Utilization Impacts of New Oral Substitutes for Parenteral Cancer-Related Therapies

Parenteral administration has been the most commonly used route for antineoplastic agents and many supportive care drugs used in oncology. Yet, over the past decade, multiple new oral drugs have been approved by the FDA for use in cancer patients. The National Comprehensive Cancer Network (NCCN) 2008 Task Force Report on Oral Chemotherapy estimated that one-fourth of the 400 new antineoplastic drugs in the pipeline were oral formulations^1. Many of these new oral agents represent unique therapeutic entities, whereas others provide close substitutes for existing parenteral therapies.

Oral drugs may differ from parenteral therapies in a variety of clinically important ways, including pharmacokinetics, pharmacodynamics, and toxicities. Because the logistics of therapy are often much simpler for oral compared with parenteral medications, oral drugs are often preferred by patients as long as efficacy is equivalent.² Although the convenience and toxicity profile of oral agents may expand the pool of patients able to receive therapy, adherence, defined along the dimensions of both duration and dose adequacy, requires patient cognitive skills to manage medications and self-monitor for treatment-related side effects, compared with the equivalent parenteral therapy.

Although many studies have addressed adherence to oral medications, few empirical studies have examined how utilization and adherence change when an oral formulation of an existing cancer-related therapy is introduced. Examples of direct oral substitutes for parenteral therapy are still relatively scarce. Where there are close substitutes (eg, capecitabine as a substitute for 5-fluoruracil), data to compare utilization patterns for both oral and parenteral versions are limited.³

Myelodysplastic syndromes (MDSs) comprise a heterogeneous group of clonal hematopoietic stem cell disorders characterized by ineffective hematopoiesis, leading to peripheral blood cytopenias and variable risk of progression to acute myeloid leukemia (AML). Despite the significant clinical associated with disease-modifying agents—such as the hypomethylating agents (HMAs), and anemia-palliating agents such as lenalidomide—MDS patients eventually experience disease progression.^4,5 Most patients will require packed red blood cell (PRBC) transfusions during the course of the disease, and as many as 40% will develop transfusion dependence.⁶ Because each PRBC unit contains 200 to 250 mg of elemental iron, and the human body does not have a physiologic mechanism to eliminate iron, chronic transfusion therapy places these patients at significant risk of developing transfusional iron overload and its associated complications.⁷

Although no data clearly demonstrate a benefit to iron chelation therapy (ICT) in MDS patients with ongoing transfusion requirements, current practice has adopted ICT, extrapolating from experience with chronically transfused thalassemia patients.⁸ Current NCCN guidelines recommend ICT in MDS patients with a serum ferritin level greater than 2500 mg/mL.^9,10 Both parenteral and oral ICT are available. Deferoxamine (DFO), approved by the FDA in 1968,¹¹ is administered by subcutaneous infusion over 8 to 12 hours, 5 to 7 days each week, although it may also be administered intravenously or by intramuscular injection. Deferasirox (DFX) is an oral iron chelator dosed once daily as tablets dissolved in liquid that was approved by the FDA in November 2005.¹²

Although administration of DFX is easier and more convenient, Medicare reimbursement policies may make DFX receipt difficult for many patients. DFO is reimbursed through Medicare Part B with a deductible and 20% coinsurance. However, 90% of Medicare beneficiaries have supplemental medical coverage that may limit the level of cost sharing associated with DFO.¹³ In contrast, DFX is reimbursed through Medicare Part D, which covered approximately 63% of Medicare beneficiaries in 2012.¹⁴ Many Part D plans have tiered formularies with variable cost sharing. High-cost drugs, such as DFX, are commonly placed on tiers requiring large co-payments and/or high coinsurance rates. Only beneficiaries receiving the Part D Low Income Subsidy (LIS) are protected from most cost-sharing requirements.¹⁵

In this study, we examined ICT use in a cohort of Medicare beneficiaries with extensive transfusion exposure due to anemia associated with MDS. Specifically, we examined whether ICT utilization rates were affected by the introduction of DFX into the market, and whether duration of therapy and dose adequacy improved with the use of DFX.

