8 included younger participants (mean age 35.7 years) than the three other studies (median age 49–51 years). Duration of CML from diagnosis to imatinib therapy ranged from 14 to 133 months in the RCT, but was not reported in any of the single-arm cohort studies. Baseline genetic and haematological data were not consistently reported across the four studies and are therefore difficult to compare. Only the RCT provided baseline data on the proportion of participants with a major cytogenetic response or a complete haematological response (CHR). The proportion of participants with BCR–ABL mutations at baseline was slightly lower in the RCT high-dose imatinib participant group (22.4%) than in the only single-arm cohort study, by Rajappa and collegues,8 that provided comparable data (32.2%).

The previous imatinib therapy received by participants in each of the four high-dose imatinib studies is summarised in Table 8. The duration of prior imatinib therapy ranged from 0.6 to 70 months (median 18 to 36 months) in the single-arm cohort studies, and from < 1 year to > 3 years in the RCT (not reported more precisely). 7 In two of the single-arm cohort studies all participants had previously received only standard-dose imatinib (400 mg/day). 8,9 In the remaining single-arm cohort study the majority of participants (90%) had received 400 mg/day, although 10% (the accelerated- and blast-phase participants) received 400–600 mg/day. 10 In the RCT the majority of participants (69%) had received 600 mg/day and the remainder (29%) received 400 mg/day (one participant received 500 mg/day). In addition to imatinib, the majority of participants in the RCT had received hydroxycarbamide or anagrelide (93.9%) and/or interferon alfa (67.3%), with some (36.7%) having received chemotherapy or, in a minority of cases (4.1%), stem cell transplantation. Two single-arm cohort studies reported that, in addition to imatinib, participants had previously received interferon alfa (one study, 29.7%) or hydroxycarbamide (one study, % not stated) only.

The characteristics of the studies reviewed are shown in the PenTAG AR2 (see pp. 57–75).

Critical appraisal of included evidence

A summary of the critical appraisal of the RCT is provided in the PenTAG AR2 (see section 3.2.3.1, p. 84). As noted in the PenTAG AR,2 the RCT is flawed, which has implications for interpreting effectiveness and safety information. The updated publication by Kantarjian and colleagues7 provides new information on two aspects of the RCT methodology that were not reported in the previous publications and which therefore do not currently appear in the PenTAG AR:2

• Kantarjian and colleagues7 explained how the sample size was calculated. However, the approach, which is based on arbitrary maximum widths of confidence intervals (CIs) for the primary outcome, was not considered in relation to the statistical power of the trial. This explanation appears to be an attempt to justify the sample size retrospectively.

• Kantarjian and colleagues7 stated that dasatinib and high-dose imatinib groups were stratified by study site and cytogenetic response (CyR) on previous imatinib.

In the PenTAG AR,2 analyses conducted in the RCT were considered appropriate. Although the statistical methods used were generally appropriate, the way in which they were applied does have serious shortcomings. Specifically, the analyses were not planned a priori and were not adjusted for multiple comparisons. However, data from the individual arms are not compared in this report.

Overall, the new information available from the Kantarjian and colleagues publication7 does not alter the judgement that the RCT was substantially flawed. Of particular relevance to the high-dose imatinib arm of the RCT was that 80% of high-dose imatinib participants with inadequate responses crossed over to the dasatinib arm at a median time of 13 weeks (range 1–68 weeks). Conversely, 20% of participants with inadequate responses to dasatinib crossed over to the high-dose imatinib arm at a median time of 28 weeks (range 1–56 weeks). As a result, outcomes for the high-dose imatinib arm reported at a median of 26 months would include an unknown (not reported) proportion of participants who had predominantly received dasatinib. Interpretation of the outcome data for high-dose imatinib in the RCT is also difficult because follow-up times varied considerably and were reported only as the median and range.

Critical appraisal of the three single-arm cohort studies8–10 of high-dose imatinib is summarised in Table 9. The assessment criteria in Table 9 reflect aspects of study design relevant to interpretation of generalisability and some types of bias, which may help to assess the relative strengths and weaknesses of the individual studies. All three studies have risk of selection bias owing to a lack of any randomised procedures for allocation to study groups, and risk of performance bias owing to a lack of allocation concealment and blinding.

The reporting of these single-arm cohort studies was generally superficial. 8–10 Only the study by Koh and colleagues10 could be clearly identified as a prospective study, although the other two studies were judged to be prospective by reviewers. 8,9 Only the study by Rajappa and colleagues8 reported whether or not participants were recruited consecutively. Breccia and colleagues9 and Rajappa and colleagues8 failed to adequately report the inclusion criteria for their studies, which is a major impediment to interpreting generalisability and selection bias. Although generalisability of the study by Koh and colleagues10 appears to be stronger than for the other two studies, none of the single-arm cohort studies was conducted in the UK. It is, therefore, unclear how relevant the findings from these studies would be to UK patients with chronic-phase CML.

In these single-arm cohort studies the participants within a study did not all receive exactly the same intervention, as dose escalations occurred at different times for individual participants, or subgroups of participants received different dose changes. It is unclear whether or not any participants received concurrent therapies alongside high-dose imatinib, as this was not reported in any of the three studies. None of the studies reported explicitly whether or not all participants allocated to treatment were analysed and whether or not the analyses included attrition [intention-to-treat (ITT) approach]. Only one of the studies, by Koh and colleagues10 adequately reported participant attrition.

Overall, owing to the inherent limitations of a single-arm study design, compounded by generally poor reporting of the methodology, the three single-arm cohort studies of high-dose imatinib appear to be at high risk of bias and limited or unclear relevance to CML patients in a UK setting.

Relationship of identified evidence to research question

The research questions could not be directly addressed by the PenTAG AR2 (see section 3.2.4, p. 90), as there was no comparative evidence available. Similarly, in our update of the PenTAG AR2 for those with imatinib-resistant CML we have not identified any comparative evidence for any of the three interventions of interest. Therefore, caution continues to be recommended in the interpretation of the evidence now presented.

Evidence reported in this review is relevant to patients with chronic-phase CML only. Owing to the paucity of data the review has been unable to consider whether or not the level of previous response to imatinib has any bearing on outcome, and the evidence does not allow the adoption of an early stopping rule to be considered.

Cytogenetic response

Complete cytogenetic response improved from 39.6% at median 15 months’ follow-up to 43.6% at median 26 months’ follow-up (Table 10). Major cytogenetic response was similar between the two follow-up periods (52.5% at 15 months6 and 53.5% at 26 months7). Major cytogenetic response was similar between patients with (34/62, 55%) and without (20/39, 51%) a previous CyR on standard-dose imatinib.

Duration of major cytogenetic response

The 2007 publication reported the probability of a maintained response at 1 year as being 0.98. 6 With longer follow-up, the proportion with a maintained major cytogenetic response at 18 months was 90% (95% CI 82% to 98%). 7

Major molecular response

Kantarjian and colleagues7 reported a major molecular response (MMR) in 28.7% of participants (29/101) receiving dasatinib, and in 63.4% (28/44) of those who had a complete cytogenetic response and a molecular response assessment (MMR: generally defined as ≥ 3-log reduction in the level of BCR–ABL transcripts or a BCR–ABL ratio of ≤ 0.05%).

Time to treatment failure and proportion without treatment failure

Median time to treatment failure was not reached with dasatinib in the 2007 publication by Kantarjian and colleagues,6 but was not reported in the 2009 study with longer follow-up. 7 However, the authors reported that the estimated proportion of participants without treatment failure at 24 months was 59%. 7

Adverse events

The PenTAG AR2 (see p. 121) stated that in most of the included evidence, neutropenia and thrombocytopenia each affected in the order of 50% ± 10% of individuals taking dasatinib. The proportion of individuals affected by neutropenia in the RCT is slightly higher with longer follow-up7 (Table 11).

The PenTAG AR2 (see p. 122) described the AEs (of any grade) most commonly reported by its included studies as diarrhoea, dyspnoea, fatigue, headache, nausea, pleural effusion, and rash, at frequencies in the range 10%–40%. The 2009 update7 also reports fluid retention (39%), bleeding (18%), infection (14%) and upper respiratory tract infection or inflammation (11%), and an increase in superficial oedema from 15% to 20% (Table 12).

The PenTAG AR2 (see p. 122) reported that grades 3–4 AEs appeared to be fairly rare in the included studies, with only dyspnoea and pleural effusion occurring in more than 5% of participants in any of the included studies. However, grades 3–4 fluid retention occurred in 7% of individuals in the 2009 update publication7 (Table 13).

Rates of treatment discontinuation due to AEs in the four studies included in the PenTAG AR2 (see p. 124) that reported this outcome were described as ranging from approximately 5% to 15%. Discontinuations due to AEs increased from 15.8% in the 2007 Kantarjian and colleagues publication6 to 22.8% in the 2009 publication7 (Table 14).

Summary of effectiveness of dasatinib

• No new studies of dasatinib were identified by the updated searches.

