Rising Drug Cost Impacts on Cost-Effectiveness of Two Chemotherapy Regimens for Intermediate Risk Rhabdomyosarcoma: A Report from the Children’s Oncology Group

Abstract

Background:

The Children’s Oncology Group clinical trial for intermediate risk rhabdomyosarcoma randomized participants to a combination of vincristine, dactinomycin, and cyclophosphamide (VAC) alone or VAC alternating with vincristine plus irinotecan (VAC/VI). Clinical outcomes were similar but toxicity profiles differed. This study estimates the cost differences between arms from the health care system’s perspective.

Methods:

A decision-analytic model was used to estimate the incremental cost-effectiveness ratio (ICER) of VAC versus VAC/VI. Protocol-required or recommended medications and laboratory studies were included. Costs were obtained from national databases or supporting literature and inflated to 2019 US$. Demographic and outcome data were obtained from the clinical trial and directed chart reviews. Life-years (LY) were estimated from life expectancy tables and discounted by 3% annually. Probabilistic sensitivity analyses and alternative clinical scenarios identified factors driving costs.

Results:

Mean direct medical costs of VAC were $164,757 and of VAC/VI were $102,303. VAC was associated with an additional 0.97 LY and an ICER of $64,386/LY compared to VAC/VI. The ICER was sensitive to survival estimations and to alternative clinical scenarios including outpatient cyclophosphamide delivery (ICER $49,037/LY) or substitution of alternative hematopoetic growth factor schedules (ICER $73,191-$91,579/LY). Applying drug prices from 2012 decreased the total costs of VAC by 20% and VAC/VI by 15% due to changes in dactinomycin and pegfilgrastim prices.

Conclusion:

Neither arm was clearly more cost-effective. Pharmaceutical pricing and location of treatment drove costs and may inform future treatment decisions. Rising pharmaceutical costs added $30,000 per patient, a finding important for future drug-pricing policy decisions.

Lay summary:

Two chemotherapy regimens recently tested side-by-side for rhabdomyosarcoma had similar tumor outcomes but different side effects. We compared the health care costs of each regimen and found that neither was clearly more cost effective. However, the costs of each treatment changed dramatically with choices of supportive medicines and location of treatment. Costs of treatment rose by 15-20% because of rising US drug costs not associated with the clinical trial.

Precise:

Neither vincristine, dactinomyin, and cyclosphosphamide alone nor alternating with vincristine plus irinotecan was clearly more cost effective for intermediate risk rhadomyosarcoma. Pharmaceutical pricing and location of treatment drove costs and may influence future treatment decisions.

Introduction

Treatment protocols for childhood rhabdomyosarcoma (RMS) rely on a chemotherapy backbone of vincristine, dactinomycin, and cyclophosphamide (VAC).¹ Irinotecan in combination with vincristine showed substantial activity in phase II trials against previously untreated² or recurrent RMS.³ The toxicity profile of irinotecan differs from dactinomycin and cyclophosphamide with more gastrointestinal and fewer myelosuppressive or infectious complications. Because of irinotecan’s favorable toxicity profile and anti-tumor activity, the recent Children’s Oncology Group (COG) trial, ARST0531 (ClinicalTrials.gov Identifier: NCT00354835), randomized frontline treatment for intermediate risk RMS to either VAC alone (VAC arm) or alternating cycles of VAC and vincristine plus five consecutive days of irinotecan (VAC/VI arm). Tumor outcomes were similar between arms on this trial.⁴

Costs of health care, particularly cancer care, are a cause of international concern. While much attention has been paid to the high cost of novel anti-cancer agents,^5–7 costs associated with established and generic agents have also increased.^8,9 Delivering chemotherapy and supporting a child through side-effects are significant efforts and highly dependent on the treatment regimen. Understanding the drivers of health care resource use may provide additional information for future decisions, especially if different treatment options have similar oncologic outcomes. In our previous evaluations of VAC, cyclophosphamide significantly impacted health care resource use due to requirements for renal and hematopoietic support and dactinomycin because of its cost-per-vial.^10,11 Because these agents were incorporated differently on the two arms of ARST0531, we estimated the costs of each arm using a cost-effectiveness analysis. We also examined potential effects of increasing drug prices by substituting previous prices of the same agents.

