Resource utilization and cost effectiveness of treating acute promyelocytic leukaemia using generic arsenic trioxide

Summary

Arsenic trioxide (ATO)-based regimens are the standard of care for treating acute promyelocytic leukaemia (APL) and have replaced chemotherapy-based approaches. However, the cost of “patented” ATO is prohibitive because of patent rights. “Generic” ATO has been used in a few countries, but its implications for health resource utilization (HRU) and cost of treatment are unknown. We hypothesized that treating APL patients using generic ATO (APL-ATO) will be cost effective compared to the chemotherapy-based regimen (APL-CT). In a single-centre retrospective study, we used a bottom-up costing method to compare the direct medical cost of treatment and HRU between APL-ATO and APL-CT. These costs and the survival and relapse probabilities were imputed in a three-state Markov decision model to estimate the cost effectiveness of APL-ATO compared to APL-CT. The mean cost of treatment for APL-ATO (n = 30, $8500 ± 2078) was significantly less than for APL-CT (n = 30, $22 600 ± 5528) (P < 0·001). APL-ATO reduced hospitalization, antibiotic and antifungal usage (P < 0·001). In the Markov model, five-year treatment costs were significantly lower for APL-ATO ($11 131) than for APL-CT ($17 926) (P < 0·001). Treatment cost and health resource utilization were significantly lower for generic ATO-treated APL patients compared to the chemotherapy-based regimen.

Acute promyelocytic leukaemia (APL) is a distinct subtype of acute myeloid leukaemia (AML) and makes up 10–12% of AML cases. Once a disease with a dismal prognosis, APL now shows excellent cure rates (>80% five-year overall survival) with arsenic trioxide (ATO) (Lo-Coco et al., 2013). ATO-based regimens thus have become the standard of care and replaced the more intensive chemotherapy-based strategy. However, the cost of patented ATO is prohibitive, and remains a major hurdle in widespread use of ATO for treatment of APL (Lo-Coco et al., 2013; Tallman et al., 2015; Philip et al., 2015b).

Arsenic and its derivatives have been used as medicine for thousands of years (Frith, 2013). The first clinical use of ATO in APL was reported in the Chinese literature (Sun, 1992; Huang et al., 1995; Chen et al., 1996) but its production was patented at once in the USA after successful replication of its efficacy in the western world. This increased the cost of ATO to $550 per 10 mg dose (MSKCC, 2019). As a result, ATO is now the major driver of cost of treatment of APL with ATO-based regimens ($50 000 for the entire treatment). The expenditure on pharmacy itself comprises about 50–80% of the total cost of treatment with ATO compared to 12–15% with the all-trans-retinoic acid (ATRA) + chemotherapy-based regimen (Kruse et al., 2015; Tallman et al., 2015). Despite these high costs, patented ATO remains cost effective because of its excellent clinical efficacy (Schlenk et al., 2014; Kruse et al., 2015). Understandably, if we reduce the cost of ATO, we can control the total cost treatment for APL.

One potential way of achieving this is by substituting patented ATO with a less expensive “generic” ATO. Concerns with generic ATO include breach of intellectual property rights and the quality of the product. Experts have argued that neither of these are valid though, as generic ATO has been used in several centres before the patent, including that of the authors, with equivalent clinical outcomes (Cyranoski, 2007; Mathews et al., 2010). Notably, less expensive “generic” ATO has resulted in greater affordability of this curative treatment (Philip et al., 2015a).

Given the potential of “generic” ATO in increasing the access and reducing the cost of APL treatment in both developed and low- and middle-income (LMICs) countries (Kerr, 2013; Mathews, 2014), we need to evaluate its cost effectiveness. The aims of this study were (i) to estimate the total healthcare costs and health resource utilization (HRU) for APL patients treated using generic ATO, and (ii) to evaluate the cost effectiveness of generic ATO in comparison to a chemotherapy-based regimen.