• Additional follow-up data for some outcomes were available for the RCT by Kantarjian and colleagues 6 and included in the PenTAG AR. 2

• The RCT was methodologically flawed, with a high level of crossovers between treatment arms. As such, the individual treatment arms were considered separately as non-comparative evidence. This is in line with the approach taken by the PenTAG AR. 2

• Complete cytogenetic response improved slightly from 39.6%6 to 43.6%,7 and major cytogenetic response was similar (52.5%6 to 53.5%7) with longer follow-up.

• A MMR was reported in 28.7% of participants. 7

• Additional AEs were reported with longer follow-up (fluid retention, bleeding, infection, upper respiratory tract infection or inflammation), and grades 3–4 fluid retention occurred in 7% of individuals in the update paper. 7

Cytogenetic response

Table 15 provides a summary of the available data detailing CyR to high-dose imatinib in CML (see Appendix 4 for further details). 6–10 All four studies were in participants with chronic-phase CML, with the exception of one study which also included a small number of participants in accelerated-phase CML (n = 3) and blast-crisis CML (n = 4). 10 Three studies reported complete, partial and major cytogenetic response rates or provided enough information to enable the deduction of each, whereas one study reported only complete cytogenetic response and major cytogenetic response. Minor and minimal responses were not reported by these studies. The definition of response for each category was consistent across studies.

It should be noted that some participants already had some degree of CyR at baseline. Rajappa and colleagues8 reported that 44.5% of participants had achieved major cytogenetic response (although it is unclear whether this is the proportion at study entry)8 and 43.7% of participants were in partial cytogenetic response in the study by Koh and colleagues. 7 None of the participants in the high-dose imatinib arm of the RCT by Kantarjian and colleagues7 was in major cytogenetic response at baseline. However, 44% (15/34) of participants with a previous CyR on standard-dose imatinib achieved a major cytogenetic response on high-dose imatinib, whereas 7% (1/15) of those without a previous CyR on standard-dose imatinib achieved a major cytogenetic response. CyR was not reported at baseline by Breccia and colleagues. 9

Complete cytogenetic response rates ranged from 18.4%7 to 36.4%. 9 Major cytogenetic response rates ranged from 32.7%7 to 63.5%. 9

Duration of major cytogenetic response

Kantarjian and colleagues7 reported that 74% (95% CI 49% to 100%) of individuals maintained their major cytogenetic response at 18 months. This outcome was not reported by the other three studies. 8–10

Haematological response

Only two of the studies reporting CHR provided a definition,7,9 and there were minor differences between these definitions (Table 16).

Table 17 provides a summary of the available data detailing HR to high-dose imatinib in CML. Three of the included studies reported CHR, with response rates ranging from 55.5% (18-month follow-up)8 to 91.8% (36-month follow-up). 9 It should be noted that 55.1% of participants in the high-dose imatinib arm of the RCT by Kantarjian and colleagues were in CHR at study entry,7 but this was not reported by the other two studies. The CHR in participants without CHR at baseline was 16/22 = 72%. 7

Duration of major or complete haematological response

These were not reported by the included studies.

Molecular response

Three studies reported molecular response and, as can be seen in Table 18, the definitions of molecular response varied between the studies. A summary of molecular response to high-dose imatinib can be seen in Table 19. Kantarjian and colleagues7 reported a MMR in 12.2% of participants receiving high-dose imatinib, and in 55.6% (5/9) of those who had a complete cytogenetic response and a molecular response assessment. A complete molecular response was found in 13.5% of participants by Breccia and colleagues,9 whereas Koh and colleagues reported that 56.3% of participants achieved a molecular reduction > 50% within 6 months. 10

Time to treatment failure

Median time to treatment failure was 18.0 months (range not reported) for all participants (n = 71) in the study by Koh and colleagues. 10 For chronic phase participants (n = 64) it was 27 months, for accelerated-phase participants (n = 3) it was 2.5 months and for blast-crisis-phase participants (n = 4) it was 4.0 months. 10 Median time to treatment failure was 3.5 months (95% CI 3.3 to 3.8 months) with high-dose imatinib in the 2007 publication by Kantarjian and colleagues,6 but was not reported in the 2009 study with longer follow-up. 7 However, the authors reported that the estimated proportion of participants without treatment failure at 24 months was 18%. 7

Progression-free or event-free survival

Three of the included studies reported progression-free survival (progression-free survival) or event-free survival. 6–9 These are reported together, as the definitions, although differing slightly (Table 20), appear to be measuring similar outcomes. Table 21 provides a summary of the available data detailing progression-free survival with high-dose imatinib. Rajappa and colleagues8 reported estimated event-free survival of 34% at 2 years, whereas a higher estimated progression-free survival is reported by the other two studies (65%7 and 87%9).

Overall survival

Two studies reported overall survival (Table 22) of individuals in chronic-phase CML,8,9 although the definitions differ. Breccia and colleagues9 defined overall survival as the time from diagnosis to death or date of last follow-up, and reported an estimated 2-year overall survival of 85%. Rajappa and colleagues8 defined overall survival as time from dose escalation to death owing to any cause, and reported an estimated 2-year overall survival of 93%.

Adverse events

Haematological AEs were reported by all included studies (Table 23). Breccia and colleagues9 reported a low proportion of participants experiencing anaemia and neutropenia (grade not reported). The other three studies reported grades 3–4 haematological AEs, with anaemia occurring in 8%7 to 30%8 of participants, neutropenia in 18%10 to 39%7,8 of participants, leucopenia in 16%7 to 31%8 of participants, and thrombocytopenia in 0%10 to 21%8 of participants.

The most commonly reported AEs of any grade were anorexia, diarrhoea, fatigue, muscle spasms, musculoskeletal pain, nausea, superficial or peripheral oedema and rash (Table 24); however, the reported proportions varied between the studies. Grades 3 and 4 AEs appeared to be fairly rare, with none occurring in more than 4% of the cohort in the two studies reporting this outcome. 7,10

Treatment discontinuation due to AEs was reported by three6–8,10 of the four studies and ranged from 0%10 to 20.4%7 (Table 25).

Summary of effectiveness of high-dose imatinib

• Four studies (one RCT and three single-arm cohort studies) provided data on the effectiveness of high-dose imatinib.

• The RCT had serious methodological flaws. A high proportion (80%) of participants in the high-dose imatinib arm of the RCT crossed over to the alternative treatment after a median duration of 13 weeks. As such, the individual treatment arms were considered as non-comparative evidence. This is in line with the approach taken in the PenTAG AR. 2

• The single-arm studies appear to be at high risk of bias and may be of limited relevance to CML patients in a UK setting.

• Complete cytogenetic response was achieved by 18–36% of individuals.

• Major cytogenetic response was achieved by 33–64% of individuals.

• One study reported that around three-quarters of individuals maintained their major cytogenetic response at 18 months.

• Complete haematological response was achieved by 56–92% of individuals.

• Event-free survival of 2 years or more occurred in only 34% of individuals in one study. progression-free survival was estimated as 65–87% in two other studies.

• Overall survival was reported by two studies, which found that 85–93% of people are expected to survive 2 years or more.

• Haematological AEs (grades 3–4) occurred in up to 40% of individuals.

• Non-haematological events also occurred, with anorexia, diarrhoea, fatigue, muscle spasms, musculoskeletal pain, superficial oedema and rash reported in varying proportions.

• Grades 3–4 non-haematological AEs were fairly rare, with none occurring in more than 5% of individuals.

• Between 0% and 20% of study participants discontinued high-dose imatinib owing to AEs.

• These results should be interpreted with caution owing to the methodological limitations of the included studies.

Methods of the systematic review

The methods used for the systematic review, including the search strategy, inclusion criteria and data extraction, are shown in Chapter 2. Quality assessment was undertaken using a critical appraisal checklist adapted by the review authors from checklists by Philips and colleagues,12 Drummond and colleagues13 and the NICE reference case requirements. 14

Quantity of existing cost-effectiveness literature

A total of 154 potentially relevant references were identified by the cost-effectiveness searches. Of these, the full text of one paper was retrieved and this study met the a priori inclusion criteria. A summary of the selection process is presented in Figure 2. Full data extraction of the study is given in Appendix 5.

FIGURE 2.

Flow chart of identification of studies for inclusion in the review of cost-effectiveness.

Modelling approach of Ghatnekar and colleagues

A Markov cost-effectiveness model was developed to calculate the costs and effects associated with dasatinib treatment compared with high-dose imatinib among patients who were confirmed to be resistant to lower doses (≤ 600 mg) of imatinib (Figure 3). The model is an adaptation to Swedish treatment practice of a model developed for the Scottish Medicines Consortia. It uses monthly cycles with probabilities of a health state change, and all patients were assumed to start treatment in chronic phase. The response to treatment after an initial 12-week treatment period determines the disease progression within the four health states: chronic phase, accelerated phase, blast-crisis phase and death (from either CML- or non-CML-related causes). At each monthly cycle the patients face the probability of staying in the same health state or moving to the next. Progression data are taken from several published sources. Patients enter the model at the age of 60 years.

