We used a markov model to simulate severe thalassemia patients receiving RI-HSCT compared to BT-ICT, which is a standard practice of thalassemia treatment. The study was undertaken using a societal perspective as recommended by the Thailand’s health technology assessment guidelines 9 . We performed a cost-utility analysis with incremental cost per QALY gained.

The model simulates for the life time horizon. Clinical effectiveness and economic outcomes was directly obtained from an ongoing clinical investigation of RI-HSCT in 18 thalassemia patients in Thailand. The simulated cohort’s characteristics of the model were based on the characteristics of 18 patients. Ethical approval was granted by the Committee on Human Rights Related to Research Involving Human Subjects, Mahidol University.

A Markov model, as shown in Figure 1, was used to estimate the relevant costs and health outcomes. The health states of patients assigned to RI-HSCT consisted of eight states, which were categorized into five groups to mimic the natural history of thalassemia and clinical practice; (i) the first year of RI-HSCT consisted of four states [Q₁-Q₄], with the length of three months each. The first six months [Q₁-Q₂] is the period with the highest costs, highest graft rejections and worst quality of life (QoL). We assumed that the rejection rate in this period is consistent. The remaining six months [Q₃-Q₄] is when the patients needed close follow up in order to monitor side effects, resulting in more frequent visits. No graft rejection occured in this period; (ii) The second and third year of RI-HSCT [Iron chelation] was when patients had higher costs due to either regular phlebotomy or chelation therapy and immunosuppressive therapy; (iii) following years after successful RI-HSCT [Post RI-HSCT] (iv) RI-HSCT failure resulting in a switch to BT-ICT. After graft rejection, patients were to receive blood transfusions and subcutaneous infusion iron chealtion therapy only. Since cardiac disease is the major cause of death, we constructed the BT-ICT health state with two sub-health states: thalassemia patients without complication and thalassemia patients with cardiac complication; and (v) death Blood transfusion dependent patients had two health states (i.e. BT-ICT and death). The arrows represent the possible transitions from one state to the other. Treatment complications were included within all health states as they typically took far less than one year to resolve. The simulation estimated the costs and health outcomes over a lifetime time horizon. All transition probabilities, costs and outcomes variables are shown in Table 1.

RI-HSCT: Reduced intensity hematopoietic stem cell, BT-ICT: Blood transfusion combined with subcutaneous iron chelating therapy, Pre-BMT: Pre-bone marrow transplantation.

* ($US, year 2011 value).

^# Charges were converted to cost using a cost-to-charge ratio of 1.37 which was derived from Ramathibodi hospital for base-case analysis and 0.8-1.5 27 for sensitivity analysis.

^! Costs in Q₁-Q₄ is calculated for 3 month period, while costs beyond Q4 are on an annual basis.

The first cohort of 8 patients received conditioning regimen as described previously 8 and the remaining 10 patients received a new reduced intensity conditioning regimen consisting of intravevous busulfan 130 mg/m² for 4 days, fludarabine 35 mg/m² for 6 days and antithymocyte globulin (Thymoglobulin®) 1.5 mg/kg for 4 days and they also received fludarabine 30 mg/m² for 5 days and dexamethasone 25 mg/m² for 5 days 1 month prior to transplant for immunosuppression rationale. The reduced intensity regimen was designed for patients who were classified as having Lucarelli class 3. All these patients were given hydroxyurea 20 mg/kg/d ≥ 3 months before BMT to decrease erythroid expansion and thus prevent graft rejection 19 20 21 .

Stem cell grafts are prepared by the following procedures. T-cell nondepleted peripheral blood stem cell graft was infused on day 0. Only 1 patient received purified CD34⁺ cells that were derived from 2 antigens and HLA-mismatched maternal peripheral blood stem cells with additional CD3⁺ cells to a total of 1 × 10⁵ CD3⁺ cells/kg 21 . These purified CD34⁺ cells were obtained with Clinimacs (Miltenyi, Bergisch, Germany). All donors received subcutaneous granulocyte colony-stimulating factor 10 μg/kg/d for 4 days before leukapheresis procedures.

For graft-versus-host disease (GVHD) prophylaxis, Three days before transplantation, patients received cyclosporin at a dose that was adjusted to achieve a plasma concentration of 250–350 ng/mL or tacrolimus at a dose that was adjusted to achieve a plasma concentration of 5–15 ng/mL. All received a short course of methotrexate 22 . Engraftment Criteria; 3 consecutive days with an absolute neutrophil count >.5 × 10⁹/L or 7 consecutive days with a platelet count >20 × 10⁹/L without transfusion were taken as evidence of engraftment.

In order to prevent Pneumocystis carinii pneumonia, patients received a preventive regimen of trimethoprim-sulfamethoxazole (cotrimoxazole) at 5 mg/kg/d divided into 2 doses for 3 days a week. Cytomegalovirus (CMV) and Ebstein-Barr virus (EBV) reactivations were monitored by CMV and EBV viral loads. Patients with a CMV viral load >10³ copies/mL were treated with ganciclovir 23 and intravenous immunoglobulin 23 and patients with an EBV viral load >10³ copies/mL were treated with rituximab 24 .