Methods

Decision Analytic Model

A decision-analytic model was constructed using deidentified aggregate results obtained from the 450 eligible and evaluable participants enrolled on ARST0531, targeted chart reviews, publicly available cost data, and existing literature. Hypothetical patients were randomized to receive age-adjusted VAC (14 cycles) or VAC/VI (7 cycles of each) as described in Hawkins, et al⁴ (Figure 1). This analysis was performed from the perspective of the health care system. The primary outcome was the incremental cost-effectiveness ratio (ICER) between VAC and VAC/VI treatment arms. All analyses were performed in TreeAge Pro (Williamstown, MA). Approval for this research was obtained from the Baylor College of Medicine Institutional Review Board and by the institutions performing primary medical record reviews. Additional model assumptions are outlined in eTable 1.

Figure 1:

Model of Treatment Demonstrating (A) a Simplified Decision-analytic Model and (B) the Age-based Dosing of Chemotherapies.

Inputs

The model considered non-research medical interventions during and up to two months after treatment on ARST0531. Pharmaceutical doses were estimated using height, weight, and body surface area for a median-aged child of each age range.¹² In our primary analysis we assumed cyclophosphamide would be given during a one-day hospitalization and all other chemotherapy would be given in the outpatient setting. Protocol-required medications and laboratory studies aimed to minimize toxicities were included. This regimen required hematopoietic growth factor (HGF) support after each dose of cyclophosphamide. The primary analysis assumed patients would receive 10 daily doses of filgrastim. Imaging, surgery, radiation therapy, and tests performed specifically for verification of eligibility onto the clinical trial were excluded because recommendations were the same between the two treatment arms and highly dependent on tumor location.

Failure, defined as cancer relapse or progression, secondary malignant neoplasm, or death of any cause, was assessed at weeks 15, 30, and 43 (end of treatment), and at 4 years from diagnosis (eTable 2) using clinical trial data. We used US life expectancy tables¹³ adjusted for potential impacts of childhood cancer and treatment^14,15 to estimate the remaining years of life for each age group (eTable 3). Survival was discounted by 3% annually.¹⁶

We included the following toxicities during or within two months of completion of chemotherapy: transfusions, admissions for infection or diarrhea, and oral antibiotics for irinotecan-induced diarrhea prophylaxis.^17,18 Specific details regarding these toxicities are not quantitated via the Common Terminology Criteria for Adverse Events version 4.0¹⁹ toxicity attribution scoring system used for this trial. To characterize toxicities and supportive care measures in more detail, we reviewed the primary medical records at four institutions for 31 patients enrolled on ARST0531 (eTable 4). Toxicities such as hepatopathy, anorexia, and neurotoxicity were excluded because of limited documentation and established interventions.

Costs

Costs for each input unit were obtained from national databases or supporting literature and inflated to 2019 US dollars (US$) using the Consumer Price Index.²⁰ Pharmaceutical costs were estimated by Average Wholesale Price (AWP)²¹ and rounded to the next vial size. Costs of laboratory tests, outpatient chemotherapy administration and physician services were estimated from Center for Medicaid and Medicare Services reimbursement rates for 2019 (eTable 5). Because the treatment duration was less than one year, no discounting was applied to costs.

Sensitivity Analysis:

We performed probabilistic, multivariate sensitivity analysis using 10,000 first-order Monte-Carlo micro-simulation trials. If parameters were not available for a specific input we assigned a standard deviation of 20%.²² Three alternative clinical scenarios accounted for common practice variations: 1) cyclophosphamide delivered in an outpatient rather than inpatient setting, 2) filgrastim vials were entered once only, and 3) single-dose pegfilgrastim replaced filgrastim.

We tested survival assumptions by applying discounting factors of 1.5% and 6%, event-free-survival (EFS) probabilities at 2, 3, and 5 years, and alternative life expectancies (eTable 3). Infants were excluded in another scenario. Potentials for toxicities were varied to the extremes of possibilities. To explore the impact of pharmaceutical pricing we replaced contemporary costs with those from 2012¹⁰ inflated to 2019 US$.

Results

Primary Analysis and Alternative Clinical Scenarios

For the primary analysis, the mean direct medical costs of VAC were $164,757 and $102,303 for VAC/VI, an incremental increase of $62,454/patient more for VAC (Table 2, eFigure 1). The average life year (LY) gained on the VAC arm was approximately one year greater than VAC/VI owing to a slightly higher but statistically insignificant EFS at 4 years.⁴ The ICER for VAC was $64,386/life year saved (LYS) gained compared to VAC/VI.

