Currently several tyrosine kinase inhibitors (TKIs) are approved for treatment of chronic myeloid leukemia (CML). Our goal was to identify the optimal sequential treatment strategy in terms of effectiveness and cost-effectiveness for CML patients within the US health care context. We evaluated 18 treatment strategies regarding survival, quality-adjusted survival, and costs. For model parameters, the literature data, expert surveys, registry data, and economic databases were used. Evaluated strategies included imatinib, dasatinib, nilotinib, bosutinib, ponatinib, stem-cell transplantation (SCT), and chemotherapy. We developed a Markov state-transition model, which was analyzed as a cohort simulation over a lifelong time horizon with a third-party payer perspective and discount rate of 3%. Remaining life expectancies ranged from 5.4 years (3.9 quality-adjusted life years (QALYs)) for chemotherapy treatment without TKI to 14.4 years (11.1 QALYs) for nilotinib
Chronic myeloid leukemia (CML) is the third most common type of leukemia [
The price of imatinib has tripled since 2001 in the US, from $30,000 per year to $92,000 [
When faced with choosing between the number of available treatment strategies with the goal to balance benefits, harms, and costs, decision-analytic modeling can be used as a supportive instrument in the decision making process [
Hence, we developed and updated a Markov state-transition model [
To represent available treatment options within the US health care context, 18 different sequential treatment strategies (Table
Sequential treatment strategies.
First-line TKI | Second-line TKI | Chemotherapy/stem-cell transplantation |
---|---|---|
Chemotherapy |
— | — |
Imatinib | — | Chemotherapy/stem-cell transplantation |
Bosutinib | — | Chemotherapy/stem-cell transplantation |
Dasatinib | — | Chemotherapy/stem-cell transplantation |
Nilotinib | — | Chemotherapy/stem-cell transplantation |
Imatinib | Bosutinib | Chemotherapy/stem-cell transplantation |
Imatinib | Dasatinib | Chemotherapy/stem-cell transplantation |
Imatinib | Nilotinib | Chemotherapy/stem-cell transplantation |
Imatinib | Ponatinib | Chemotherapy/stem-cell transplantation |
Bosutinib | Dasatinib | Chemotherapy/stem-cell transplantation |
Bosutinib | Nilotinib | Chemotherapy/stem-cell transplantation |
Bosutinib | Ponatinib | Chemotherapy/stem-cell transplantation |
Dasatinib | Bosutinib | Chemotherapy/stem-cell transplantation |
Dasatinib | Nilotinib | Chemotherapy/stem-cell transplantation |
Dasatinib | Ponatinib | Chemotherapy/stem-cell transplantation |
Nilotinib | Bosutinib | Chemotherapy/stem-cell transplantation |
Nilotinib | Dasatinib | Chemotherapy/stem-cell transplantation |
Nilotinib | Ponatinib | Chemotherapy/stem-cell transplantation |
TKI: tyrosine kinase inhibitor.
We expanded, updated, and adapted our previously developed Markov state-transition model [
State-transition diagram of the Markov state-transition model. AP, accelerated phase; BP, blast phase; CP, chronic phase; SCT, stem-cell transplantation; and TKI, tyrosine kinase inhibitor.
The Markov model was analyzed as a cohort simulation. For model evaluation, the absolute and incremental outcome measures life years (LYs), quality-adjusted life years (QALYs), and costs, as well as incremental cost-effectiveness ratios (ICERs in $/LY gained) and incremental cost-utility ratios (ICURs in $/QALY gained) were used. A third-party payer perspective was adopted and a discount rate of 3% was applied for both costs and health outcomes [
The following important assumptions and simplifications were made: in advanced phases of the disease, chemotherapy was the only treatment option. Additionally, the sequential application of only two different TKIs was considered in the model and dose modification was not allowed within the model. After receiving SCT, patients were assumed to either survive without relapse or die.
The model was programmed and analyzed in TreeAge 2015 (TreeAge Software, Inc. Williamstown, MA, USA).
