Out-of-Pocket Spending Not Associated with Oral Oncolytic Survival Benefit

BACKGROUND: With total and out-of-pocket spending for oral oncolytics rising, there is increased interest in choosing oncology treatments based on their clinical value relative to cost. OBJECTIVE: To determine if out-of-pocket spending varied for higher versus lower benefit oral oncology drugs reimbursed by commercial insurers. METHODS: This study was a retrospective analysis of commercial insurer prescription drug claims filed between 2007 and 2014 for 13 oral oncolytics approved before 2009. We calculated mean monthly out-of-pocket payment for each fill by patient. We then categorized oral oncolytics by their overall and progression-free survival benefits for each FDA-approved indication, using evidence from published studies. We assessed the relationship of survival benefit with mean monthly out-of-pocket payment, adjusting for demographic and plan characteristics. RESULTS: Our population included 44,113 patients aged 18-65 years (mean 52.5 [SD 9.4]) with a cancer diagnosis who filled 731,354 prescriptions. The most commonly represented oncolytics were imatinib (37.4% of fills), lenalidomide (17.7% of fills), and dasatinib (10.0% of fills). Approximately 32.3% of fills were for drug-indication pairs with an overall survival benefit of 4+ years. In adjusted analyses, there was no clear pattern to suggest that out-of-pocket payments differed with drug indication-specific survival benefits. Drugs for indications providing > 0 to 1 year of overall survival benefit were significantly more likely to have a lower out-of-pocket payment versus those prescribed off-label, but there were no significant differences in out-of-pocket payments between drugs and associated indications in any other survival category versus drugs used off-label. CONCLUSIONS: Out-of-pocket payments for oral oncolytics were not clearly related to indication-specific value in commercially insured patients. This finding suggests that despite increased attention to value- and indication-based drug pricing, cost sharing for oral oncolytics does not currently reflect these goals.

C ancer drug costs continue to rise in the United States, 1 and spending on targeted oral oncolytics has increased particularly rapidly, with mean monthly expenditures in the year of product launch increasing from $1,869 in 2000 to $11,325 in 2014. 2 Rising drug prices and expanded use of cost sharing have translated to higher out-of-pocket spending for targeted oral oncolytics. 2,3 For example, 1 study found that average out-of-pocket monthly expenditures rose from $88 in 2001 to $198 in 2011. 4 Concurrently, there is increasing interest in choosing treatments based on their value-or health outcomes relative to their cost 5 -with definitions of outcomes in oncology integrating clinical benefits, toxicities, and patient preferences. 6,7 Evidence of the relationship between drug prices and outcomes in oncology is mixed but largely suggests that overall price is not correlated with benefit. [8][9][10] With regards to prices that patients see, there is a lack of evidence that insurers have employed cost sharing to preferentially encourage use of higher-versus lowerefficacy oncology drugs. Institution of lower cost-sharing levels for interventions with greater evidence of efficacy is known as value-based insurance design and has previously been shown to increase adherence to medications and improve clinical outcomes. 11 Given the known inverse relationship between out-ofpocket spending and adherence to or continuation of oncology drugs, 12,13 use of value-based cost sharing could potentially drive patient adherence to high-efficacy oral oncolytics and thus enhance outcomes.
Within this context, we sought to describe out-of-pocket spending by patients with commercial insurance who had a diagnosis of cancer during 2007-2014 and were treated with oral cancer drugs approved before 2009. We specifically assessed if patients using drugs with higher versus lower benefit had lower out-of-pocket spending, with benefit defined in terms of overall or progression-free survival as per the American Society of Clinical Oncology's (ASCO) 2016 Value Framework. 6 • Spending on cancer drugs, particularly oral oncolytics, continues to rise in the United States, with 1 study showing an average monthly out-of-pocket spending increase from $88 in 2001 to $198 in 2011. • There is interest in choosing treatments based on their value, and institution of lower-cost sharing levels for interventions with greater efficacy has been shown to increase adherence and improve clinical outcomes.