Table 2:

Results of Cost and Effectiveness for Primary Analysis and Alternative Clinical Strategies

	VAC		VAC/VI		Incremental cost	Incremental LY (years)	ICER (95%CI) ($/LY saved)
	Total cost (95% CI)	Total LY (95% CI)	Total cost (95% CI)	Total LY (95% CI)
Primary Analysis	$164,757 ($138,516–$197,835)	17.01 (16.80–17.21)	$102,303 ($85,550-$124,727)	16.04 (15.80-16.25)	$62,454	0.97	$64,386 ($45,244 - $96,796)
Cyclophosphamide given outpatient	$135,355 ($108,180-$168,299)		$87,789 ($70,238-$110,149)		$47,566		$49,037 ($32,656 - $76,664)
Filgrastim single use vial only	$191,594 ($162,577 – $226,394)		$119,867 ($99,965 - $144,878)		$71,727		$73,191 ($52,081 - $110,552)
Pegfilgrastim instead of filgrastim	$240,938 ($195,047-$295,333)		$152,106 ($122,535-$187,169)		$88,832		$91,579 ($64,877 -$136,308)

Abbreviations: LY: Life year, VAC: vincristine, dactinomycin, cyclophosphamide treatment arm; VAC/VI: VAC alternating with vincristine plus irinotecan treatment arm

In the alternative clinical scenario of delivering cyclophosphamide in the outpatient setting the costs of the VAC and VAC/VI arms decreased by 17.9% and 14.2% respectively, and the ICER decreased by 23.8% to $49,037/LYS. Conversely, when filgrastim vials were only entered once, the costs of VAC and VAC/VI increased by 17.1% and 14.9% respectively and increasing the ICER by 14% to $73,191/LYS. Single dose pegfilgrastim increased the costs of VAC and VAC/VI by 46.2% and 48.7% respectively while increasing the ICER by 42% to $91,579/LYS.

Sensitivity Analysis

ICERs demonstrated sensitivity to survival assumptions. Varying annual discount factors applied to survival from 1.5-6% resulted in ICERs from $47,712-$115,558. If the difference in survival was assessed at 2 years the ICER was modestly higher than other time points (eTable 6). Removing infants from the primary analysis decreased the difference in life years between arms to 0.65 and increased the ICER to $95,960/LYS.

The effects of toxicities on costs were studied by varying the probability of individual toxicities to extremes (Figure 2, eTable 7). Transfusions had minimal impact on overall costs or cost differences between arms. ICERs ranged from $31,129 if all patients receiving VAC and no patients receiving VAC/VI had admissions for toxicity to $89,930 if the inverse scenario occurred.

Figure 2:

Sensitivity of Incremental Cost Effectiveness Ratios to Toxicities

Abbreviation: A: Dactinomycin, C: Cyclophosphamide, I: Irinotecan, V: Vincristine

2012 versus 2019

When 2012 drug prices were applied the total average cost of VAC decreased by $33,495 (20.3%), VAC/VI decreased by $14,947 (14.6%), and the ICER decreased by 30% to $45,264. The majority of this decrease was due to increases in the per-vial-cost of dactinomycin (Figure 3, eTable 8). In the alternative clinical scenario using pegfilgrastim the total mean cost of VAC decreased by $35,443 (14.7%), VAC/VI decreased by $23,163 (15%) and the ICER decreased by 13.8% to $78,919.

Figure 3:

Change in Total Cost and ICER by Substituting 2012 Prices of Pharmaceutical Agents

Discussion

We examined the health care costs of two treatments for intermediate risk RMS which reported similar survival outcomes in a previously reported multi-institutional randomized trial. The VAC treatment arm was more costly than VAC/VI in all common clinical scenarios by approximately $47,000-$89,000, and toxicity profiles were not major determinants of cost differences between arms. Pharmaceutical pricing and location of chemotherapy delivery contributed significantly to the costs of each arm and to incremental costs.

The United States does not have a formal cuttoff for what should be considered cost-effective via an ICER calculation. A threshhold of $100,000 is frequently referenced but does not necessarily represent societal beliefs considering how the nation actually spends its health care dollars.²³ Studies on pediatric therapies have additional intricacies because of inherent but difficult-to-quantify societal values of children.²⁴ Although the ICER from our primary analysis was below $100,000 we do not conclude that VAC is a better choice based on cost-effectiveness. However, the factors affecting the results of this model provide important insight into future use of these regimens.