As this study was undertaken using the societal perspective, each group consisted of direct medical costs, direct non medical costs, and indirect costs. All costs were converted and reported in 2011 US $ (US $1= 30.50 THB 26 ) using the consumer price index (CPI) 15 . We divided costs for RI-HSCT groups into 2 parts which were cost of pre-bone marrow transplantation (Pre-BMT) and costs of transplantation and follow up. Direct medical costs were estimated using micro-costing approach 16 . Direct medical costs for pre-bone marrow transplantation included laboratory tests and investigation from patients and donors were obtained from hospital databases. Direct medical costs for transplantation and follow up were calculated starting from the first date of patient’s admission during pre-transplantation management. Costs were divided in line with the heath states in the model, for example; costs in the first year [Q1-Q4] were estimated in 3 month-period, while costs of iron chelation and post RI-HSCT were calculated annually. Direct medical costs for RI-HSCT group were obtained from two data sources. First, costs of patients receiving RI-HSCT were retrieved from a hospital database at a teaching hospital between 2003 and 2011. Cost to charge ratio was used to convert hospital charges to cost by multiplying charges to cost to charge ratio. Conversion of hospital charge to costs is a reasonable trade-off between accuracy and the resources used for costing 16 . Charges were converted to cost using a cost-to-charge ratio of 1.37 which was derived from Ramathibodi hospital for base-case analysis and 0.8-1.5 27 for sensitivity analysis. Other direct medical costs for example: out of pocket expenses which related to medications were collected using questionnaires. Both direct non-medical and indirect costs for RI-HSCT groups were collected by interviewing severe thalassemia patients and their caregivers at Ramathibodi hospital after obtaining informed consent. Eighteen RI-HSCT patients were interviewed by questionnaire. Direct non-medical costs included; costs of transportation, meals, accommodation, facilities, other self care expenses and informal care. Indirect costs referred to productivity losses due to sick leave. Resource cost parameters are presented in Table 1.

Because of the lack of separated costs data for severe thalassemia patients with or without cardiac complications, we decided to use costs of severe thalassemia patients based on a study of Leelahavarong et al. and Riewpariboon et al. as cost inputs in our model. The costs for severe thalassemia patients were obtained from two literature studies. First, direct medical costs were derived from a cost analysis by Riewpaiboon et al. 28 which is the largest cost-of-illness of thalassemia patients study in Thailand. The study population consisted of 124 severe thalassemia patients. Mean direct medical costs per year were US $ 1,187 (SE = 137). For direct non-medical and indirect costs, we obtained data from Leelahavarong et al. 29 which were collected from severe thalassemia patients (same as our study). This study interviewed 28 severe thalassemia patients who received hyper-blood transfusion and iron chelation treatment at Ramathibodi hospital using questionnaire. Direct non-medical costs were estimated US $ 1,240 (SE = 232) and US $ 636 (SE = 221) for indirect costs.

We measured health-related quality of life of RI-HSCT patients using SF-36V2 [Thai version of the SF-36 (short-form survey) version 2.0] and EQ5D (EuroQoL EQ-5D). In Thailand, the validated official Thai version of EQ-5D is available. The answer from Thai version of EQ-5D questionnaire can be converted into 0–1 utility score by Thai preference score for EQ-5D health state 30 . Because of its practicality and accepted reliability, EQ-5D is the recommended utility measurement method in guideline of Thailand’s health technology assessment 9 . The other advantage of EQ-5D is that it is easier for patients to answer the questionnaires compared to decision making in preference elicitation techniques. We used utility derived from EQ-5D as main analysis and SF-36 as sensitivity analysis. The SF-36 is a multipurpose, short-form health survey with only 36 questions. It yields an eight-scale profile of scores as well as physical and mental health summary measures 17 . The questionnaire consists of 36 items that cover 8 aspects of health-related quality of life including physical functioning, role limitations due to physical problems, social functioning, bodily pain, general mental health, role limitation due to emotional problems, vitality (energy/fatigue) and general health perceptions. SF-36V2 31 (Thai version) is a widely used instrument with high validity and reliability. The utility scores were estimated via SF-6D derived from the SF-36. SF-6D was converted using SF-36 data at the individual level to single utility score using regression model 32 . We adopted the approach of Delea et al. 12 in deriving utility values for severe thalassemia patients without cardiac complication from the literature. The utility value was 0.61 (0.66). These values were taken from an Australian study 33 , using a time trade-off (TTO) exercise to elicit utility values from a sample of 120 individuals of various demographic characteristics. For health state with cardiac complications, we also adopted the approach of Delea et al. 12 and deducted a utility of 0.15. This value was obtained from a cohort study which estimatd a disutility value of heart failure based on 1365 random samples of US adults in a community 34 .