Material and methods

Study setting, inclusion criteria and duration

We conducted a single-centre retrospective observational study at a tertiary referral centre in India. All newly diagnosed APL patients from 1 January 2011 until 31 December 2013 were included. For comparison of outcomes, AML patients treated with a chemotherapy-based regimen were chosen as the control. We chose this control group as our centre does not use patented ATO or chemotherapy-based regimens for treating APL. We assumed that if APL patients were treated using a chemotherapy-based regimen, the cost of treatment and HRU would be similar to that in this control group. We followed patients until 31 December 2016 (to allow a follow-up of at-least three years for each patient) or death. A sample size of 25 was calculated for 0·1 (10%) precision and a significance level α < 0·05 with a co-efficient of variation of 0·25.

Treatment protocols and cost of treatment

Protocols are detailed in Figure S1. Newly diagnosed patients with APL were treated on generic ATO-based protocol (APL-ATO) comprising an induction phase, one cycle of consolidation and six cycles of maintenance. The control cohort (APL-CT) received 3 + 7 regimen (Murphy & Yee, 2017) for induction followed by three cycles of high-dose cytarabine (Schwarer et al., 2018) for consolidation. Relapsed APL patients in both groups (R-APL) received an ATO-based clinical trial (Ganesan et al., 2016) followed by an autologous stem cell transplant (auto-SCT).

We used an activity-based costing method (Fig 1) with the perspective of the health service provider (HSP). Four major cost activities were identified that delivered care to patients with APL and incurred costs to HSPs. The unit price of each cost activity was calculated using expenditure on four cost drivers (bottom-up costing, see Data S1, Table SI). The product of frequency of utilization of cost activities and its unit price gave the total expenditure for each cost activity during treatment. Finally, the total cost of treatment was obtained by summing up the expenditure incurred for each of four cost activities and expressed in USD (2016, 1 USD = 61·86 Indian rupees). Costs were estimated for all three phases of treatment induction, consolidation and maintenance.

Fig 1.

Overview of bottom-up costing methodology, treatment protocols, and the three-state Markov model. Bottom-up costing method: Four cost activities which incurred cost to the health service provider each time a patient utilized this activity were identified. The unit cost of each of this cost activity was in turn determined using the expenditure on four cost drivers, thus termed as bottom-up costing. The frequency of cost activity utilization was obtained from electronic medical records and multiplied by the unit price resulting in total expenditure incurred for that cost activity. The sum of expenditure on providing each of the four cost activities during the entire treatment and follow-up period resulted in total treatment costs. Treatment protocols: Costs were calculated for each of the three treatment protocols (newly diagnosed APL treated with generic ATO, APL-ATO; relapsed APL, R-APL; newly diagnosed APL treated on chemotherapy, APL-CT) separately and presented as total and in phase-specific manner. Three-state Markov model: These costs were used to estimate five-year cost of treatment using a Markov model that was constructed with three possible health states: Stable disease, relapsed disease and death. Monthly transition probabilities were obtained from the survival data from this centre and from published literature. The outputs from this model were: five-year cost of treatment to for each patient using APL-ATO and APL-CT, and five-year costs to health service provider (HSP) assuming one new APL patient is treated each month by the HSP.

Markov model for cost effectiveness

Three-state decision model

To estimate cost effectiveness by modelling disease events such as relapse and death during treatment, a Markov decision analysis model (Fig 1) was constructed from the perspective of the HSP. The time horizon was five years. The basis for this model is from a published study on patented ATO (Kruse et al., 2015). It comprises a one-month-long cycle, at the end of which three mutually exclusive outcomes of treatment (stable disease, relapse or death) were assigned to each patient. All newly diagnosed APL patients entered the model in a stable disease state and at the end of the one-month cycle either remained in stable disease or had relapsed or died depending upon the transition probabilities (see below). Relapse was defined as (i) morphological (>5% blasts in the bone marrow) or (ii) molecular [positive real-time quantitative polymerase chain reaction (RQ-PCR) test for promyelocytic leukemia-retinoic acid receptor alpha (PML-RARA) transcript], or (iii) failure to achieve morphological remission at the end of induction or molecular remission at the end of consolidation. Relapsed patients were switched to an ATO-based clinical trial and remained in the relapse state in the model and could not move into stable disease. Treatment for second relapse (third-line treatment) was not modelled.