FIGURE 3.

Markov model structure starting with imatinib-resistant chronic-phase CML (adapted from Ghatnekar et al. 15). CCyR, complete cytogenetic response (typically defined as no Ph+ chromosomes in metaphase in bone marrow); CP-CML, chronic-phase chronic myeloid leukaemia; PCyR, partial cytogenetic response (typically defined as between 1% and 35% of Philadelphia-positive (Ph+) chromosomes in metaphase in bone marrow).

Assumptions made by Ghatnekar and colleagues

• The better the initial response to treatment, the slower the expected cohort disease progression.

• It is not possible to move from chronic phase to blast-crisis phase directly; the probability of CML-related death is dependent on the health state and the treatment response of the patient.

• Utilities are assumed to be the same for both study groups.

• Adverse event rates are limited to the first month only; no disutility weights are used for AEs and patients are assumed to continue with study medication.

• Patients are treated until disease progression.

Effectiveness of intervention

The effectiveness outcome of the interventions used in the model was best initial response rate, taken from a 12-week head-to-head clinical trial of dasatinib versus high-dose imatinib (Kantarjian and colleagues,6 see Chapter 3, Effectiveness of dasatinib: update of Peninsula Technology Assessment Group assessment report). The proportions of patients in the dasatinib group and the high-dose imatinib group, respectively, were 7.9% and 18.4% for no response; 57.4% and 53.1% for CHR; 13.9% and 20.4% for partial cytogenetic response; and 20.8% and 8.2% for complete cytogenetic response.

Estimation of quality-adjusted life-years

Utility weights for each health state were reported to have been elicited from a time trade-off (TTO) technique using the European Quality of Life-5 Dimensions (EQ-5D) instrument among 100 laypersons in the UK and applied to both the dasatinib and imatinib arms by Levy and colleagues. 16 For the base-case analysis these were 0.90 for chronic phase responder, 0.72 for chronic phase non-responder, 0.53 for accelerated phase and 0.29 for blast-crisis phase.

Estimation of costs

The average number of resources used per patient and month in each health state was elicited from two Swedish clinical haematologists. Direct health care-related costs for drugs and other health-care resources were taken from published Swedish statistics. Costs for study medication are added each month the patient is in chronic phase. Inpatient costs correspond to a bed-day at a haematological clinic plus one haematologist visit per day. The cost of thrombocyte transfusion is based on a regional cost-per-patient study, inflated to year 2008.

Indirect costs in terms of production loss were estimated using the human capital approach, with an average monthly salary for individuals aged 45–64 years including payroll taxes of 41%. The workforce participation was assumed to be 85% among patients with CML patients who were under 65 years, recommended by clinical experts, as not all patients are in the labour market for reasons other than CML diagnosis. The expected increase in public consumption owing to extended survival resulting from either treatment was included in the analysis [increased survival costs equal total consumption less total production during life-years gained (LYGs), according to Swedish guidelines on economic evaluation].

Cost-effectiveness results

The results showed that patients with chronic-phase CML, who are resistant to standard-dose imatinib gain, on average, 0.67 LYs or 0.62 QALYs when treated with dasatinib compared with high-dose imatinib. The incremental societal cost amounts to €4250 during the lifetime period or €6880 per QALY gained. The indirect costs of production losses and increased public consumption almost cancel out.

In the one-way sensitivity analysis, dasatinib is a dominant treatment option in a 10-year time horizon (both cost saving and generating more benefit). Probabilistic sensitivity analysis results fall below the derived willingness to pay (WTP) for a QALY in Sweden.

Summary of key issues

• It is unclear how generalisable the model parameters and results are to the UK as the study was conducted in Sweden and takes a societal perspective. In particular, non-medical costs are included, and it is unclear what the results would be if the study were adapted for the UK.

• Adequate details are provided on the model structure and the methodology used, and results obtained seem credible.

• There are some methodological limitations and uncertainties, such as the use of a surrogate measure of effectiveness (response rate) from a single Phase II trial and probability data are taken from a range of sources.

• No base treatment has been used so the study is slightly different from the scope of this appraisal.

• Patients receive treatment until disease progression.

Characteristics of manufacturers’ models and the PenTAG assessment report model

Description of each of the modelling approaches and results

Novartis

Modelling approach

A Markov model was developed to simulate the transition of a hypothetical cohort of 1000 patients resistant to standard-dose imatinib over their lifetime (Figure 4). The cycle length was 1 month for the first six cycles and then 3 months. An equal number of male and female patients in chronic phase, aged 57 years, enter the model. At each cycle, patients have the probability of remaining in chronic phase or progressing to accelerated-phase CML. Patients failing on second-line treatment may remain in chronic phase and receive further treatment (stem cell transplantation if eligible or hydroxycarbamide) before progressing to accelerated phase/blast-crisis phase. Patients can then progress from accelerated phase to blast-crisis phase and, finally, from blast-crisis phase to death. Patients are able to remain in chronic phase, accelerated phase or blast-crisis phase for more than one cycle and they may die from other causes in chronic phase and accelerated phase. Patients may die from CML only in blast-crisis phase. On progression to accelerated phase, patients receive hydroxycarbamide, based on clinical opinion. TTD and overall survival were used to predict lifetime costs, LYG and QALYs.

FIGURE 4.

Markov model health states used in the Novartis model. 4

Assumptions used in the Novartis model4

• All patients in either the nilotinib arm or high-dose imatinib arm are assumed to receive treatment until treatment failure, when it is assumed they receive allo-stem cell transplantation as a third-line option, if eligible, otherwise hydroxycarbamide; this is assumed to occur before progression to accelerated phase. Patients in both arms who progress to accelerated phase or blast crisis receive hydroxycarbamide.

• All patients are assumed to have died of CML or other causes by the age of 100 years.

• Patients may stop taking nilotinib prior to progression to the next phase of treatment, so TTD of treatment is used in the model, rather than progression-free survival, to provide an estimate of time on nilotinib.

• It was assumed that 10% of patients who discontinued treatment owing to AEs would progress from chronic phase to accelerated phase.

• Utilities were assumed to be independent of drug therapy and time; also utility values for accelerated phase and blast-crisis phase were assumed to be the same.

Estimation of effectiveness

Overall survival for nilotinib was estimated as 86% from the clinical study (CAMN107A2101) at 24 months’ follow-up. TTD is a Kaplan–Meier estimate of duration of exposure defined as the time difference (days) between the first dose and last dose. Long-term survival was extrapolated from the study data using an exponential curve as this provided a good fit to current data. Overall survival and TTD data were taken from a study by Kantarjian and colleagues17 for high-dose imatinib.

To model stem cell transplantation after failure of nilotinib or high-dose imatinib the outcome of stem cell transplantation was based on a risk score in order to take account of patient characteristics. Thus, the model uses outcomes based on patients with a risk factor score of 4 with a 5-year survival of 34%. 18 Time on hydroxycarbamide following nilotinib failure was estimated based on the time in chronic phase following discontinuation of nilotinib or imatinib, by considering the difference between progression-free survival and TTD curves. An exponential curve was fitted to this difference and provided an estimate of 5-year survival of 16%. The effectiveness of hydroxycarbamide is assumed to be the same regardless of the positioning of hydroxycarbamide in the treatment pathway and the same data are used for the exploratory analysis.

Estimation of quality-adjusted life-years

Utility values were assigned to the different health states derived from a study by Reed and colleagues19,20 [the International randomized study of interferon versus ST1571 (IRIS)] using the EQ-5D. A utility decrement was used for patients experiencing grade 3 or 4 AEs during the first 18 months of treatment in chronic phase, taken from evidence in the nilotinib clinical trial (CAMN107A2101). A decrement was applied to the long-term utility for 52% patients following stem cell transplantation.

The utility values for chronic phase, accelerated phase and blast-crisis phase were 0.854, 0.595 and 0.595, respectively. Disutilities for AEs, taken from various sources, were 0.049 for nilotinib, 0.027 for high-dose imatinib, 0.00 for hydroxycarbamide and 0.079 for stem cell transplantation. The authors assumed the same utilities for both the accelerated-phase/blast-crisis-phase health states.

Estimation of costs

Costs, adjusted to 2009–10 prices, include routine appointments (taken from NHS reference costs 2006/7),21 end-of-life care and treatment for managing AEs. The costs of the drugs were taken from the British National Formulary (BNF) 2010, except for high-dose imatinib, which used a cost that reflected a future cost increase (commercial in confidence at time of review). Where published data were not available, advice was sought from clinical experts. Most of the uncertainty was around the costs for stem cell transplantation; however, this was stated not to affect the overall results given that upon failure of nilotinib or high-dose imatinib patients follow a similar treatment pathway.

The quarterly cost of nilotinib is £7928 and of high-dose imatinib is £10,490.

Cost-effectiveness results

The results for patients with chronic-phase CML who are resistant to imatinib indicate that nilotinib dominates high-dose imatinib, i.e. nilotinib is less costly and more effective than high-dose imatinib. The incremental cost per QALY gained is –£30,513 (Table 30). In the exploratory analysis compared with stem cell transplantation/hydroxycarbamide, the incremental cost per QALY gained for nilotinib is £44,028.