What is already known about this subject
• Out-of-pocket payments for oral oncolytics from 2007-2014 were not clearly related to indication-specific value. • Despite increasing interest in value-based pricing, out-of-pocket payment design is not congruent with the goal of aligning costs with health outcomes. survival), we used estimates from cost-effectiveness analyses and, where appropriate, focused on the survival measure for which a significant effect had been shown in randomized controlled trials. Drugs and associated indications were partitioned into the following 8 categories of increasing evidence for benefit, based on the relative distribution of survival estimates for all fills: (1) off-label use (i.e., prescribed to a patient with a diagnosis that was not an FDA-approved indication for that drug); (2) no survival benefit for an on-label indication; (3) > 0-5 months progression-free survival benefit; (4) > 5-10 months progression-free survival benefit; (5) 10+ months progression-free survival benefit; (6) > 0-12 months overall survival benefit; (7) > 12-48 months overall survival benefit; and (8) 48+ months overall survival benefit. If a drug was associated with statistically significant overall survival and progression-free survival benefits, it was only put into an overall survival category, given the clinical superiority of overall survival benefits over progression-free survival benefits. The distribution of each drug and its associated assignment into these categories, as well as the study from which information about benefit was derived, is described in Table 1. Each of these categories was assigned a sequential numeric code, which was applied individually to each fill for each patient based on the drug and associated cancer diagnosis for that fill.
As an example of this process, pazopanib is approved for use in soft tissue sarcoma. In a 2012 study by van der Graaf et al., 17 use of the drug was associated with a progression-free survival benefit of 4.6 months, while use of placebo was associated with a progression-free survival benefit of 1.6 months. We thus calculated the relative progression-free survival benefit as 3 months. This placed pazopanib for an indication of soft-tissue sarcoma in category number 3 for > 0-5 months of progressionfree survival benefit.
In the case of imatinib, since median overall and progression-free survival was not reached even in trials with longerterm follow-up, 18 we used estimates of these metrics from economic modeling studies based on randomized controlled trial data. 19 Given that randomized controlled trials had demonstrated an overall survival benefit of imatinib versus interferon alpha, we defined the relative survival benefit of imatinib as the difference between the 19.1 remaining life-years associated with first-line imatinib versus the 9.1 life-years associated with first-line interferon alpha. 19 Covariates. For each patient in each year, we documented age, health plan type, and census region, as reported in the MarketScan database. We also calculated previous year comorbidity based on the Klabunde modification of the Charlson comorbidity score, using diagnoses reported in MarketScan for each patient. 20 We also measured total monthly drug reimbursement (out-of-pocket and plan payments for each fill

■■ Methods Data and Patients
We used the MarketScan Commercial Claims and Encounters Database for 2007-2014 to examine drug use and expenditures among cancer patients who had filled any orally administered, targeted nonhormonal anticancer medications approved before 2009. 4,14 We restricted our analysis to the 13 drugs approved before 2009 to ensure availability of 5 years of data for analysis, thus, allowing for adjustment for changes in spending over time. The MarketScan database represents > 50 million employees and their dependents enrolled in commercial health insurance plans sponsored by approximately 100 large and medium-sized U.S.-based employers. About 56% of the nonelderly U.S. population was covered by employer-sponsored coverage in 2014, making the MarketScan database largely representative of our population of interest. 15 We identified 731,354 prescription drug claims filled by 44,113 patients with a cancer diagnosis in the year of the fill. We excluded 5,063 claims with patient or plan payment < $0 or days supplied < 0.

Measures
Out-of-Pocket Expenditures. We summed monthly copayments, coinsurance, and deductibles for each fill. We adjusted fills for > 1-month supply to a single month supply to generate mean monthly total out-of-pocket payments. We did not consider employee contributions to commercial health insurance coverage via monthly paycheck deductions because we sought to estimate the cost to patients of an oral oncolytic at the time of prescription fill. Patient cost sharing at the point of service is a formulary tool that may influence a patient's use of a prescription drug. 16 Clinical Benefit of Oral Oncolytics. We categorized drugs by overall survival benefit (defined as time from randomization in a trial to death) and progression-free survival benefit (defined as time from randomization in a trial to disease progression or death; used if overall survival data were lacking) for each indication approved by the U.S. Food and Drug Administration (FDA) for the particular oncolytic (as described in the product label for that oncolytic). We examined evidence from randomized controlled clinical trials that compared the drug against standard of care or placebo for each particular FDA-approved indication. We reported overall or progression-free survival benefit as a relative survival benefit (i.e., overall survival or progression-free survival with drug minus overall or progression-free survival with standard of care or placebo, depending on what was reported in the included trial). If multiple similar trials were available for a drug and indication, we selected the trial presenting the largest estimate of survival differences between patients receiving treatment versus control. If median overall survival or progression-free survival were not reached in published trials (most commonly due to high overall  adjusted to a per-month value) to account for differences in drug prices. Variables are categorized in Table 2.