In the US, the median age at diagnosis for CML is approximately 64 years [
Data for natural history, effectiveness, adverse events, and costs are reported in Table
Data for natural history, effectiveness, adverse events, and costs.
Input parameter | Value | Source (references) |
---|---|---|
|
||
Probability of staying on 1st-line chemotherapy | Weibull, shape: 1.17, scale: 53.15 | [ |
Probability of staying on 1st-line bosutinib | Weibull, shape: 0.92, scale: 54.79 | [ |
Probability of staying on 1st-line dasatinib | Weibull, shape: 0.92, scale: 102.4 | [ |
Probability of staying on 1st-line imatinib | Weibull, shape: 0.92, scale: 79.65 | [ |
Probability of staying on 1st-line nilotinib | Weibull, shape: 0.92, scale: 106 | [ |
Probability of staying on 2nd-line bosutinib | Exponential, 0.97 | [ |
Probability of staying on 2nd-line dasatinib | Exponential, 0.98 | [ |
Probability of staying on 2nd-line nilotinib | Exponential, 0.97 | [ |
Probability of staying on 2nd-line ponatinib | Exponential, 0.97 | [ |
Probability of staying in CP on chemotherapy after TKI failure | Exponential, 0.01 | [ |
Probability of staying in AP on chemotherapy | Exponential, 0.11 | [ |
Probability of dying from CML in BP on chemotherapy | Exponential, 0.09 | [ |
|
||
Chronic phase | 0.92 × (age-dependent utility general population) | [ |
Accelerated phase | 0.79 × (age-dependent utility general population) | [ |
Blast phase | 0.57 × (age-dependent utility general population) | [ |
After SCT without GvHD | 0.98 × (age-dependent utility general population) | [ |
After SCT with GvHD | 0.9 × (age-dependent utility general population) | [ |
|
||
Imatinib 1st line 400 mg qd | $10,057.04 per month | [ |
Dasatinib 1st line 100 mg qd | $11,021.20 per month | [ |
Nilotinib 1st line 300 mg bid | $10,436.08 per month | [ |
Bosutinib 1st line 500 mg qd | $11,277.36 per month | [ |
Dasatinib 2nd line 100 mg qd | $11,021.20 per month | [ |
Nilotinib 2nd line 400 mg bid | $10,436.00 per month | [ |
Bosutinib 2nd line 500 mg qd | $11,277.36 per month | [ |
Ponatinib 2nd line 45 mg qd | $12,611.04 per month | [ |
Hydroxyurea 2000 mg qd | $655.24 per month | [ |
Tacrolimus 2 mg/day | $313.20 per month | [ |
Mycophenolate 2000 mg/day | $1,887.35 per month | [ |
|
||
Outpatient in CP | $162.52 per month | [ |
Inpatient in CP | $323.09 per month | [ |
Outpatient in AP | $261.98 per month | [ |
Inpatient in AP | $2,173.10 per month | [ |
Outpatient in BP | $261.98 per month | [ |
Inpatient in BP | $1,890.59 per month | [ |
|
||
Acute GvHD | $66,821.50 | [ |
Chronic GvHD | $10,082.11 | [ |
Follow-up care within the first year after SCT | $556.15 per month | [ |
Follow-up care beyond the first year after SCT | $485.61 per month | [ |
Transplant from live related donor (occurs just once) | $90,234.54 | [ |
Transplant from live unrelated donor (occurs just once) | $131,976.34 | [ |
|
||
Abdominal pain | $5,176.34 per inpatient stay | [ |
Anemia | $4,919.15 per inpatient stay | [ |
Diarrhea | $5,389.65 per inpatient stay | [ |
Hypertension | $6,845.76 per inpatient stay | [ |
Leukocytopenia | $6,424.44 per inpatient stay | [ |
Neutropenia | $8,400.54 per inpatient stay | [ |
Pancreatitis | $7,656.35 per inpatient stay | [ |
Rash | $3,915.32 per inpatient stay | [ |
Thrombocytopenia | $5,846.95 per inpatient stay | [ |
AP: accelerated phase; bid: twice a day; BP: blast phase; CML: chronic myeloid leukemia; CP: chronic phase; GvHD: graft-versus-host disease; qd: every day; SCT: stem cell transplantation.