TABLE 30 - Incremental cost-effectiveness ratio (ICER) for imatinib-resistant patients aged 57 years

HDI, high-dose imatinib; HU, hydroxycarbamide; SCT, stem cell transplantation.

Exploratory analysis is shown in italic text.

Intervention	Costs (£)	LYs (£)	QALYs (£)	ICER (LYG) (£)	ICER (QALY) (£)
HDI	146,234	5.53	4.28
Nilotinib	139,216	5.80	4.51	–26,006	–30,513
SCT/HU	80,933	4.21	3.18	36,748	44,028

A range of efficacy assumptions, health utilities, costs and other parameters were considered in sensitivity analyses. For the deterministic sensitivity analyses, most incremental cost-effectiveness ratios (ICERs) are close to the base-case result of –£30,000, except for the 5-year time horizon, which gives an ICER of –£82,000 (due to delayed treatment benefit), and for extending high-dose imatinib TTD from 14 months to 19.4 months, which gives an ICER is £201,871 (higher costs of high-dose imatinib treatment with marginal QALY gain for high-dose imatinib vs nilotinib).

Probabilistic sensitivity analysis was undertaken to explore the impact of joint uncertainty in all model parameters on the cost-effectiveness results. The uncertainty around costs was represented based on the interquartile ranges presented within the NHS reference costs where available. Costs were assumed to have a gamma distribution. Quality of life (QoL) of the health states was varied using the uncertainty reported by Reed and colleagues. 19,20 The uncertainty within the TTD curves was based on beta distributions, with alpha and beta parameters representing the number of people who have discontinued and not discontinued, respectively. The uncertainty relating to nilotinib and high-dose imatinib is assumed to be correlated. Results give an ICER of –£86,413 per QALY gained. From cost-effective acceptability curves nilotinib is predicted to be cost-effective at a threshold of over £10,000 per QALY.

Summary of key issues

• The evaluation compared nilotinib, high-dose imatinib and stem cell transplantation/hydroxycarbamide, but did not include dasatinib.

• It is not clear if stem cell transplantation/hydroxycarbamide is an appropriate comparator.

• The derivation of parameters, such as progression to accelerated phase and blast-crisis phase, is poorly explained.

• Overall survival estimates appear low, but the reasons for this are unclear.

Bristol-Myers Squibb

Modelling approach

A Markov model was developed to predict the health changes and resulting costs for patients starting dasatinib treatment in each of the three phases of CML: chronic phase, accelerated phase and blast-crisis phase (Figure 5). It is not stated whether this is a newly developed model or has been adapted from a previously reported model. The model uses monthly cycles.

FIGURE 5.

The BMS3 model. SAE, serious adverse event.

The authors state that the Markov process was considered appropriate, as it allows the incorporation of the three disease phases, different response categories and the different rates of disease progression characteristic of CML, and the approach used has been used in previous economic analyses in CML.

For chronic phase and accelerated phase, the model consists of three health states: ‘stable disease’, ‘progressed disease’ and ‘death’. For blast-crisis phase, the model consists of two states: ‘stable disease’ and ‘death’. In each health state, five types of response to treatment are used in the model: no response to treatment (NR); achieve CHR; achieve partial cytogenetic and CHR (partial cytogenetic response); achieve complete cytogenetic and CHR (complete cytogenetic response); and achieve a molecular response (MR).

The modelling incorporates two stages: (1) initial assessment of the patient’s initial best response to treatment and (2) determination of the prognosis of the patient, based on his or her initial best response. Initial best response rates were based on clinical trial data and, in some cases, clinical opinion. As such, once the patients’ initial best response has been determined, they enter a specific ‘submodel’ that links response with long-term prognosis rates (e.g. progression-free survival, overall survival). The prognosis rates for different response groups were based on evidence from the BMS 034 trial. 22

Assumptions used in the Bristol-Myers Squibb3 model

• Response to treatment is assessed in the initial period; after that, it is assumed to remain at the same level until disease progression.

• The efficacy of 800 mg imatinib is equivalent to 600 mg in accelerated phase and blast-crisis phases.

• The efficacy of standard-dose imatinib and interferon alfa is zero.

• Patients cannot return to the chronic phase from advanced phases of CML.

• The probability of progressing to the next CML phase and death was estimated from the progression-free survival and overall survival data for patients in a dasatinib trial (i.e. BMS trial 034). 22 The probability of progression or death was (other than by response) independent of treatment.

• Beyond the trial period, progression rates were assumed to remain constant, at a rate equal to that during the final year of follow-up.

• After failing imatinib, dasatinib or nilotinib, patients receive post-failure treatment (PFT).

• Progression rates and other input parameters for patients receiving PFT are assumed equal to those used for non-responders.

• Patients receiving PFT incur the cost, but not the utility benefits.

• Utility values do not change over time, as long as the patient remains in the same health state.

• Where utility estimates for serious adverse events (SAEs) were not available from the non-CML literature a 5% (–0.05) decrement was assumed.

• Where resource use associated with an AE was not known, a cost of £100 was assumed.

• Monthly cost of bone marrow stem cell transplantation is based on an aggregate figure to reflect the average costs for different prognoses post stem cell transplantation.

• Different utility values were used for response and no response groups.

Effectiveness of the interventions

Effectiveness data used in the model are the patient’s initial best response to treatment, which was taken from various studies for the different interventions and is shown in Table 31.

TABLE 31 - Initial best response rate (chronic phase)

CCyR, complete cytogenetic response; CP, chronic phase; IFN, interferon; NA, not applicable; PCyR, partial cytogenetic response; SCT, stem cell transplantation.

Assumption.

Kantarjian et al. J Clin Oncol 2009b;27:abstract 7029.

PCyR, typically defined as between 1% and 35% of Philadelphia-positive (Ph+) chromosomes in metaphase in bone marrow (study definitions may vary). CCyR, typically defined as no Ph+ chromosomes in metaphase in bone marrow.

CP	No response (%)	CHR (%)	PCyR (%)	CCyR (%)	Survive SCT (%)	Mortality (%)	Follow-up (months)
Dasatinib22	8.1	33.1	15.3	43.5	0.0	0.0	24
Imatinib 400 mga	100.0	0.0	0.0	0.0	0.0	0.0	NA
Imatinib 600 mg17	56.4	15.4	28.2	0.0	0.0	0.0	12
Imatinib 800 mg23	32.1	13.3	14.1	40.5	0.0	0.0	61
Nilotinibb	6.0	35.0	18.0	41.0	0.0	0.0	19
IFNa	100.0	0.0	0.0	0.0	0.0	0.0	NA
Bone marrow SCTb	0.0	0.0	0.0	0.0	65.0	35.0	25

The progression-free survival rates associated with each level of response (as observed in the BMS 034 trial)22 are shown in Table 32.

TABLE 32 - Progression-free survival rates associated with level of response (from BMS study 034)22

CCyR, complete cytogenetic response; MCyR, major cytogenetic response; PCyR, partial cytogenetic response.

CCyR typically defined as no Philadelphia-positive (Ph+) chromosomes in metaphase in bone marrow. PCyR typically defined as between 1% and 35% of Ph+ chromosomes in metaphase in bone marrow. MCyR typically defined as ≤ 35% Ph+ chromosomes in metaphase in bone marrow (study definitions may vary).

Month	No response (%)	CHR (%)	PCyR (%)	CCyR (%)	MCyR (%)
0	100.0	100.0	100.0	100.0	100.0
6	30.0	94.9.	100.0	100.0	99.7
12	30.0	84.1	94.4	98.2	98.2
18	30.0	77.7	83.3	98.2	97.5
24	30.0	63.6	83.3	94.2	94.2
30	30.0	55.9	83.3	94.2	94.2
36	30.0	38.7	77.8	94.2	94.2
42	25.8	25.8	71.3	94.2	94.2
48	24.1	25.8	59.4	94.2	93.9

Estimation of quality-adjusted life-years

A cross-sectional study was commissioned to calculate utility values for the purpose of the BMS analysis3 (Szabo and colleagues 201024). Ratings for health states and response were elicited from a representative sample of 100 unaffected individuals in the UK using the TTO method and the EQ-5D instrument. The impact of the SAEs on health utility was captured in the model using utility decrements identified in the non-CML literature. Utility values were 0.85 for chronic phase with response, 0.68 for chronic phase no response, 0.79 for accelerated-phase response, 0.50 for accelerated phase no response, 0.50 for blast-crisis-phase response and 0.31 for blast-crisis phase no response. However, there was a mistake in the BMS model3 so that for the accelerated-phase/blast-crisis-phase health states only the value for blast-crisis phase no response was used.

Estimation of costs

Drug costs were estimated based on the recommended doses from their Summary of Product Characteristics and prices were from the BNF (2009). Monthly costs for the interventions were £2504, £3208 and £2613 for dasatinib, high-dose imatinib and nilotinib, respectively (see Table 28), and £863 for interferon alfa and £2400 for stem cell transplantation.