Our analysis was not designed to assess anti-tumor outcomes of the treatment arms; that was the purpose of the original clinical trial. However, we found that VAC, while more costly, was also associated with approximately one additional LYS over VAC/VI. This model combined clinical data for outcomes²⁵ with assumptions about potential long-term survival. Patients who survive childhood cancer are at risk for numerous long-term consequences linked to the therapy they received, but the influence of either arm of this study on late effects is unknown. When predicting long-term outcomes and survival after treatment we applied conservative methods of adjusting potential overall life expectancy equally to both arms. Changes to our assumptions, such as alternative discounting methods and timing of survival assesment after finishing treatment, changed the incremental survival differences. When we removed infants from the model during sensitivity analyses, for example, the difference in long-term survival decreased from 0.97 LYS to 0.65 LYS and the ICER increased to $95,960/LYS. Infants were less than 8% of all patients, but more failed during treatment on VAC/VI than on VAC. They also have a longer predicted life expectancy than older children. These small differences heavily weighted the average life years gained. Clinical decisions regarding the efficacy of VAC/VI to infants should not be made using this model. Rather, the increase in the ICER by removing infants from the model demonstrates sensitivity of this model to life expectancy assumptions.

The fundamental difference in the two treatments is the substitution of six doses of dactinomycin and seven doses of cyclophosphamide with irinotecan over the 42 weeks of treatment. Dactinomycin is an established and successful agent in the treatment of RMS, it is also the most costly of the four chemotherapy agents included. The number of dactinomycin doses and age and weight-based dosing was a driver of differential costs in this model and for a previous analysis of low risk RMS treatment.¹⁰

This regimen calls for HGFs, well-known costly supportive components of cancer treatment.^26,27 Guidelines exist for use of HGF as prophylaxis in adult oncology chemotherapy regimens when the expected incidence of febrile neutropenia is at least 20% or if a patient has had previous episodes of febrile neutropenia.²⁸ Pediatric regimens incorporate HGF as primary prophylaxis to maintain dose intensity in the absence of well-established rates of febrile neutropenia. This analysis was not designed to consider the merits of HGFs in this treatment plan. Our goal was to estimate the impact on costs of HGFs on costs.

Both arms increased in total cost when we substituted alternative HGF dosing (pegfilgrastim or single entry filgrastim), but VAC more so than VAC/VI because of the number of cyclophsphamide/HGF cycles in each arm. Initial pricing for pegfilgrastim was designed to replace a course of daily dose filgrastim for adult patients; cost comparisons for these two agents in children are more complex due to billing requirements for single-use pharmaceutial products.²⁹

Our exploration of HGF highlights the potential for wasted drug. We assumed patients would enter filgrastim vials twice, a common practice despite the single-use design. Following the FDA recommended single-use design increased the total cost by $17,000 to $27,000 (because twice as many pre-filled vials were required) but no additional agent was given to patients.³⁰ A single dose of pegfilgrastim is an appealing replacement for multiple daily doses of filgrastim to reduce the number of injections, ensure adherence, and improve quality of life in pediatric patients.³¹ Per the prescribing information of pegfilgrastim,³² patients weighing less than 45 kg must be dosed by weight.¹² We estimate that approximately 68% of children on ARST0531 would be less than 45 kg and partially waste a 6 mg prefilled syringe. We did not include any additional costs of pharmacy preparation for HGF, potentially underestimating the health care costs for smaller patients.

We previously published the 2012 AWPs of the VAC agents in an analysis of low-risk RMS treatment.¹⁰ When substituting the 2012 AWP of these agents into the model of ARST0531, the average costs of treatment were 15-20% lower than with 2019 prices. This change was almost entirely due to increases in AWP of dactinomycin and pegfilgrastim between 2012 and 2019. These results highlight the impacts of rising generic drug prices on overall treatment costs and assumptions of cost-effectiveness over time. We used the AWP of pegfilgrastim for this analysis as it was the agent available during the trial. Additional biosimilars have since entered the market with AWPs lower than the original product. These lower prices may decrease the total cost of the treatments, they are higher than filgrastim and remain a substanital cost driver.