Cost inputs

The monthly cost inputs for each state in the Markov model were obtained using bottom-up costing. Costs of adverse events related to treatment regimen were not modelled as those were already included in the cost activities.

Transition probabilities

The transition probabilities included in the model were event-free survival (EFS) and overall survival (OS). We based these on previously published data from this centre for APL-ATO (Mathews et al., 2010), and from the GIMEMA study (Lo-Coco et al., 2010) for APL-CT. The monthly estimates for three transition probabilities—from stable disease to relapse, from stable disease to death and from relapse to death —were obtained from the Kaplan–Meier (KM) curves for EFS and OS. The graphical curve-fitting method using Microsoft Visual Basic in Excel was used to estimate the proportion of patients alive (OS) and alive without a relapse (EFS) at monthly time intervals from the KM curves fitted onto the Markov model. Microsoft Excel Solver calculated the transition probabilities (see attached excel file) for the Markov model such that the sum of the percentage absolute value difference between EFS and OS in the KM curves and the Markov model was minimized.

Model outputs

This model estimated the five-year treatment cost to the HSP for treating each APL-ATO and APL-CT patient assuming the current estimates for the EFS and OS. It also estimated the five-year cost to the HSP assuming one newly diagnosed patient with APL was treated every month for next five years.

Sensitivity analysis

We conducted a series of one-way (deterministic) sensitivity analyses to test the impact of changing specific input parameter values or model assumptions on the results and report in a tornado diagram.

Results

Patients, disease characteristics and clinical outcomes

Sixty patients were newly diagnosed with APL on or after 1 January 2011 until 31 December 2013, and received treatment with generic ATO (APL-ATO). The mean age was 34 + 3·2 years and 51% were male. In APL-ATO, at the end of the follow-up, 40 patients (66·6%) were in first complete molecular remission (CMR: undetectable PML-RARA transcript levels). Four patients (6·6%) experienced early induction deaths (<30 days), and sixteen (26·6%) had a relapse. All relapsed patients were salvaged with an ATO-based clinical trial followed by auto-SCT and achieved a second CMR. Median follow-up for surviving patients was 3·5 years [interquartile range (IQR) 2·9–4·5]. Out of the 40 patients in first CMR, thirty patients were randomly chosen for analysis. All sixteen relapsed patients were included for estimation of the cost for relapsed APL (R-APL). The control group (APL-CT) comprised thirty patients randomly chosen from the newly diagnosed and treated non-M3 AML patients during the same period. Their mean age was 44·2 + 5·3 years and 69% were male. The median follow-up duration for surviving patients was 11 months (IQR 1·4-15). In the APL-ATO versus APL-CT group, the baseline comorbidities were not different when compared using the variables in the Charlson Comorbidity Index. Only one patient in APL-ATO had diabetes mellitus, and in APL-CT, one patient each had a stroke, diabetes mellitus and HIV-AIDS.

Health-resource utilization

During the induction phase (Fig 2A), patients with APL-ATO showed shorter length of hospitalization (P < 0·001) and higher outpatient treatment days (P < 0·001) as compared to APL-CT. No difference was noted in the length of Intensive Care Unit (ICU) stay (P = 0·07). During the consolidation phase (Fig 2B), in APL-ATO, the hospitalizations and ICU stay were significantly lower (P < 0·001) whereas outpatient days were significantly higher (P < 0·001) as compared with APL-CT.

Fig 2.

Comparison of heath resource utilization between APL-ATO and APL-CT. (A) shows comparison of median number of days for hospitalization, ICU stay, emergency and outpatient visits for induction phase. APL-ATO (n = 30) received induction with ATO-based regimen and APL-CT (n = 30) received induction with standard 3 + 7 regimen. R-APL (n = 16) are relapsed patients who received reinduction using ATO-based phase II clinical trial. (B) shows comparison of median number of days for hospitalization, ICU stay, emergency and outpatient visits for consolidation phase. APL-ATO (n = 26, n = 4 early deaths during induction excluded) received one cycle of consolidation with ATO-based regimen and APL-CT (n = 30) received three cycles of consolidation with high dose cytarabine regimen. R-APL (n = 16) are relapsed patients who received consolidation using ATO-based phase II clinical trial followed by autologous transplant. (C, D) represents median duration of antibiotics and antifungals usage for treatment of infectious complications during induction (C) and consolidation (D) phase for the three treatment protocols. No patients received antibiotic or antifungal prophylaxis. (E, F) represents median number of blood transfusions (Packed red blood cells, Platelet Rich Concentrates and Fresh Frozen Plasma) during induction (E) and consolidation (F) phase for three treatment protocols. For all comparisons, P < 0·05 is considered significant. Three-way anova was used for statistical comparison. ***, indicates Statistically significant difference.