Besides survival, we considered preference-based health outcomes. Survival was adjusted for health-related quality of life (QoL) using quality of life indices (i.e., utilities). Life years were multiplied by utilities to derive QALYs. Utility values can range from one (perfect health) to zero (death) [
The costing index year for the analysis was 2014. All cost data were adjusted for inflation to US dollars for the price level of June 2014 using the consumer price index (CPI) for medical services and the CPI for medical care commodities [
Drug prices were derived from the Redbook [
Comparative clinical effectiveness was estimated as survival and quality-adjusted survival for each treatment strategy. In the cost-effectiveness analysis, we compared clinical outcomes to costs and calculated incremental cost-effectiveness ratios and incremental cost-utility comparing the different strategies. To calculate incremental ratios, first, strategies were ordered according to their costs (i.e., starting with the least expensive strategy). Subsequently, strategies that were less effective and more or equally expensive as the comparator were excluded due to dominance. After removal of the dominated strategies, ICERs and ICURs were calculated by applying the following formula: [Costs(Strategy1) – Costs(Strategy2)]/[Effectiveness(Strategy1) – Effectiveness(Strategy2)]. Afterwards, the ICERs and ICURs were compared and strategies that were weakly dominated were excluded as well. Weak dominance or “extended dominance rules out any strategy with a higher incremental cost-effectiveness ratio (ICER), which is greater than that of a more effective strategy. That is, extended dominance applies to strategies that are not cost-effective because another available strategy provides more units of benefit at a lower cost per unit of benefit [
In 2015, it is expected that imatinib will lose patent protection [
In the comparative effectiveness analysis, remaining life expectancies (undiscounted) ranged from 5.4 years (3.9 QALYs) to 14.4 years (11.1 QALYs). Our analyses showed a large gain in life expectancy when using a treatment strategy that includes a TKI instead of only chemotherapy. The least effective strategy including a TKI was bosutinib
The cost-effectiveness plane in Figure
Cost-effectiveness plane base-case analysis. (a) Cost-effectiveness frontier including only nondominated strategies. (b) Cost-effectiveness frontier including all strategies except chemotherapy alone. Chemo: chemotherapy; QALYs: quality-adjusted life years; SCT: stem-cell transplantation. Letters next to the symbols in (b) indicate the second-line TKI: B: bosutinib; D: dasatinib; P: ponatinib; N: nilotinib; the shape of the symbols explained in the legend beneath the graph indicates the first-line TKI. The cost-effectiveness plane presents simultaneously costs (
The remaining nondominated strategies resulted in the following ICERs and ICURs. Imatinib without second-line TKI came to an ICUR of $171,700 per QALY gained (ICER $137,900 per LY gained) compared to the baseline strategy chemotherapy. Imatinib
Cost-effectiveness results base-case analysis.