Our intitial model assumed cyclophosphamide would be delivered in the inpatient setting. Rapid hydration, oral mesna, and/or prolonged day-hospital options allow safe³³ and less costly¹¹ outpatient delivery of this regimen’s cyclophosphamide dose where institutionally possible. When we substituted outpatient delivery into our model the total costs of VAC decreased by $29,000 and VAC/VI by $14,500 demonstrating potential health care saving of outpatient versus inpatient delivery.

Toxicities had only modest effects on the overall costs of treatment on either arm or to the ICER. A limitation of retrospectively accounting for toxicities was the use of the CTCAE criteria for grading toxicities^19,34 and targetted collection of only grade 3 or 4 toxicities. This scoring method captures the most severe grade of each toxicity for the reporting period, not the resources used. For example, a reported grade 3 anemia indicates that the participant’s hemoglobin was <8.0 g/dL at least once during that reporting period but does not capture the number of transfusions given. We reviewed the primary medical records of a subpopulation of ARST0531 participants to supplement the CTCAE grading. Our results from these reviews match our clinical experiences, but remain limited because of the inherent challenges of retrospective data collection on a rare tumor. To attempt to overcome some of these limitations we varied the probability of toxicites to extreme, unlikely conditions. Our ability to accurately assess the cost and resource impacts of toxicities will continue to be a limitation unless we address prospective data collection. VAC/VI was associated with less myelosuppression and infection and was incorporated into the subsequent COG RMS trial, ARST1431.³⁵ Compared to historical results, the local control rate was worse on ARST0531 for patients with embryonal RMS.³⁶ Cyclophosphamide dosing and timing of radiation are two hypothetical reasons behind the worse outcomes, but in the absence of a side-by-side comparison it is difficult to know for sure. Results from our current analysis may be useful to inform the design of future comparisons.

This analysis focused on the perspective of the health care system by comparing direct medical costs of treatment and life expectancy. The choice to use this perspective stemmed from data availablilty. The financial and emotional burdens families incurred during treatment such as travel, out-of-pocket expenses, or lost earnings are substantial and potentially devastating.^37,38 Lost wages and travel costs are significant components of family financial burden^39–41 and increase with frequency of scheduled and unplanned medical encounters.⁴² Outpatient encounters are not necessarily less burdensome than inpatient encounters.⁴³ Despite the concerns about treatment impact on families and society and the recommendations that these perspectives be included in cost analyses,⁴⁴ patient or societal costs are rarely considered.^44,45 Unfortunately the current methods of capturing patient costs and personal utility are burdensome in themselves^39,46,47 and therefore rarely performed in prospective clinical trials, including ARST0531. Easier methods to measure family burden are needed; in the meantime prospective trials should avail themselves of the available tools and compare the impact on families. If VAC/VI is indeed more burdensome for families, reverting to VAC is not the only alternative. Other solutions could include home health delivery or oral irinotecan dosing.^48,49

In conclusion, the costs of RMS treatment are increasing. There are opportunities to decrease overall costs at the individual provider level (e.g. delivering care in the outpatient setting, choice of HGF). However, rising costs of generic established therapies are outside of the direct influence of the health care provider and support a need to address greater drug-pricing policy issues.

Table 1.

Average Wholesale Prices of Pharmaceuticals

Cost Inputs	2019		2012^
	Mean	Range	Mean	Range
Chemotherapy
Vincristine 1 mg vial	$6.90	$5.52-$8.28	$6.60	$5.28-$7.92
Dactinomycin 0.2 mg vial	$1,730	$1,384-$2,076	$799	$639-$959
Cyclophosphamide 500 mg vial	$396	$317-$475	$166	$133-$199
Irinotecan 40 mg vial	$17.45	$13.95-$20.9	#
Supportive Drugs
Mesna 1000 mg	$36.20	$28.96-$43.44	$68.93	$55.10-$82.73
Ondansetron 4 mg tablet	$22.67	$18.14–$27.20	$43.54	$34.84-$52.25
Filgrastim 300 mcg vial	$389	$311 - $467	$438	$350-$525
Pegfilgrastim 6 mg vial&	$7,477	$5,982-$8,978	$4,904	$3,923-$5,885

^{^}Inflated to 2019 US$

^#2012 Irinotecan average wholesale prices not previously published

^&Amgen Inc