METHODSData Source and Cohort Selection

We used a unique database of Medicare beneficiaries with MDS generated from 100% of Medicare enrollment and claims files from 2004 through 2008. Claims for Medicare Parts A, B, and D included detailed information on dates, diagnoses, and services provided based on International Classification of Diseases, 9th Revision, Clinical Modification (ICM-9-CM) procedure codes, Healthcare Common Procedure Coding System (HCPCS) codes, and/ or National Drug Codes (NDCs). Patients were included if they had at least 1 inpatient claim with an MDS diagnosis (ICM-9-CM code 238.7 before October 2006, code 238.72-5 beginning October 2006). In the absence of an inpatient claim, we searched for a minimum of 2 outpatient claims separated by at least 29 days, but no more than 12 months. The gap between outpatient claims was designed to minimize the inclusion of patients with an unconfirmed diagnosis. The MDS index date was set at January 1, 2005 for those who qualified for MDS in 2004, otherwise at the date of the first claim with a qualifying diagnosis. Patients were required to meet a threshold of either 10 transfusion episodes or 20 PRBC units during the period starting from 12 months before their index MDS date, representing the minimum PRBC exposure needed to experience elevated serum ferritin levels. eAppendix A (available at www.ajmc. com) illustrates the sample selection process. The observation period began the week immediately after the MDS patient met the transfusion threshold (cohort entry date) and continued until death or end of study. Patients who met the transfusion threshold before their MDS index date, received ICT before the observation period; those who progressed to AML before cohort entry were excluded.

Beneficiaries had to be enrolled in Medicare Parts A and B, but not enrolled in a Medicare Advantage plan, for the entire observation period to ensure completeness of medical claims data. In addition, to ensure that all prescription drug transactions would be captured, the cohort was limited to those patients who were continuously enrolled in Medicare Part D from January 2006 or the cohort entry date, whichever was later.

eAppendix B

summarizes the cohort selection results.

Measurement

The main independent measure was the period of ICT-eligible cohort entry relative to DFX market entry in 2006. We summarized cohort entry dates in 6-month intervals. Baseline patient characteristics were determined from Medicare enrollment files. The MDS risk group was determined based on ICD-9-CM diagnostic codes on claims that qualified the patient for the study. Patients diagnosed before October 2006 were all assigned to MDS not otherwise specified (NOS) because the ICD-9-CM code (238.7) did not distinguish risk group. Otherwise, patients were assigned to lower-risk (238.72), chromosome 5q deletions (del 5q-MDS; 238.74), higher-risk (238.73), or NOS (238.75) groups. To capture Part D program cost sharing, we identified beneficiaries receiving the LIS during the full observation period. Zip code-level sociodemographic factors were linked, including median household income, education, proportion of households with English language difficulty, region of the country, and urbanicity.^16,17 Information on additional PRBC units transfused after cohort entry and medication use for MDS management, including exposure to erythropoiesis-stimulating agents, HMAs, and lenalidomide, was captured on a weekly basis, and cumulated over time. We captured baseline and time-varying indicators of comorbid conditions that might be considered as contraindications to ICT treatment or indicators of overall poor prognosis. Disability status is a claims-based measure that uses regression weights to generate an indicator for a high probability of poor functional status.¹⁸ Clinical condition indicators were measured at cohort entry, whereas weekly indicators updated the health status information to account for subsequent development of new conditions.

ICT utilization was characterized as receipt of any therapy, dosage form (DFO or DFX), duration of treatment, and average dose, based on specific HCPCS and NDC codes in Part B claims, and NDC codes in Part D claims. Duration of therapy was defined as the number of episode-weeks with evidence of ICT as specified in the claims. eAppendix C describes the measurement of treatment episode-weeks and illustrates the episode measurement process. The average weekly dose was calculated for ICT episodes by summing the charged quantity of DFO or DFX from all claims during the follow-up period and then iding that sum by the episode duration. Assuming an average weight of 70 kg per patient, we categorized DFO dose into 3 groups according to criteria modified from the study by Rose and colleagues¹⁹: adequate (>8000 mg/week), low (3000-8000 mg/week) and very low (<3000 mg/week). Similarly, a standard of 1000 mg/day of DFX was used to define an adequate DFX dose; we did not distinguish low from very low doses of DFX because there was insufficient sample to stratify further.

Analysis

Univariate and bivariate analyses described patient characteristics and ICT exposure. Cox proportional hazards models, using the time-varying (weekly) observations, estimated the effect of cohort entry period on risk (probability) of ICT initiation. Use of time-to-event models permitted us to follow patients over variable observation periods. Competing risks models were estimated separately for DFO and DFX, with use of the alternative ICT or death considered as competing risks. Among patients using only 1 drug (DFO or DFX), we used Kaplan-Meier analysis and Cox proportional hazards models to compare times to discontinuation. All analyses were conducted using SAS version 9.2 (SAS Institute Inc, Cary, North Carolina) and Stata version 12 (StataCorp, College Station, Texas). The study was approved by the University of Maryland Institutional Review Board. Through the authorship role of Rahul Shenolikar, the sponsor participated in the design of the study, the interpretation of the data, and the review of the manuscript.