Primary outcomes of interest were lifetime costs, QALYs gained, and ICER in US ($) per QALY gained. For base-case analysis, we calculated the expected lifetime costs and outcomes from each group of treatment. All future costs and outcomes were discounted at a rate of 3% per annum 35 . The results were presented as ICER for RI-HSCT vs BT-ICT

ICER = Cos t RI − HSCT − Cos t BT − ICT QALY s RI − HSCT − QALY s BT − ICT

The interpretation of cost-effectiveness of our findings is based on the statement of the Subcommittee for Development of the National List of Essential Drugs and the Subcommittee for Development of the Health Benefit Package and Service Delivery of the NHSO in 2007. The subcommittee set up the societal willingness to pay (WTP) threshold in Thailand to one to three times per capita Gross Domestic Product (GDP) 36 , approximating US $ 4,210 (128,403 THB) to US $ 12,263 (385,209 THB) per QALY gained in 2010 18 . These values are in accordance with the recommendations of the Commission on Macroeconomics and Health of World Health Organization, suggesting that health technologies with the incremental cost-effectiveness ratio (ICER) below the per capita GDP are considered very cost-effective, while those between one and three times per capita GDP being cost-effective, and ICER above three times per capita GDP indicate that a health technology is not cost-effective 37 .

One-way sensitivity analyses were performed to investigate the effects of altering parameters within plausible ranges including survival parameters, time to failure parameters, all costs, utilities, costs to charge ratio and discounting rate. The results of one-way sensitivity were presented using a tornado diagram. For the sensitivity analysis of discounting rate, we used the recommendation from WHO 35 including 0% discount costs and health outcomes, 0% discount health outcomes and 6% discount costs, and 6% discount cost and health outcomes. In addition, a probabilistic sensitivity analysis (PSA) was conducted to examine the effect of all parameter uncertainty simultaneously using a Monte Carlo simulation performed by Microsoft Excel 2003 (Microsoft Corp., Redmond, WA) 25 . The distributions of each probability were assigned following 38 : (a.) probability and utility parameters, in which their values ranged between zero and one, were specified to beta-distributions, (b.) costs, their characters were positively skewed and values were above zero, were assigned to gamma-distributions, and (c.) survival parameters were given to a log-normal distribution. A Monte Carlo simulation was run for 1,000 sets of the simulation to give a range of values for total costs, outcomes, and ICERs. Results of the PSA were presented as cost-effectiveness acceptability curves. The expected net monetary benefit (NMB) was calculated for each WTP threshold in Thailand (one and three GDP per capita).

A total of 18 patients receiving RI-HSCT were included in this study. Mean age was 13.7 years; range 9–18 years. Peripheral Blood Stem Cell (PBSC) was used as sources for stem cell transplantation for all patients. Patients were classified as class III according to Lucarelli or Pesaro classification 2 . Of 18 patients, two patients have comorbidity diseases (hypothyroid and diabetes). The average duration of follow-up is 2 years with the range from 1 to 8 years. Demographic description of these patients was shown in Table 2. Based on our cohort, 2 of 18 patients (11%) were failed of RI-HSCT. One was failed within 50 days, while another was 131 days. Only one patients died of transplantation-related complications (namely pulmonary infection).The Kaplan-Meier curve was displayed in Figure 2.

CSMBS: Civil servant medical benefit scheme; UC: Universal coverage; PBSC: Peripheral blood stem cell.

Our base-case analysis demonstrated the estimated life time cost for BT-ICT and RI-HSCT as US $ 73,928 (2,254,806 Thai baht) and US$ 114,000 (3,476,988 Thai baht), respectively, while the estimated QALYs were 14.11, and 26.49 QALYs, respectively. (Table 3) Compared to BT-ICT, the conventional therapy of severe thalassemia with hyper-blood transfusion and iron-chelating therapy, an incremental cost per QALY gained for RI-HSCT was US$ 3,236 per QALY.

QALY = Quality adjusted life year; RI-HSCT: Reduced intensity hematopoietic stem cell; BT-ICT: Blood transfusion combined with subcutaneous iron chelating therapy.

* ($US, year 2011 value).

A series of one-way sensitivity analyses as shown in Figure 3 showed that the most influential parameter was utility of severe thalassemia patients without cardiac complication patients. When varying the utility of severe thalassemia patients without cardiac complication patients from 0.3 to 0.9, the cost-utility ratio was shifted to US$ 1,843 and 5,364. Discounting rate was the second influential parameter.

When varying the discounting rate from 3% to 0% and 6%, the cost-utility value was shifted to US$ 1,792 and 5,087, respectively. When varying the cost to charge ratio from 0.8 to 1.5, the cost-utility ratio was shifted to US$ 641 and 3,689.

The result of 1,000 simulations of PSA showed that RI-HSCT was estimated to have higher cost and more effective compared to conventional therapy (BT-ICT) (Figure 4). Cost effectiveness acceptability curves (CEAC) showed the threshold values of US$ 4,210 (128,403 THB) and US$ 12,263 (385,209 THB), RI-HSCT have 71.3% and 99.9%, respectively of being cost-effective when compared with BT-ICT (Figure 5).