The use of antibiotics and antifungal agents during both induction and consolidation (Fig 2C, 2) were significantly lower (P < 0·001) in the APL-ATO as compared to APL-CT. APL-ATO received a higher number of blood transfusions compared to APL-CT (P < 0·001) during induction (Fig 2E). During the consolidation phase, when coagulopathy was no longer an issue, APL-ATO patients received significantly fewer blood transfusions as compared to APL-CT (P < 0·001) (Fig 2F). Figure 3 is a Sankey visualization comparing the flow of HRU in which the width of the arrows represents the total frequency of HRU.

Fig 3.

Sankey diagram showing flow of health resources during entire treatment for APL-CT and APL-ATO. This diagram visualizes the flow of cost of treatment and health-resource utilization in APL patients treated with ATO (APL-ATO)versus chemotherapy (APL-CT) using Sankey diagram. The width of each flow pictured represents the proportional quantity.

Cost of treatment

The unit cost of each of the four cost activities estimated using the four cost drivers is shown in Table SII. The cost of generic ATO was US$3·0 for each 10 mg dose. The unit (per day) cost of hospitalization for chemotherapy was $52·6, $52·7 and $108·0 for APL-ATO, R-APL and APL-CT respectively. The unit cost of ICU stay, antibiotic, antifungal administration, supportive care and laboratory/imaging tests was same for both the groups.

The mean total cost of treatment for APL-ATO (n = 30) was $8500 ± 2078, while for APL-CT (n = 30), the cost was three times higher at $22 600 ± 5528 (P < 0·001) (Fig 4A). In APL-ATO patients, the mean cost of induction and consolidation was $4400 ± 1581 and $900 ± 212 respectively, while in APL-CT, it was $9800 ± 3239 (P < 0·001) and $11 800 ± 4872 (P < 0·001) respectively (Fig 4B, 4). The APL-CT group incurred higher costs due to longer hospitalization ($1000 vs. $4215, P < 0·001) and higher complications including neutropenic infections ($1174 vs. $3776, P < 0·001) (Fig 4B, 4). In the maintenance phase of APL-ATO, the entire chemotherapy was given on an outpatient basis, with an average cost of $2100 ± 146 (Fig 4C). The cost of follow-up for three years including molecular monitoring was $700 ± 289 (Fig 4D). For R-APL (n = 16), the generic ATO-based clinical trial incurred a mean cost of $3500 ± 1220 and $2300 ± 372 during induction and consolidation respectively (Fig 4B, 4). The mean cost of an autologous-SCT was $4300 ± 159. The average total cost of treatment of relapsed APL patients using the generic ATO-based regimen followed by an auto-SCT was $11 300 ± 4915 (Fig 4A).

Fig 4.

Comparison of phase-specific and total cost of treatment for APL-ATO and APL-CT. (A) Represents the contribution of cost activities (chemotherapy administration and treatment of complications, supportive care-blood transfusions and G-CSF administration, and laboratory and imaging) to the mean cost of treatment during induction phase. Comparison across three different protocols is shown. APL-ATO (n = 30) received induction with ATO-based regimen, APL-CT (n = 30) received induction with standard 3 + 7 regimen and R-APL (n = 16) are relapsed patients who received reinduction using generic ATO-based phase II clinical trial. (B) Represents the contribution of cost activities to the mean cost of treatment during consolidation phase. Comparison across three different protocols is shown. APL-ATO (n = 26, 4 early deaths during induction excluded) received one cycle of consolidation with ATO-based regimen and APL-CT (n = 30) received three cycles of consolidation with high dose cytarabine regimen. R-APL (n = 16) are relapsed patients who received consolidation with ATO-based phase II clinical trial followed by autologous transplant. (C) Represents the contribution of cost activities to the mean cost of treatment during maintenance phase. Comparison across three different protocols is shown. APL-ATO (n = 26, n = 4 early deaths during induction excluded) received six cycles of maintenance (each cycle for 28 days) with ATO-based regimen; APL-CT did not receive maintenance; and R-APL (n = 16) are relapsed patients who received maintenance on the ATO-based clinical trial for six cycles. (D) Total and phase-specific treatment cost for the three different protocols are compared. For all comparisons, P < 0 05 is considered significant. Three-way anova was used for statistical comparison.