Costs ($) | Life years | QALYs | ICERs ($/LY) | ICURs ($/QALYs) | |
---|---|---|---|---|---|
Chemo | 94,492 | 4.86 | 3.47 | — | — |
Bosutinib → chemo/SCT | 676,243 | 9.06 | 6.86 | Weakly dominated | Weakly dominated |
Imatinib → chemo/SCT | 749,272 | 9.61 | 7.29 | 137,900 | 171,700 |
Nilotinib → chemo/SCT | 884,222 | 10.08 | 7.65 | Weakly dominated | Weakly dominated |
Dasatinib → chemo/SCT | 912,367 | 10.02 | 7.61 | Dominated | Dominated |
Bosutinib → nilotinib → chemo/SCT | 913,682 | 9.96 | 7.82 | Dominated | Weakly dominated |
Bosutinib → ponatinib → chemo/SCT | 947,136 | 9.92 | 7.80 | Dominated | Dominated |
Imatinib → nilotinib → chemo/SCT | 965,597 | 10.44 | 8.14 | 260,800 | 253,500 |
Imatinib → ponatinib → chemo/SCT | 995,868 | 10.40 | 8.12 | Dominated | Dominated |
Imatinib → bosutinib → chemo/SCT | 1,020,857 | 10.57 | 8.22 | Weakly dominated | Weakly dominated |
Bosutinib → dasatinib → chemo/SCT | 1,062,220 | 10.45 | 8.14 | Dominated | Dominated |
Imatinib → dasatinib → chemo/SCT | 1,099,065 | 10.88 | 8.43 | Weakly dominated | Weakly dominated |
Nilotinib → ponatinib → chemo/SCT | 1,108,291 | 10.80 | 8.39 | Dominated | Dominated |
Dasatinib → nilotinib → chemo/SCT | 1,111,549 | 10.79 | 8.38 | Dominated | Dominated |
Nilotinib → bosutinib → chemo/SCT | 1,130,750 | 10.95 | 8.48 | Weakly dominated | Weakly dominated |
Dasatinib → ponatinib → chemo/SCT | 1,139,314 | 10.75 | 8.35 | Dominated | Dominated |
Dasatinib → bosutinib → chemo/SCT | 1,162,092 | 10.90 | 8.45 | Dominated | Dominated |
Nilotinib → dasatinib → chemo/SCT | 1,200,921 | 11.23 | 8.67 | 299,800 | 445,100 |
Chemo: chemotherapy; ICERs: incremental cost-effectiveness ratios; ICURs: incremental cost-utility ratios; QALY: quality-adjusted life years; SCT: stem-cell transplantation.
We investigated the scenario of generic drug pricing of imatinib. Table
Scenario analysis generic pricing imatinib.
Scenario | A: base-case | B: imatinib 60% of original cost | C: imatinib 40% of original cost | ||||||
---|---|---|---|---|---|---|---|---|---|
Cost (US$) | Effectiveness (QALYs) | ICUR (US$/QALY) | Cost (US$) | Effectiveness (QALYs) | ICUR (US$/QALY) | Cost (US$) | Effectiveness (QALYs) | ICUR (US$/QALY) | |
Chemo | 94,492 | 3.47 | 94,492 | 3.47 | 94,492 | 3.47 | |||
Imatinib → chemo/SCT | 749,272 | 7.29 | 171,700 | 510,214 | 7.29 | 109,000 | 390,685 | 7.29 | 77,600 |
Imatinib → nilotinib → chemo/SCT | 965,597 | 8.14 | 253,500 | 726,539 | 8.14 | 253,500 | 607,010 | 8.14 | 253,500 |
Imatinib → dasatinib → chemo/SCT | 1,099,065 | 8.43 | — | 860,007 | 8.43 | 463,800 | 740,477 | 8.43 | 463,800 |
Nilotinib → dasatinib → chemo/SCT | 1,200,921 | 8.67 | 445,100 | 1,200,921 | 8.67 | 1,415,200 | 1,200,921 | 8.67 | 1,911,400 |
Chemo: chemotherapy; QALY: quality-adjusted life years; SCT: stem-cell transplantation.