Analysis
After describing out-of-pocket expenditures by drug benefit and patient characteristics, we estimated a 2-part model, 21 predicting changes in out-of-pocket payments by drug survival benefit and adjusting for the previously described covariates. The first part of the model consisted of a logit function that assessed the association of drug survival benefit with having any out-ofpocket spending (vs. none). The second part used a generalized linear model with a gamma distribution and a log-link function to assess the association of drug survival benefit with level of non-zero out-of-pocket spending. Estimates from these 2 models were combined using a margins function. 21 In sensitivity analyses, we conducted stratified analyses, first, considering overall survival as a continuous variable among patients treated with drugs with evidence for an overall survival benefit and, second, considering progression-free survival as a continuous variable among patients treated with drugs, with evidence for only progression-free survival benefit. Finally, in a sensitivity analysis, we conducted a version of the original 2-part model excluding the top 5% of monthly outof-pocket spending claims. All dollars were inflation adjusted to 2014 U.S. dollars based on the medical component of the Consumer Price Index.
Because pricing decisions are often made at launch when only 1 indication for a drug is known, it is possible that cost sharing may be more closely related to the first indication's benefit rather than that of subsequent indications. 9 Accordingly, in a secondary analysis, we reassigned the survival benefit for each drug and all its associated indications based on the benefits associated with the drug's first approved indication. Finally, we repeated analyses within subgroups of drugs or indications, including among tyrosine kinase inhibitors only, among drugs for treating renal cell carcinoma, and for a single drug (i.e., lenalidomide) when considering all of its indications.
Two-sided P values < 0.05 were considered statistically significant. Data were analyzed using SAS version 9.4 (SAS Institute, Cary, NC). This study was deemed not human subjects research by the Harvard University Institutional Review Board.

Indications, Survival Benefit Categories, and Survival Comparisons for Targeted Oral Oncolytics Approved, 1995-2009 (continued)
and associated indications with > 0-5 months of progressionfree survival benefit; 9.7% for drugs and associated indications with > 5-10 months of progression-free survival benefit; and 5.4% for those drugs and associated indications with 10+ months of progression-free survival benefit. Finally, 16.8% of fills were for drugs and associated indications with > 0-1 year of overall survival benefit; 10.1% for those with > 1-4 years of overall survival benefit; and 32.3% for those with more than 4 years of overall survival benefit. Monthly per-fill out-of-pocket payments ranged from $0 to $14,167.30, with a mean of $82.82 (SD $353.45). The mean monthly total (plan plus patient) was $5,884 (SD $2,901.46). Table 3 depicts numbers of patient-fills and mean out-of-pocket payment by drug and indication. In adjusted analyses, there was no clear pattern to suggest that out-of-pocket payments differed with drug indicationspecific survival benefits (Table 4). While drugs for indications providing > 0-1 year of overall survival benefit were significantly more likely to have a lower copayment versus those prescribed off-label, there were no significant differences in out-of-pocket payments between drugs and associated indications in any other survival category versus drugs used off-label. Having a more recent year of fill was associated with lower out-of-pocket spending (P < 0.001). In addition, fills for drugs with higher total costs (plan plus patient payment) were associated with higher out-of-pocket spending (P < 0.001; Table 4). Notably, the correlation coefficient for total drug cost and survival benefit was 0.06, suggesting that total drug prices were not associated with benefit).
Results of sensitivity analyses including only (a) patients filling drugs with evidence of an overall survival benefit or (b) patients filling drugs with a progression-free survival benefit and examining overall or progression-free survival as continuous variables similarly revealed no significant relationship between out-of-pocket payment and survival benefit (change in mean monthly out-of-pocket spending for each additional month of progression-free survival = 0.11, 95% confidence interval [CI] -0.42-0.63, P = 0.69; change for each additional month of overall survival = 0.01, 95% CI = -0.04-0.05, P = 0.77).
Results of sensitivity analyses excluding payments in the top 5% yielded results similar to those in our primary model (not shown). Our findings also held within subgroups of drugs or indications such as among only tyrosine kinase inhibitors, only drugs for renal cell carcinoma, and only for lenalidomide when considering all of its indications (data not shown; available from authors).
Finally, analyses examining survival benefit based on the benefit reported for the drug's first approved indication also showed no association of out-of-pocket payments with benefit (Table 5).