Direct costs were also included for outpatient visits, tests, hospitalisation and other interventions (such as blood transfusion). AE costs were included for treatment-related grades 3–4 SAEs, the most common being neutropenia, thrombocytopenia and leucopenia.

Cost-effectiveness results

The results of the base-case analyses for chronic-phase CML are shown in Table 33. Dasatinib dominates high-dose imatinib, nilotinib and stem cell transplantation. The total cost for dasatinib is £314,413 and the total number of QALYs is 6.425.

TABLE 33 - Base-case ICERs for dasatinib vs other treatments (chronic-phase CML)

IFN-α, interferon alfa; SCT, stem cell transplantation.

Dasatinib vs:	Incremental cost (£)	Incremental QALYs	ICER (£)
Imatinib 400 mg	179,087	4.940	36,251
Imatinib 600 mg	140,707	4.031	34,907
Imatinib 800 mg	–35,952	0.515	Dasatinib dominant
Nilotinib	–4565	0.190	Dasatinib dominant
IFN-α	185,121	4.762	38,877
SCT	–9821	1.687	Dasatinib dominant

Parameters used in the model, which were varied in the deterministic sensitivity analysis, were costs, utilities, starting age, time horizon and discounting. The key impact factors were the utility of responders, starting age and the time horizon of the model. However, the sensitivity analyses were not presented in the normal way and are difficult to interpret.

For the probabilistic sensitivity analysis, 1000 iterations were performed, with beta distributions used for probabilities and gamma distributions for costs. For initial response to treatment, the beta distribution was parameterised from the evidence in the selected clinical trials; for survival rates, one single random ‘seed’ was generated and subsequently used to generate the values for the model. The results of the probabilistic sensitivity analysis showed that the probability of dasatinib being cost-effective compared with stem cell transplantation was 81% for a WTP of £30,000. Cost-effectiveness acceptability curves were not presented for all the drugs together and results were not shown for the probability that dasatinib was cost-effective compared with all of its alternatives.

Summary of key issues

• There is uncertainty around the approach taken with respect to initial best response and extrapolation from this to progression-free survival, with values based on a single trial for surrogate outcomes (using a submodel).

• A number of assumptions have been made, such as best response rate for imatinib, because of lack of data.

• Patients are treated until progression, which does not appear to be correct and affects treatment duration and costs.

PenTAG assessment report evaluation

Modelling approach

The PenTAG AR2 cost-effectiveness analysis used a survival model developed in Microsoft Excel (Microsoft Corporation, Redmond, WA, USA). The model closely resembles a Markov state-transition approach, using an ‘area under the curve’ method. The number of patients in each state of the model over time is determined by using survival curve data to apportion the overall cohort population between the states at each successive cycle of the model. The structure of the model is shown in Figure 6. The model comprises the following health states: chronic phase (on treatment), chronic phase (following discontinuation of treatment), accelerated phase, blast-crisis phase and death. Patients in chronic-phase CML have either major cytogenetic response or no major cytogenetic response. Patients enter the model in chronic-phase CML and then subsequently progress to accelerated phase, then to blast-crisis phase and then, finally, to death from CML causes. Patients may also die in the chronic phase and accelerated phase states from non-CML causes.

FIGURE 6.

Structure for the PenTAG AR2 CML cost-effectiveness model. Major cytogenetic response (MCyR) typically defined as ≥ 35% Ph+ chromosomes in metaphase in bone marrow (study definitions may vary).

Overall survival is estimated for each of the treatments on the basis of the proportion of major cytogenetic response (responders), with survival duration estimated for major cytogenetic response responders and major cytogenetic response non-responders. A hazard ratio for overall survival of non-responders versus responders was pooled from several studies (hazard ratio = 0.37).

Patients move between the states chronic phase (on treatment) and chronic phase (no treatment) according to estimates for the progression-free survival and the number of individuals who prematurely discontinue treatment. Patients were assumed to have no drug costs in the chronic phase (no treatment) state. Patients spend a predefined time for each of the accelerated-phase and blast-crisis-phase states. The time spent in the chronic phase (no treatment) was adjusted by subtracting the durations for the chronic phase (treatment), accelerated-phase and blast-crisis-phase states from the overall survival duration. Utility values were applied to each of the treatment states to estimate the total QALYs for each of the treatments.

Assumptions in the PenTAG assessment report

• Overall survival is predicted on the basis of major cytogenetic response and the relationship between major cytogenetic response and overall survival is the same for all treatments, and not affected by the timing, duration and depth of CyR. The hazard ratio for the overall survival for the major cytogenetic response versus non-major cytogenetic response groups is based upon first-line therapy and is still valid for second-line treatments, and is constant over time.

• Duration of treatment is estimated on the basis of progression-free survival with a deduction to account for premature discontinuations.

• Times spent in accelerated phase and blast-crisis phases is independent of chronic-phase treatment, i.e. is identical across comparators.

• Treatment-related AEs incur no utility decrement and no additional costs.

• Duration of chronic phase (no treatment) is estimated by deducting time spent in chronic-phase (treatment), accelerated-phase and blast-crisis-phase states from overall survival.

Effectiveness of the interventions

A summary of the clinical effectiveness parameter estimates used in the PenTAG AR2 model is shown in Table 34. Parameter estimates were derived through a clinical effectiveness review. Overall survival was estimated for each of the treatments according to the proportion with major cytogenetic response. The data for major cytogenetic response used in the PenTAG AR2 model differed slightly from those reported in the clinical effectiveness sources (which are shown in parentheses). Overall survival rates for responders and non-responders were calibrated using a retrospective study by Jabbour and colleagues23 of the long-term efficacy of high-dose imatinib in a population that had failed standard-dose imatinib, as this source provided the longest available estimates of overall survival for responders and non-responders. For interferon alfa, overall survival was derived using an estimate for the major cytogenetic response rate from the IRIS trial. 19,20

TABLE 34 - Summary of the clinical effectiveness parameter estimates used in the PenTAG AR2 model

AP, accelerated phase; BC, blast crisis; HDI, high-dose imatinib; IFN-α, interferon alfa; MCyR, major cytogenetic response; OS, overall survival; PFS, progression-free survival.

Shah et al. J Clin Oncol 2008;26:320–12.

O’Brien et al. N Engl J Med 2003;348:994–1004.

Cervantes et al. Eur J Haematol 1996;57:286–91.

Kantarjian et al. Cancer 2001;92:250–7.

MCyR typically defined as ≤ 35% Philadelphia-positive chromosomes in metaphase in bone marrow (study definitions may vary).

Parameter	Estimate, PenTAG AR2 (SHTAC)	Sourcea	Justification
MCyR rate: dasatinib	58.1% (59%)	Shah et al. (2008)a	Only available estimate at currently recommended dose of dasatinib
MCyR rate: nilotinib	52.4% (56%)	Kantarjian et al. (2007)6	Only available estimate
MCyR rate: HDI	44.0% (54%)	Jabbour et al. (2009)23	Estimates of OS, PFS and MCyR rate all originate from the same study
MCyR rate: IFN-α	22%	O’Brien et al. (2003)b	Most recent large trial of IFN-α
Proportion of patients who discontinue dasatinib prematurely	10.2%	Shah et al. (2008)a	Treatment discontinuations are included in PFS
Proportion of patients who discontinue nilotinib prematurely	23.2%	Kantarjian et al. (2007)6	Only available estimated
Proportion of patients who discontinue HDI prematurely	14.8%	Pooled estimate	Pooled from all available data
Overall survival hazard ratio between people who achieve a MCyR to imatinib and those who do not	0.370	Pooled estimate	Pooling all data sources reduces the uncertainty in the data
Mean time spent in AP	0.80 years	Reed et al. (2004)19 and Cervantes et al. (1996)c	Large, relevant data source
Mean time spent in BC	1.09 years	Reed et al. (2004)19 and Kantarjian et al. (2001)d	Large, relevant data source

Progression-free survival was derived through fitting survival curves to available progression-free survival data. progression-free survival for dasatinib was estimated from a single data point of 0.77 corresponding to 24-month progression-free survival provided in the submission BMS3 made to NICE at that time. progression-free survival for nilotinib was estimated from the submission Novartis4 made to NICE at that time: 0.864, 0.769 and 0.632 for progression-free survival probability at 6, 12 and 18 months, respectively. progression-free survival for high-dose imatinib was derived from Jabbour and colleagues’23 single-centre retrospective study: 0.865, 0.81, 0.71, and 0.57 for progression-free survival probability at 6, 12, 18 and 24 months, respectively.

Estimation of quality-adjusted life-years

The PenTAG AR2 reviewed QoL studies to use in the economic model. From these data, the authors chose the IRIS study by Reed and colleagues,20 which was also used by Dalziel and colleagues25 in a previous assessment of imatinib for CML, as the most appropriate. These data were drawn from a large sample of patients and reported EQ-5D values. The utility values used for the health states are chronic phase (treatment or no treatment) 0.85; accelerated phase 0.73; and blast crisis 0.52.