RESULTS

Characteristics of the ICT-eligible cohort at baseline are summarized in

Table 1

. The full cohort of 3843 beneficiaries was 46.3% male and 90.5% white, with a median age of 78 years. By ICD-9-CM codes, 13.6% had lower-risk MDS, 2.1% had higher-risk MDS, 5.5% had chromosome 5q deletions (del 5q-MDS), and the remaining patients had MDS-NOS. Cohort entry was skewed to the post 2006 period, but 445 (14.2%) qualified for ICT in 2005. Relatively few beneficiaries received ICT (532, 13.8%); of this group, most (434, 81.6%) received DFX, and 157 (29.5%) received DFO. A small number (59, 11.1%) received both agents. Relative to the latter half of 2005, the probability (risk or hazard) of ICT initiation increased by 2007, with a hazard ratio (HR) of 1.814 (95% confidence interval [95% CI], 1.15- 2.87; P = .012) associated with cohort entry in the latter half of 2008 (Table 2). Additional predictors of increased ICT initiation risk included cumulative PRBC units (HR, 1.0003; 95% CI, 1.0002-1.0004; P <.001), prior HMA use, and black race. Females (HR, 0.73; 95% CI, 0.61-0.88; P =001) and those with predicted poor disability status (HR 0.62; 95% CI, 0.44-0.88; P = .007) were less likely to initiate ICT. Conditions including arrhythmia and renal disease, events including gastrointestinal hemorrhage or hemorrhage at other sites, a new diagnosis of pancytopenia, and hospitalization for sepsis were associated with reduced likelihood of ICT initiation. We did not find an effect associated with Part D LIS receipt.

<

Table 3

further explores the role of cohort entry period in initiation of ICT, with DFO and DFX initiation as competing events and with censoring at death or end of study. We observed lower rates of DFO use after 2006, although there is no clear time trend and the estimates were not significant. The estimated effects of time period on DFX initiation were similar to those of ICT initiation overall, with increased use starting in 2007. Similar to results for any ICT, cumulative PRBC units and HMA exposure were associated with increased use of both DFO and DFX, while gastrointestinal hemorrhage and pancytopenia were associated with decreased use. Poor disability status and new diagnoses of arrhythmia, renal disease, and AML were associated with decreased use of DFX but not DFO. A new diagnosis of diabetes was associated with increased use of DFO only.

Characteristics of DFO and DFX users are reported in eAppendix D. Treatment duration and dose for beneficiaries who received only DFO or DFX during the observation period are reported in Table 4. Treatment duration was longer for beneficiaries using DFX compared with DFO (median 47 vs 17 weeks; P <.01). The Figure illustrates the relationship between the agent used and the duration of therapy. Mean dose of DFO was 2695 mg/week; fewer than 10% of users received an adequate dose. Mean dose of DFX was 1334 mg daily, with 72.8% receiving an adequate dose.

DISCUSSION

There is growing literature on the role of patient preferences and socioeconomic characteristics in adherence to oral cancer-related therapy,^20-22 but most empirical estimates of adherence are reported within the context of clinical trials, where utilization is carefully monitored. Additionally, clinical trial participants may be subject to the Hawthorne effect, whereby utilization may be influenced by trial participation.²³ This is the first population-based observational study to examine initiation of and adherence to a cancer-related therapy when an oral substitute for a parenteral formulation becomes available for patient use. Our findings, focusing on ICT use in Medicare beneficiaries with MDS who met a minimum threshold of PRBC, suggest that market entry of the oral agent DFX was associated with increased use of therapy. Furthermore, patients using the oral agent were more likely to adhere to therapy and receive chelation therapy at an adequate dose. The overall effect on ICT use represents the combined process of small declines in DFO initiation and a large increase in DFX use. The net effect is an expanded pool of ICT users.