Cost effectiveness and five-year cost to the HSP

Using Markov analysis (Fig 5A, 5, attached Excel file), the five-year cost of treatment of each newly diagnosed APL patient (Fig 5C) using the generic ATO-based regimen was estimated to be $11 131 with the observed probabilities of event-free and overall survival for generic ATO. This was significantly less than that for treatment with chemotherapy ($17 926, P < 0·001). Furthermore, assuming the health service provider treats one new patient with APL every month, the cost to the HSP for the next five years was substantially lower (Fig 5D, $555 670 for APL-ATO versus $1 012 998 for APL-CT). The sensitivity analysis (Fig 5E) was conducted by changing the cost of each phase of treatment in APL-ATO by 25% to assess the impact on the five-year cost of treatment to the HSP. The largest change in the five-year treatment cost was observed by varying the cost of the induction phase followed by a maintenance phase for newly diagnosed APL patients (stable disease). This shows that the cost of the first-line induction phase is the major cost driver in treating APL patients.

Fig 5.

Predicted five-year cost of treatment for each newly diagnosed APL patient and to five-year cost to health service provider for treating one new APL patient per month. (A, B) Shows the estimation of transition probabilities for the Markov model using Microsoft excel solver and trial-based Kaplan-Meir curves for event-free and OS. (C) Represents the comparison of five-year cost of treatment of APL using generic ATO (APL-ATO) versus chemotherapy (APL-CT). Markov analysis was used to incorporate the costs of progression on generic ATO as compared to chemotherapy. (D) Represents the comparison of estimated five-year cost of to health service provider for treating APL patients using generic ATO versus chemotherapy, assuming one new patient with APL starts treatment every month. For all comparisons, P < 0 05 is considered significant. Student's t-test was used for statistical comparison. All costs in US dollars. (E) One-way deterministic sensitivity analysis was conducted to assess the impact of changing specific input parameter values or model assumptions on the results and reported in a tornado diagram. Cost of induction and maintenance phase in newly diagnosed patients (stable disease) show highest impact on five-year cost of treatment to the HSP. ***, indicates Statistically significant difference.

Discussion

Randomized controlled trials have shown that the ATO + ATRA-based regimen for treatment of newly diagnosed APL could decrease cytotoxic chemotherapy without compromising the excellent outcomes achieved by ATRA + chemotherapy protocols. This led to approval of ATO as first-line agent for treating APL (Lo-Coco et al., 2010; Burnett et al., 2015). However, the cost of “patented” ATO is high. The price of 10 mg ATO is $550 in the United States, CAD530 in Canada, and €390 in Europe (Cyranoski, 2007). The patent on ATO expired in 2017, providing an opportunity for generic ATO to be marketed (Warrell, 2007). In India, generic ATO has been available commercially for clinical use since the year 2004 (Arsenox; Intas Pharmaceuticals Ltd, Ahmedabad, India).