Estimation of costs

Cost estimates in the economic evaluation include drug costs, administration costs of interferon alfa and cytarabine, outpatient visits, bone marrow tests, radiography, computerised tomography (CT) scans, blood transfusion and inpatient terminal care. All costs were inflated to 2009–10 values where necessary. The cost of treating patients with AEs has not been modelled. The drug costs are taken from the BNF and are shown above in Table 28. The dosage use differed from that suggested in the BNF, and the usage was taken from the relevant clinical studies. Although the dosage for dasatinib and nilotinib was not affected, dosage for high-dose imatinib (738 mg/day) and interferon alfa (4.8 million units per day) was lower.

Cost-effectiveness results

The results presented in the main PenTAG AR2 are supplemented with additional analyses in appendices, which include the comparator interferon alfa. Table 35 presents the aggregated totals for the base-case model results for the four treatments. Outputs are shown for total LYs (undiscounted), and total discounted QALYs and costs for each treatment over the time horizon of the model.

TABLE 35 - The PenTAG AR2 aggregated base-case results

AP, accelerated phase; BC, blast crisis; CP, chronic phase; HDI, high-dose imatinib; IFN, interferon.

	Dasatinib	Nilotinib	HDI	IFN
LYs: mean, undiscounted
CP treated	6.50	2.44	2.68	2.04
CP not treated	5.00	8.65	7.79	6.82
AP	0.80	0.80	0.80	0.80
BC	1.09	1.09	1.09	1.09
Total (mean)	13.40	12.98	12.37	10.75
Total (median)	10.76	10.21	9.45	7.75
QALYs: mean, discounted
CP treated	4.50	1.89	2.10	1.27
CP not treated	2.62	5.00	4.46	4.16
AP	0.37	0.38	0.38	0.41
BC	0.36	0.36	0.37	0.39
Total	7.846	7.630	7.311	6.229
Costs (£): mean, discounted
Drug costs	161,432	70,143	88,883	15,936
Drug administration	0	0	0	4390
Monitoring outpatient appointment	6818	6728	6597	6259
Bone marrow tests	6518	2732	3038	2199
Radiography	726	736	752	795
CT scans	428	434	444	469
Blood transfusions	4058	4117	4205	4445
Inpatient palliative care	41,346	76,439	68,496	64,325
Total	221,325	161,330	172,415	98,818

The incremental cost–utility of dasatinib, nilotinib, high-dose imatinib and interferon alfa, as estimated in the PenTAG AR2 model, is shown in Table 36. The PenTAG AR2 appendix results use interferon alfa as the base case, as it is the least costly of the alternatives. The ICER for nilotinib compared with interferon alfa is more than £30,000 per QALY gained. high-dose imatinib is dominated by nilotinib, i.e. nilotinib is predicted to be both cheaper and more effective, so the PenTAG AR2 model suggests that high-dose imatinib would not be considered a viable option. Compared with nilotinib, dasatinib provides only a small additional QALY at substantial extra cost and thus the ICER is more than £250,000 per QALY gained.

TABLE 36 - Deterministic base-case results for the PenTAG AR2 model (discounted)

HDI, high-dose imatinib; IFN-α, interferon alfa.

Dasatinib vs nilotinib.

Treatment	Cost (£)	Utility (QALY)	Incremental cost (£)	Incremental utility (QALY)	Incremental £/QALY (ICER)
IFN-α	98,800	6.229
Nilotinib	161,300	7.630	62,500	1.401	44,600
HDI	172,400	7.311	11,100	–0.318	Dominated
Dasatinib	221,300	7.846	60,000	0.216	277,700a

One-way deterministic sensitivity analyses for nilotinib and dasatinib versus high-dose imatinib were performed by varying single parameters. Deterministic sensitivity analyses were not performed that included interferon alfa. In the majority of sensitivity analyses, nilotinib dominated high-dose imatinib. In only one case was nilotinib not cost-effective compared with high-dose imatinib, where progression-free survival for nilotinib was assumed to be identical to that with high-dose imatinib, i.e. treatment duration was approximately equal. The base-case conclusion for dasatinib is not affected by changes to the parameter values except for changing the treatment duration to be the same as for either high-dose imatinib or nilotinib. In these cases, dasatinib becomes cost-effective compared with these treatments.

Probabilistic sensitivity analyses were run with 1000 Monte Carlo simulations. A cost-effectiveness acceptability curve was constructed to show the probability that each treatment would be considered to most cost-effective, for a range of WTPs thresholds. At a conventional WTP threshold of £30,000 per QALY, the PenTAG AR2 model estimates the probability of interferon alfa providing optimal cost–utility at 97%, with corresponding likelihoods for nilotinib, high-dose imatinib and dasatinib of 3%, 0% and 0%, respectively. At a WTP threshold of around £45,000 per QALY, nilotinib is predicted to be the optimal choice. The model predicts that it is unlikely that dasatinib would be considered the best option; even when WTP approaches £150,000 per QALY, the probability of dasatinib being most cost-effective is < 20%.

Summary of key issues

• Overall survival is based on the surrogate outcome of major cytogenetic response. However, data used for major cytogenetic response do not appear to match data in trials.

• Although overall survival is based on major cytogenetic response, progression-free survival is not and so there is no link between overall survival and progression-free survival.

• Survival for interferon is likely to be over-optimistic (according to comments from clinical advisers).

Summary

• There are some similarities and differences between the model results.

• There are concerns with the PenTAG AR2 model with respect to the overall survival for interferon alfa and the length of time on treatment with dasatinib.

• There are concerns with the BMS model3 with respect to the time spent on treatment, and the utility values used for accelerated phase and blast-crisis phase.

• There are concerns with the Novartis4 model with respect to survival estimates, which are much lower than for the other two models owing to the assumed high mortality associated with stem cell transplantation, which is used with hydroxycarbamide as a comparator.

Model structure and approach

The PenTAG AR2 model is described in detail above (see Description and results of the published economic evaluation). No structural changes have been made for the SHTAC analyses. It is a lifetime survival model with five health states: chronic phase (on treatment), chronic phase (following discontinuation of treatment), accelerated phase, blast-crisis phase and death. Patients enter the model in chronic-phase CML and progress to accelerated phase, then to blast-crisis phase and then, finally, death from CML causes. The model uses best initial response to treatment to predict overall survival, with survival duration estimated for major cytogenetic response responders and major cytogenetic response non-responders. The relationship between major cytogenetic response and overall survival is assumed to be the same for all treatments. Patients move between the states chronic phase (on treatment) and chronic phase (no treatment) according to estimates for progression-free survival and the number of patients who prematurely discontinue treatment. Trial data are extrapolated for treatment duration and progression-free survival.

Data inputs for the interventions

In our update systematic review we have not identified any new clinical effectiveness data that provide better estimates than the data used in the PenTAG AR2 [described above (see Description and results of the published economic evaluation) and shown in Tables 28, 34, 40 and 41]. The parameters therefore used for the interventions are mostly as for the PenTAG AR,2 with the exception of parameter values for progression-free survival and treatment duration for dasatinib. The costs of the drugs were taken from the BNF (2010). The cost of high-dose imatinib is due to increase in a subsequent edition of the BNF (unpublished at the time of the review). The effect of this cost increase for high-dose imatinib is presented in Appendix 6.

There were concerns over the parameter values chosen in the PenTAG AR,2 which gave treatment duration of dasatinib as double that for nilotinib (data taken from poor-quality studies). One of our clinical experts advised us that no difference in efficacy has been shown between nilotinib and dasatinib and, hence, no difference would be expected between these drugs in progression-free survival and treatment duration. During the consultation process for the PenTAG AR,2 BMS asserted that: ‘patients treated with dasatinib would not be expected to remain on treatment for (on average) 4 years longer than those receiving nilotinib, instead the duration of treatment for the two drugs should be considerably more similar.’ PenTAG defended their choice of treatment duration as evidence based. They pointed out that the treatment duration was based upon progression-free survival, which was clearly shorter with nilotinib than with dasatinib (0.63 at 18 months compared with 0.77 at 24 months, respectively). However, PenTAG conceded that differences between populations and outcome definitions make comparisons between these data precarious.

In our base-case analyses we assume that progression-free survival for dasatinib is the same as that for nilotinib (Table 40), based on the view of our clinical expert. An additional scenario analysis is presented subsequently, which uses the progression-free survival parameter values for dasatinib from the PenTAG AR2 model.

Although our review of clinical effectiveness identified a range of new data for high-dose imatinib (reported in Chapter 3), we have used the PenTAG AR2 as the data source for high-dose imatinib because the values chosen by PenTAG are mid-way in this range.