One of the motivations for the pharmaceutical industry to develop new oral agents was the expanded access to prescription drug benefits in the Medicare population associated with the implementation of the Medicare Part D program. An important question is, how will the availability of oral formulations of cancer-related therapies affect spending by the Medicare program, whether through Part B or Part D? The budget impact of a new drug is a function of the cost of the new oral regimens compared with the parenteral therapies, the number of users, and the quantity of use.²⁴ Prior estimates of cost for a full year of ICT for MDS, assuming full dose, were $21,048 and $46,008 for DFO and DFX, respectively, suggesting a doubling in Medicare spending associated with DFX availability. ²⁵

To quantify the full effect of utilization and price changes, we calculated expected cost of a course of ICT per user, incorporating unit drug costs and observed dose and duration (adjusted for censoring). The cost of DFO, including daily administration, was abstracted from Medicare fee schedules; for DFX, we calculated total reimbursement per unit from Part D claims. The calculated treatment cost per beneficiary, assuming the median duration, was $2761 for DFO and $46,142 for DFX. Our results reveal that the increased duration and dose adequacy associated with oral ICT contribute substantially to an overall increase in drug-related spending associated with use. Although we did not calculate it specifically, we expect that patient out-of-pocket spending would be similarly affected by the higher price of DFX, the higher cost sharing associated with Part D coverage (absent the LIS), and the greater quantity of ICT used. Failure to take changes in utilization into account may result in underestimates of the impact on spending associated with new oral therapies.

The decision to treat with ICT may be heavily influenced by patient-specific factors including the ability to manage therapy and the cost of treatment. Contrary to expectations, we did not find an effect of LIS receipt on ICT initiation overall or even when we looked separately at DFX. An examination of cost sharing for DFX using the 2013 Part D plan finder website indicated co-payments ranging between $35 and $90 monthly during the initial coverage phase, whereas plans using coinsurance rates required patients to cover up to 43% of the total cost.²⁶ In addition, cost sharing is substantially greater during the coverage gap; one plan required a payment of $3080 monthly, with $308 in the catastrophic zone.²⁷

In contrast, beneficiaries with LIS would be required to pay only up to $6.50 monthly, with no contribution in the catastrophic zone. It is possible that Medicare beneficiaries are not particularly responsive to out-of-pocket cost sharing when there is a strong perceived benefit of the treatment.^28,29 However, given the absence of clinical evidence for an impact of ICT on survival specific to MDS patients, physicians would not likely recommend that patients bear a large out-of-pocket burden to pay for therapy. An alternative hypothesis is that patients timed their DFX initiation to coincide with the initial coverage phase, but discontinued therapy when they reached the Part D coverage gap. Further research on price response and the role of Part D benefit design for cancer patients is clearly warranted.

Our study focused on ICT, which is a supportive-care measure in MDS. Although we report better adherence with the oral agent in our study, we acknowledge that the magnitude of our findings may not be directly applicable to antineoplastic therapy. Medication adherence is affected by perceived clinical benefit. To the extent that antineoplastic therapy is perceived by patients to have a greater benefit than a supportive-care drug, ease of administration may play a lesser role in determining adherence. Hence, the availability of an oral substitute for antineoplastic therapy may have less of an impact on use than what we observed with ICT. We also note that this study focused on cancer-related therapies. Although the pattern of substituting oral for parenteral therapy is likely most prevalent in cancer therapy, it is not unique to this clinical area. It is not clear whether the results from this study would generalize to patterns of treatment in other areas.

Limitations

We acknowledge a variety of study limitations associated with the use of administrative data. Our MDS cohort was selected using ICD-9-CM codes, which can be subject to errors, and our assignment of MDS risk categories was limited by the lack of detail before October 2006 and failure to designate specific risk categories even after the fifth-digit ICD-9-CM codes became available. With claims, we lacked information on clinical parameters such as ferritin levels. Hence, we were unable to make detailed determinations about eligibility for ICT. Our measures of ICT utilization were based on claims paid, but we cannot confirm that drugs were actually consumed by patients. Finally, our post hoc analysis of Medicare spending focused on the drug and administration costs, but did not address whether there were downstream differences in toxicities or adverse events and the costs associated with treating them; nor did it compare the cost of therapy with the value of any benefits to patients.

Despite these limitations, this analysis provides an instructive example. As new oral substitutes for parenteral drugs enter the market, the question will be whether access to appropriate therapy is improved, and whether patients are more likely to complete the recommended duration of therapy at meaningful doses. For ICT, the oral formulation is associated with better patient compliance and adherence than the parenteral drug. This appears to lead to improved delivery of therapy, and hopefully better outcomes—but at substantially greater cost.