We conducted this study to ascertain the HRU, cost of treatment, and cost effectiveness of treating newly diagnosed APL using generic ATO as compared to the ATRA + chemotherapy-based regimen which was advocated before for low- and middle-income countriess (LMICs) (Ribeiro & Rego, 2006; Rego et al., 2013). To our knowledge, this is the first study using a bottom-up activity-based costing method to estimate the cost of treatment of APL patients using the ATO-based regimen. The primary driver of the costs in APL-ATO was blood product administration (30% of the treatment cost). In contrast, the major costs in APL-CT were chemotherapy administration (42%) and management of infectious complications (38%) related to prolonged myelosuppression. The ATO-based regimen reduced the cost incurred for inpatient administration of chemotherapy during induction, consolidation for newly diagnosed APL and during reinduction for relapsed cases. This is important for LMICs where inpatient facilities are often limited in terms of population-to-bed ratio. In the consolidation phase of therapy, there was a significant difference in costs between the ATO and chemotherapy-based therapy, in favour of ATO. Finally, the Markov analysis using the survival data on generic ATO and the AIDA-2000 trial; estimated a significantly lower five-year cost for treating APL patients using generic ATO. Finally, the Markov analysis using the survival data on generic ATO and AIDA-2000 estimated a significantly lower five-year cost for treating APL patients using generic ATO.

Published literature from developed countries has shown that the cost of treatment with the patented ATO-based regimen is higher than that of the chemotherapy-based regimen. These excessive costs are within the willingness-to-pay (WTP) threshold, however, and thus cost-effective. From an Italian payer perspective, the three-year pharmacy cost for ATO + ATRA was higher (€46 700 per patient) compared to ATRA + chemotherapy (€6500 per patient); medical costs for ATO + ATRA were €12 300 per-patient versus €30 200 for ATRA + chemotherapy (Kruse et al., 2015). This corroborates our results where the ATO-based regimen has an advantage in terms of a significant reduction in the need for supportive care. In another analysis (Tallman et al., 2015), total costs per patient ranged from $96 940 for the ATRA + chemotherapy regimen to $136 170 for the ATO + ATRA regimen. However, compared to the ATRA + chemotherapy regimen, incremental cost-effectiveness ratios (ICER) for ATO + ATRA were $4512 per life year (LY) saved and $5614 per quality-adjusted life year (QALY) gained. This was below the WTP threshold ($50 000–$150 000 per QALY), leading to the conclusion that the ATO + ATRA-based regimen was cost-effective. From the Canadian perspective, ICER of ATO + ATRA compared to ATRA + chemotherapy varied between $21 294/QALY and $56 933/QALY, which was below the ICER of many other oncology treatments now in use (Lachaine et al., 2014). In the relapsed setting, the cost of treating with patented ATO alone was reported to be $168 849 with an ICER of $20 551/QALY gained. Thus, ATO-based regimens are cost-effective in developed countries even after the increased total costs because of patented ATO. Our study, in contrast to this published literature on patented ATO, shows that generic ATO significantly reduced the total cost of treatment of APL and remains cost-effective. This is because of the significant reduction in the pharmacy cost for ATO.

The challenges of treating APL in resource-poor settings involve the shortage of facilities including inpatient beds, antibiotics, blood banks, trained health care personnel and the availability of funds for treatment. An effort to improve outcomes in developing countries was the International Consortium on Acute Promyelocytic Leukemia (IC-APL). They introduced ATRA + chemotherapy-based therapy in South American countries, resulting in a 50% decrease in early mortality and a 30% improvement in OS (Rego et al., 2013). But these outcomes with ATRA + chemotherapy-based protocols have remained inferior as compared to the ATO + ATRA-based regimen, given the aforementioned challenges (Ribeiro & Rego, 2006; Rego & de Lira Benìcio, 2013). Results from the current study show that the decreased costs and HRU with the generic ATO-based regimen could make this a better choice than ATRA + chemotherapy in LMICs.

The limitations of the study were related to the costing method. First, the present study was conducted in a single tertiary care centre in India, Thus, the costs could vary depending upon differences in prices of various resources such as salaries, cost of facilities, drugs, consumables and equipment at other centres. Second, there were assumptions in calculating and comparing the costs including the non-AML-M3 cohort as the control. Third, we did not consider the direct non-medical costs and indirect costs because of the retrospective nature of the study.

In conclusion, our study shows that the generic ATO-based regimen without chemotherapy is the most cost-effective way to deliver treatment for patients with APL and probably the best strategy to achieve the goal of universal access to treatment for APL.

Supplementary Material

Additional supporting information may be found online in the Supporting Information section at the end of the article.