Data inputs for the comparators

Interferon alfa, stem cell transplantation, hydroxycarbamide and standard-dose imatinib are examined through exploratory analyses using data available from the PenTAG AR2 or the manufacturers’ models where no suitable data are available in the PenTAG AR. 2 Therefore, data for hydroxycarbamide are taken from the Novartis4 model and data for standard-dose imatinib and stem cell transplantation from the BMS model. 3 For these analyses we have derived our best estimates of the following parameters: monthly treatment cost, treatment duration, overall survival and health-state utility for the chronic-phase treatment period (see Table 41). The PenTAG AR2 estimated overall survival by extrapolating from the surrogate outcome major cytogenetic response. However, owing to the lack of data for overall survival and major cytogenetic response for these comparators, we were unable to derive survival curves in this way and so have instead taken a simple pragmatic approach and selected an estimate for overall survival (see below). In the model, an overall survival curve is derived from this overall survival estimate, by assuming a negative exponential distribution for mortality. The PenTAG AR2 estimated progression-free survival and treatment duration by fitting distributions to the clinical trial data. In the absence of any reliable data for the comparators, we were unable to provide estimates of the clinical data and so instead use plausible estimates for treatment duration for each of the parameters.

Results of SHTAC analyses

The base-case deterministic results are shown in Table 42. As the evidence for each of the comparators and interventions is poor, and many assumptions have had to be made to model long-term outcomes, the results should be treated with caution. The results are shown in Table 42 for nilotinib, dasatinib, high-dose imatinib, and each comparator against hydroxycarbamide, as hydroxycarbamide is the cheapest comparator, with the treatments ordered by increasing effectiveness. The comparators are also compared against the previous best option, i.e. a treatment that is more clinically effective and cost-effective compared with a preceding one (i.e. not dominated or extendedly dominated). At a cost-effectiveness threshold of £30,000, interferon alfa, standard-dose imatinib (400 mg) and stem cell transplantation are not cost-effective compared with hydroxycarbamide. This can be seen in Figure 15 by considering the cost-effectiveness frontier. The cost-effectiveness frontier is a line connecting the most cost-effective treatments. Treatments above this line are not cost-effective, as they are either dominated or extendedly dominated by the treatments on the frontier.

FIGURE 15.

Cost-effectiveness plane for treatments for CML. HDI, high-dose imatinib; HU, hydroxycarbamide.

These results show that each intervention (dasatinib, nilotinib and high-dose imatinib) has a similar total cost and QALY and each intervention has an ICER around £30,000 per QALY compared with hydroxycarbamide. In this analysis nilotinib and dasatinib were marginally more cost-effective compared with hydroxycarbamide than high-dose imatinib owing to lower costs and better efficacy (major cytogenetic response) (see Table 34). Owing to the uncertainty around the clinical data, it is unclear which of the interventions – dasatinib, high-dose imatinib or nilotinib – would be the most cost-effective. This is shown in Figure 15, in which high-dose imatinib is dominated by nilotinib. The cost-effectiveness of dasatinib versus nilotinib is £50,000 per QALY gained.

Sensitivity analyses

The uncertainty around the model results was explored using deterministic sensitivity analyses, threshold analyses and probabilistic sensitivity analyses. The deterministic sensitivity analyses were performed around some of the model input parameters that had been shown to have the greatest effect on the model results in the PenTAG AR. 2 The interventions dasatinib, nilotinib and high-dose imatinib are compared with hydroxycarbamide. The other comparator treatments were not included in the sensitivity analyses as they were dominated by hydroxycarbamide in base-case analyses; however, we have examined the other comparator treatments in threshold analyses.

A deterministic sensitivity analysis was performed for changes to the overall survival for hydroxycarbamide and treatment efficacy for the interventions dasatinib, nilotinib and high-dose imatinib. There are no data for the likely range in overall survival for hydroxycarbamide so a plausible range was chosen (2–6.5 years). Treatment efficacy (major cytogenetic response) has been varied between 30% and 70% for the interventions. This reflects the range of efficacy found in the clinical review (see Chapter 3, Effectiveness of high-dose imatinib) for high-dose imatinib.

We have compared the results for each intervention versus hydroxycarbamide (Table 43). Changes in the length of overall survival for hydroxycarbamide have only a small impact on the cost-effectiveness of any of the interventions. Similarly, changes to major cytogenetic response also have little impact on the results.

A two-way sensitivity analysis was performed for the effect of different overall survival times and treatment durations for nilotinib versus hydroxycarbamide (Table 44). These results show that most of the cost-effectiveness estimates are below £40,000 per QALY gained, except if the overall survival for nilotinib was much lower or the treatment duration was much longer than chosen for the base case. Similar results are obtained for dasatinib versus hydroxycarbamide and for high-dose imatinib versus hydroxycarbamide (not shown here).

In our critique of the PenTAG AR,2 we raise concerns for the values chosen for major cytogenetic response for each of the interventions, as these do not match the data cited. We ran analyses using the data values found in the studies cited, i.e. major cytogenetic response – dasatinib 59%, nilotinib 56% – and high-dose imatinib 54%: however, these changes did not impact.

Threshold analysis for interventions

Threshold sensitivity analysis was performed for high-dose imatinib versus nilotinib and for dasatinib versus nilotinib (Table 45), as nilotinib was the most cost-effective treatment in the base results. The analysis varied the most influential parameters (i.e. overall survival and treatment duration). These parameters were varied for nilotinib, whereas the parameters for dasatinib and high-dose imatinib were unchanged.

The results are fairly robust for changes in overall survival and treatment duration for nilotinib versus high-dose imatinib and dasatinib versus nilotinib. However, as there was a lack of reliable data for the appropriate value for treatment duration for dasatinib, there is large uncertainty around the results for dasatinib.

Probabilistic sensitivity analysis

The uncertainty around the model results was explored using probabilistic sensitivity analysis. This was run comparing the interventions dasatinib, nilotinib, high-dose imatinib and hydroxycarbamide using the PenTAG AR2 model. We were unable to run a probabilistic sensitivity analysis for all the comparator treatments, as the PenTAG AR2 model did not include all of the comparator treatments, and there is no evidence available for the distributions around the parameter values for many of the treatments. The probabilistic sensitivity analysis was run with 1000 iterations. For each iteration parameter values are sampled at random from their probability distributions. The parameter values and distributions used in the probabilistic sensitivity analysis are described in the PenTAG AR. 2 However, we have used hydroxycarbamide as a comparator instead of interferon alfa for consistency with our base case.

The results for the probabilistic sensitivity analysis are shown graphically in a cost-effectiveness acceptability curve (Figure 16), which shows the probability that each treatment is the most cost-effective at different WTP values. For a WTP threshold of £20,000 per QALY, hydroxycarbamide is the most cost-effective treatment (probability = 100%). For a WTP threshold of £30,000 per QALY, nilotinib, dasatinib, hydroxycarbamide and high-dose imatinib have a probability of being cost-effective of 60%, 28%, 12% and 0%, respectively. These results reflect those in the deterministic analyses, in which high-dose imatinib is dominated by nilotinib (see Table 42) and high-dose imatinib does not become the optimal treatment for any of the probabilistic sensitivity analysis iterations.

FIGURE 16.

Cost-effectiveness acceptability curve for nilotinib, dasatinib, high-dose imatinib and hydroxycarbamide.

A scatterplot of the probabilistic sensitivity analysis runs (Figure 17) shows that there is considerable overlap between the results for the interventions dasatinib, nilotinib and high-dose imatinib, but all three have a cost-effectiveness of around £30,000 per QALY versus hydroxycarbamide.

FIGURE 17.

Scatterplot for probabilistic sensitivity analysis for nilotinib, dasatinib, high-dose imatinib and hydroxycarbamide. HDI, high-dose imatinib; HU, hydroxycarbamide.

Scenario analyses

To further explore uncertainty in parameter values, we have conducted scenario analyses relating to length of progression-free survival and treatment duration, and overall survival for interferon alfa, and threshold analyses for the comparators.

Summary and conclusions to the cost-effectiveness section

• The systematic review identified one cost-effectiveness study15 that compared dasatinib with high-dose imatinib (800 mg/day). The results showed that chronic-phase CML patients resistant to standard-dose imatinib gain 0.62 QALYs when treated with dasatinib compared with high-dose imatinib and the incremental societal cost is €4250 during the lifetime period or €6880 per QALY gained. It is unclear how generalisable these results are to the UK NHS, as the study was conducted in Sweden and takes a societal perspective.

• The Novartis4 submission compared nilotinib with high-dose imatinib and also had an exploratory analysis versus stem cell transplantation/hydroxycarbamide. The results showed that nilotinib dominates high-dose imatinib (i.e. is more effective and less costly). Exploratory analysis gives an ICER of about £44,000 for nilotinib versus stem cell transplantation/hydroxycarbamide.

• The BMS3 submission compared dasatinib, nilotinib and high-dose imatinib with standard-dose imatinib, stem cell transplantation, hydroxycarbamide, interferon alfa, acute leukaemia-style chemotherapy and best supportive care. The results showed that dasatinib dominates high-dose imatinib, nilotinib and stem cell transplantation.