■■ Discussion
Despite consistently rising out-of-pocket spending for oral oncolytics and increasing interest in value-based pricing, 4,6,7 we found that out-of-pocket payments for oral oncolytics from 2007-2014 were not clearly related to indication-specific survival benefit in a commercial population. Although drugs with > 0-1 year of overall survival benefit were associated with significantly lower out-of-pocket spending, there was no difference in out-of-pocket payments for drugs in any other survival category versus drugs used off-label. In addition, while previous analyses have examined drug pricing in relation to first indication survival benefit, 9 our analysis did not show any relationship between the survival benefit of the first approved indication and out-of-pocket payment.
Cancer therapy is associated with significant financial toxicity, including bankruptcy, 22 and there is evidence that beneficiaries with higher-out-of-pocket costs are more likely to discontinue cancer therapy. 12,13 Notably, substantial financial toxicity of cancer therapy has been associated with worse outcomes, including early mortality. 23 Value-based insurance designs have been shown to improve adherence in nononcology clinical areas, [24][25][26] and ideally, cost sharing could be aligned to encourage greater use of and adherence to high-value cancer treatments. However, we saw no association between

Demographic Characteristics of Patient-Fills in Sample
for fills in more recent calendar years. As expected, we observed higher out-of-pocket payments for more expensive drugs.

Limitations
Our study has limitations that need to be considered. Given the composition of the MarketScan database, we studied only commercial health insurance claims for individuals with employer-sponsored insurance; out-of-pocket payments and their relation to drug benefit may differ for other payers, including individuals enrolled in Medicare Part D. Although the MarketScan database provides some information about employee health plan types, it does not provide more granular data regarding copay, coinsurance, or drug coverage arrangements for individual plans, which limited our ability to adjust for more granular differences in coverage. The database also out-of-pocket payments and drug survival benefit. This finding suggests that despite some payers and pharmacy benefit managers announcing or considering value-based drug pricing, 7,27 the complex U.S. drug reimbursement and distribution system still does not lend itself to cost sharing that reflects value. 7,28 Previous work has demonstrated that out-of-pocket spending increased substantially between 2001 and 2011 for oral oncolytics and that Medicare enrollees are more likely to discontinue more expensive cancer drugs. 4,12 In addition, several payers have publicly cited interest in value-based pricing for oncology drugs. 27,29 However, the fact that a relationship does not exist between value as defined by survival and out-of-pocket costs has not been previously demonstrated in the oncology realm. Possibly reflecting the increasing prevalence of oral chemotherapy parity laws, 30,31 we observed lower out-of-pocket payments

Drug
Year Indication

Patient-Fills and Out-of-Pocket Payment Statistics by Drug and Indication for Targeted Oral
Oncolytics Approved, 1995Approved, -2009 This finding highlights opportunities for commercial health insurers to address the complexities of implementing valuebased cost sharing for oncolytics in a health care landscape that will increasingly emphasize benefit reaped for dollars spent.
does not provide information regarding employee contributions to health insurance coverage, such as paycheck deductions. We were also limited to evaluating filled prescriptions, which by nature of patient-pharmacy interactions and the composition of the MarketScan database means those for which claims were paid in full. It is possible that patients prescribed lower value treatments were less likely to fill them if faced with high out-of-pocket costs. Prescriptions not filled were excluded from our analysis, thus, potentially excluding a segment of our population of interest facing significant costs.
Our analysis included only oral oncolytics approved through 2009, so findings could differ for more recently approved drugs, particularly given the recent focus on the value of care delivered in the context of health care reform. In addition, we described value in terms of survival benefit. This metric has been designated as a cornerstone of defining value in ASCO's 2016 Value Framework. 6 However, there are certainly other aspects that contribute to a therapy's clinical value, including safety, tolerability, and ease of use. Further analyses should examine the relationship between these additional measures and out-of-pocket costs.
Finally, since information about cancer stage is not available in the MarketScan database, it was assumed that any drug used for a particular FDA-approved indication was used for the correct stage of disease based on oncologist expertise. While this was largely true, there may have been exceptions not captured in our analysis.

■■ Conclusions
Our study demonstrated that for commercially insured patients, out-of-pocket payments were not associated with indicationspecific survival for oral oncolytics approved before 2009.