We performed best- and worst-case scenario assuming first that all second-line treatments have the same effectiveness as dasatinib (most effective second-line treatment in our analysis) and second a worst-case analysis assuming all second-line treatments have the same effectiveness as second-line ponatinib, which has demonstrated the least second-line effectiveness on average in our analysis (see Table
In another scenario analysis, we evaluated the impact of excluding SCT as an option. Excluding SCT as a treatment option does not have a huge impact on our outcome. Only one additional strategy becomes nondominated (imatinib
Several TKIs are approved and recommended in guidelines for first- and further-line treatment of CML in the US. Our study is the first one that analyzed 18 different combination strategies over a lifelong time horizon. When comparing the health outcomes, we showed that adding a TKI rather than only using chemotherapy increased the life expectancy substantially. Additionally, sequential treatment, as recommended by current treatment guidelines, brings another additional gain in life expectancy. When considering costs as well, two nondominated strategies including a second-line TKI remained on the cost-effectiveness frontier: imatinib
In the US, there is no commonly accepted willingness-to-pay threshold as, for example, in the UK (20,000–30,000 £/QALY) [
Further modeling studies resulted in similar recommendations. Our Austrian analysis [
Additionally, we compared our model results to models described in a recently published review on CML [
A significant strength of our analysis is the systematic, evidence-based, and comprehensive evaluation of 18 different treatment strategies and the inclusion of newly approved CML treatments, such as ponatinib and bosutinib. It would be hardly possible to compare 18 different treatment strategies in a randomized controlled clinical trial with sufficient sample size. Furthermore, we extrapolated short-term trial data to a lifelong time horizon and adjusted the survival for QoL to generate comprehensive patient-relevant outcomes. Evaluating and reporting the generic measure “quality-adjusted life years” help to optimize the benefit-harm tradeoff as it combines the treatment strategies’ short- and long-term effects on duration and quality of life. Furthermore, evaluation of QALYs helps to direct health care resources most efficiently as comparisons across disease are possibly opposed to reports on disease specific outcomes measures, such as incidence of complications or response rates [
Our study has several limitations. First, there were no utilities available specifically for the US setting and also no utilities specific to each treatment line. When comparing QALYs, this is a major limitation that can only be solved by conducting utility studies. However, we do also report the results for LYs without adjustment for QoL and these results support the analyses including QALYs. Second, we did not include third-line TKI treatment. This might have influenced the absolute number of LYs but should not have a big influence in comparing the different combination strategies as all of them did not include third-line TKI treatment. Another assumption was that patients cannot relapse after SCT. We considered a higher mortality compared to the general population but not a relapse after SCT. This might have led to a slightly better outcome across all strategies; however, it should not have biased the results between the different strategies. The effectiveness data were derived from pivotal clinical trials that might not represent the real world target population. However, as the effectiveness data was applied consistently across the treatment strategies it would be unlikely to influence the differences observed. Treatment patterns and access to treatment might depend on various factors, such as physician preferences, hospital policies, or insurance coverage. Therefore, we analyzed comprehensively 18 different strategies. Additionally, the model is flexible to be adapted to other settings in the future. Another limitation of our modeling approach is that the choice of the first-line treatment might also be influenced through the long-term safety results of imatinib and recent concerns that, for example, the newer TKIs might lead to severe complications, such as pleural effusion, arterial hypertension, or vascular events [
In conclusion, the model results suggest that imatinib followed by second-line nilotinib and nilotinib followed by second-line dasatinib are candidates for cost-effective sequential treatment strategies among those including a second-line TKI for chronic phase CML in the US. The decision on the cost-effectiveness has to be made in the context of individual or society’s willingness-to-pay. These results may be used to support CML treatment decision making by clinicians and patients.
Ursula Rochau reports grants from the Austrian Research Promotion Agency (FFG), during the conduct of the study, and relevant financial activities outside the submitted work with Novartis; Martina Kluibenschaedl reports grants from the Austrian Research Promotion Agency (FFG), during the conduct of the study; David Stenehjem reports grants from Novartis, during the conduct of the study; Kuo Kuan-Ling has nothing to disclose; Jerald Radich reports relevant financial activities outside the submitted work with Novartis and Ariad; Gary Oderda reports grants from Bristol Myers Squibb, outside the submitted work; Diana Brixner reports grants from Bristol Myers Squibb, outside the submitted work; Uwe Siebert reports grants from the Austrian Research Promotion Agency (FFG), during the conduct of the study, and personal fees from AGORA-Board (Novartis), outside the submitted work.
This work was supported by the COMET Center ONCOTYROL, which is funded by the Austrian Federal Ministries BMVIT/BMWFJ (via FFG) and the Tiroler Zukunftsstiftung/Standortagentur Tirol (SAT). This paper is partly based on a Master Thesis of Martina Kluibenschaedl (M.S. degree in health sciences).