• The PenTAG AR2 economic evaluation compared dasatinib and nilotinib to high-dose imatinib. An appendix compared these treatments with interferon alfa. The results showed that nilotinib dominates high-dose imatinib and the ICER for nilotinib versus interferon alfa is about £44,600. The ICER for dasatinib versus nilotinib is > £277,000. Concerns relate to the fact that there is no link between overall survival and progression-free survival, as overall survival is based on major cytogenetic response but progression-free survival is not, and also the estimate for survival on interferon alfa appears unrealistically high.

• There are two main differences between the industry models.

  1. – In the BMS3 model, patients are treated until progression, which incurs greater costs.

  2. – In the Novartis4 model, the assumed third-line treatment is stem cell transplantation/hydroxycarbamide, which has associated high mortality and reduced overall survival.

• These key assumptions drive the differences between the models. Therefore, we felt justified in using the PenTAG AR2 model to explore further analyses using some different data inputs.

• SHTAC conducted these analyses for the interventions dasatinib, nilotinib and high-dose imatinib, and the comparators interferon alfa, standard-dose imatinib (400 mg), stem cell transplantation and hydroxycarbamide.

• The results suggest that the three interventions – dasatinib, nilotinib and high-dose imatinib – have similar costs and cost-effectiveness.

• Nilotinib, dasatinib and high-dose imatinib are all cost-effective when compared with hydroxycarbamide, for a WTP of about £30,000 per QALY.

• Nilotinib and dasatinib are slightly more cost-effective than high-dose imatinib because of slightly lower costs and better effectiveness than high-dose imatinib.

• It is not possible to derive firm conclusions about the relative cost-effectiveness of the three interventions owing to the great uncertainty around data inputs.

• The parameters that had the most impact on the model results were overall survival, treatment efficacy and treatment durations. The results were fairly robust to changes in overall survival and treatment duration for nilotinib versus high-dose imatinib and dasatinib.

• A probabilistic sensitivity analysis was run comparing the interventions dasatinib, nilotinib, high-dose imatinib and hydroxycarbamide. For a WTP threshold of £20,000 per QALY, hydroxycarbamide is the most cost-effective treatment. For a WTP threshold of £30,000 per QALY, nilotinib, dasatinib, hydroxycarbamide and high-dose imatinib have a probabilities of being cost-effective of 60%, 28%, 12% and 0%, respectively.

Clinical effectiveness

The updated searches identified one RCT of dasatinib versus high-dose imatinib and three single-arm cohort studies of high-dose imatinib in people with imatinib-resistant CML. No new studies of nilotinib were identified. Earlier publications of the RCT had been included in the PenTAG AR;2 however, owing to major limitations in the design of the RCT, only the dasatinib arm of the trial was reported and no comparative evidence was discussed. In line with this, outcomes with longer follow-up from the dasatinib and high-dose imatinib arms of the RCT were dealt with as non-comparative evidence in the present systematic review. Key limitations of the RCT include a high crossover rate of 80% from the high-dose imatinib arm to the dasatinib arm at a median of 13 weeks and a 20% crossover from dasatinib to high-dose imatinib at a median of 28 weeks. The criteria used to define imatinib failure were slightly different in each of the four included studies. All participants had chronic-phase CML, except in one of the single-arm cohort studies, which also included very small numbers with accelerated phase and blast crisis (three and four patients, respectively). The three single-arm cohort studies appeared to have a high risk of bias. The following results should be interpreted with caution owing to the methodological limitations of the included studies.

Cost-effectiveness

One cost-effectiveness study was identified for this update. This compared dasatinib with high-dose imatinib. The results showed that chronic-phase CML patients resistant to standard-dose imatinib gain 0.62 QALYs when treated with dasatinib compared with high-dose imatinib and the incremental societal cost is €4250 during the lifetime period or €6880 per QALY gained. It is unclear how generalisable these results are to the UK NHS as the study was conducted in Sweden and takes a societal perspective.

In addition, economic evaluations from two manufacturer submissions were reviewed and critiqued and the economic evaluation presented in the PenTAG AR2 is also summarised and critiqued.

The Novartis4 economic evaluation compared nilotinib with high-dose imatinib for those with chronic-phase CML who are imatinib resistant. An exploratory analysis versus stem cell transplantation/hydroxycarbamide was also undertaken. The results showed that nilotinib dominates high-dose imatinib (i.e. is more effective and less costly). Exploratory analysis gives an ICER of about £44,000 for nilotinib versus stem cell transplantation/hydroxycarbamide.

The BMS3 economic evaluation compared dasatinib, nilotinib and high-dose imatinib with standard-dose imatinib, stem cell transplantation, hydroxycarbamide, interferon alfa, acute leukaemia-style chemotherapy and best supportive care for patients with CML who are resistant to imatinib. The results showed that dasatinib dominates high-dose imatinib, nilotinib and stem cell transplantation.

The PenTAG AR2 economic evaluation compared dasatinib and nilotinib with high-dose imatinib, and also compared the three treatments with interferon alfa, for patients with CML who are resistant to imatinib. The results showed that nilotinib dominates high-dose imatinib and the ICER for nilotinib versus interferon alfa is about £44,600. The ICER for dasatinib versus nilotinib is > £277,000.

All three economic evaluations were assessed in terms of their study quality and were deemed to be appropriately conducted. There are two main differences in the assumptions of the two industry models, which drive the differences between the models:

• In the BMS3 model, patients are treated until progression, which incurs greater costs.

• In the Novartis4 model, the assumed third-line treatment is stem cell transplantation/hydroxycarbamide, which has associated high mortality and reduced overall survival.

We therefore used the PenTAG AR2 economic model to explore further analyses using some different data inputs and to include the three interventions, dasatinib, nilotinib and high-dose imatinib, and the comparators interferon alfa, standard-dose imatinib, stem cell transplantation and hydroxycarbamide.

The SHTAC analysis

An exploratory analysis, using the PenTAG AR2 economic model, suggested that dasatinib, nilotinib and high-dose imatinib have similar costs and effectiveness. Nilotinib, dasatinib and high-dose imatinib are all cost-effective when compared with hydroxycarbamide, for a WTP of about £30,000 per QALY. Nilotinib and dasatinib are slightly more cost-effective than high-dose imatinib because of slightly lower costs and better effectiveness than high-dose imatinib. However, there is great uncertainty around the data inputs and, as such, it is not possible to derive firm conclusions about the relative cost-effectiveness of the three interventions. Where possible, exploration of these uncertainties was undertaken using sensitivity analyses. Deterministic sensitivity analyses showed that changes in overall survival for hydroxycarbamide and changes in treatment efficacy of the interventions had little impact on results. A probabilistic sensitivity analysis was run comparing the interventions dasatinib, nilotinib, high-dose imatinib and hydroxycarbamide. For a WTP threshold of £20,000 per QALY, hydroxycarbamide would be the most cost-effective treatment. For a WTP threshold of £30,000 per QALY, nilotinib, dasatinib, hydroxycarbamide and high-dose imatinib have a probability of being cost-effective of 60%, 28%, 12% and 0%, respectively.

Title/abstract screening

Titles/abstracts will be screened and studies will be selected for full-paper retrieval according to the following criteria:

• patients have CML

• patients have experienced disease progression while being treated with imatinib

• dasatinib, nilotinib or high-dose imatinib are used.

Where this information is not available, and it is not fully clear that the study does not fit the inclusion criteria, full-paper copies will be retrieved for further assessment.

Study inclusion/exclusion criteria

Full papers will be assessed for inclusion with reference to the patients, interventions, comparators and outcomes described in the decision problem above. Questions for inclusion/exclusion are designed to be less specific than details stated in the decision problem so that all trials and studies that can potentially contribute relevant data will be included.

Sub-group analysis

Patients in different phases of CML will be considered separately in the main review and these are therefore not considered to be sub-groups. If evidence allows, patients with different levels of previous imatinib response will be considered in a sub-group analysis.

Cost-effectiveness review

The sources outlined above (see Search strategy) will be used to identify studies of the cost-effectiveness of dasatinib, nilotinib or high-dose imatinib in patients with imatinib-resistant CML. The inclusion and exclusion criteria for the systematic review of published cost-effectiveness studies will be identical to that applied in the review of clinical effectiveness, with the exception of study design (see Study inclusion/exclusion criteria). The quality of the included economic evaluations will be assessed using a critical appraisal checklist based upon that proposed by Drummond and colleagues13 and Philips and colleagues. 12 The data from these studies will be tabulated and discussed in a narrative review.

Economic evaluation

An assessment of dasatinib and nilotinib as the intervention treatments compared with high-dose imatinib has recently been conducted. 2 In this previous assessment, an economic model was developed in which high-dose imatinib and interferon alfa were used as comparators in the modelling approach. We will critically appraise this existing model using the same checklist as specified above (see Cost-effectiveness review). In addition we will adapt this existing model to run updated analyses for the current assessment to reflect the current scope. That is, we will compare each intervention with one another and against all comparators specified in Study inclusion/exclusion criteria, if feasible and appropriate. Additional searches will be undertaken if required to inform specific parameters; these will be determined during the review of the model.

Study details

Baseline characteristics

Results

Quality appraisal

Study details

Baseline characteristics