 reports the results of 4 studies analyzing novel aspects of the value generated by oncology agents. These 4 studies lay out new approaches to incorporating dynamic and patient-centered considerations into value and cost-effectiveness calculations. These findings may help managed care organizations better understand and quantify the value oncology therapies produce for patients with cancer.This supplement was supported by 

Leonard D. Schaeffer Center for Health Policy and Economics

Norman Topping Chair in Medicine and Public Policy

Quintiles Professor of Pharmaceutical and Regulatory Innovation

Daniel Levin Professor of Public Policy Studies

Articles

AUTHOR INDEX

Participating Faculty: New Approaches to Measuring Value in Oncology Therapy

 provides a list of therapies for chronic myeloid leukemia (CML) and the dates of approval by the US Food and Drug Administration.

The traditional calculation of a given health innovation’s survival benefits assumes no background progress in medical technology during the period in which the user’s life is being extended by the innovation. As a consequence, the survival gains that the innovation offers are measured against the prevailing life expectancy and under the assumption that no further improvements will occur.

However, this assumption may be unrealistic given substantial advances in life expectancy in general and in particular fields of medicine (such as oncology). For example, a patient with cancer who achieves a lower risk of death from cancer will typically face a decreased risk of death from hypertension, diabetes, heart disease, and other conditions over his newly extended life expectancy. These additional benefits multiply the traditionally measured effect and are not considered when valuing the cancer treatment that extended life initially.

In reality, the true value of an innovation includes an individual’s newly gained opportunity to see future medical breakthroughs in other areas, as illustrated by the calculation in 

. We refer to the difference between the true value and the traditional calculation as the option value. In finance, an option holder has the right to buy or sell a security at a contracted price in the future. Part of the value of the option is the expected benefit of having the right to buy or sell the security in the future, should its market price evolve in a favorable direction.

The traditional value is the increased chance of remaining alive, due to the technology. Intuitively, option value is the increased chance of remaining alive, multiplied by the life expectancy gains enjoyed by those who remain alive. This is, therefore, the benefit of gaining the opportunity to live longer. The sum of the traditional value and the option value provide the true value of a technology. Next, we develop the economic framework to value a health innovation and incorporate its option value.

Technical Framework for Calculating the Option Value of an Existing Innovation

The economic framework for incorporating option value into health technology assessment using historical and forecasted changes in survival due to new technologies is based on a competing risk model in which overall mortality is driven by CML and all alternative causes. In this framework, option value derives from 2 sources: (1) treatment may enable patients with CML to live to see new innovations for CML; and (2) treatment may enable patients with CML to live to see new innovations for other causes of morbidity and mortality (eg, heart disease).

 years, will survive 1 more year. This survival probability explicitly depends on person 

 having access to the medical technology available at time 

’s age, other demographics, and health status.

) denote the same probability in the presence of a new medical innovation (eg, a new drug for person 

Let A denote a person’s maximum remaining lifetime at the time of diagnosis (that is, 120 minus his current age).

In the absence of the innovation, we calculate person i’s life expectancy 

, using the standard techniques, as follows:

(Note that we are abstracting away from any discounting for simplicity.)

This difference represents the benefits of future medical progress (in curly brackets) multiplied by the probability of surviving to see those benefits, 

). In other words, the expected lifetime calculation takes into account that survival technologies will improve over time, whereas the standard life expectancy does not.

) denote the (unconditional) survival probability after having the disease a years, and consider a more general innovation which raises each year’s survival probability by some amount, ΔS(

We evaluate the option value of this innovation in terms of its survival benefit (ie, the expected years of life it provides). Let 

} denote the probability of surviving a years after diagnosis, using the life expectancy method (ie, ignoring medical advances). Similarly, let 

} denote the probability of surviving a years after diagnosis, using the expected lifetime method (ie, incorporating medical advances). The option value of the innovation is the difference between the expected lifetime benefits and the life expectancy benefits owing to the innovation. It can be approximated as follows:

Although the specifics of this calculation can be complex, some general points are worth noting. First, the option value will be positive as long as medical technology is advancing. Second, the option value can be expressed as the product of the increased chance of surviving to enjoy the benefits of medical advances multiplied by the survival value of those advances.

The reasoning behind this analytic framework is the following. Suppose the innovation is introduced at time 

. The traditional or direct value of the innovation is the change in survival probability due to the innovation itself. The option value is the increased probability of being alive, multiplied by the value of staying alive in terms of future improvements in medical technology. However, in each time period from time t onward, the option value can be calculated as the product of the change in survival probability due to the innovation multiplied by the change in survival benefits from other medical advances between time 

 and the time period being considered. The total option value of the innovation is simply the sum of these combined survival benefits over the time periods its user may live to see. The total value of the therapy should hence add this option value to the conventionally calculated value.

Estimating Past and Future Improvements in CML Survival and All-Cause Mortality

We quantified past improvements in survival from CML and all-cause mortality by fitting 2 separate Cox proportional hazard models to the Surveillance, Epidemiology, and End Results (SEER) cancer registry data.

 The first model focused exclusively on CML mortality and the second model analyzed mortality from all other causes.

Conceptually, the Cox proportional hazard model assumes that the hazard function takes the form

) is the baseline hazard (for X = 0) and X is the vector of covariates. In this way, the effect of the covariates on the hazard can be estimated without specifying an exact functional form of the baseline hazard 

, the risk of death at any point in time, relative to X

 parameterizes the mortality risk associated with X

 is estimated by maximum likelihood. This estimation allowed us to identify the key effect of innovative CML treatments on CML survival probabilities.

Covariates in each model include year of diagnosis (year 2000 is the reference category), age of diagnosis, age of diagnosis squared, number of other cancers at diagnosis, race (white is the reference category), ethnicity (non-Hispanic is the reference category), sex (female is the reference category), marital status (married is the reference category), calendar month of diagnosis, and county of residence. These patient characteristics affect mortality risk and the use of innovative CML treatments. Hence, failure to control for these observables would confound their effect on CML and all-cause mortality with the survival benefits from first- and second-line treatments. 

 presents the regression estimates of the Cox model applied to CML mortality. As a sensitivity analysis, Table 2 includes different cohorts of patients diagnosed with CML as early as 1995. Patients diagnosed before 1995 were excluded from this sensitivity analysis given their limited comparability to the post-2001 treatment groups in terms of unobservable characteristics (such as frailty, wealth, education, and environmental factors). Overall, these different specifications show quantitatively similar results. The most notable difference is that year coefficients decreased in magnitude as cohorts diagnosed earlier than 2000 are included since these patients received treatment at a more advanced disease stage. Regression model 6 considered patients diagnosed between 2000 and 2008 to measure the benefit/cost of accelerating/delaying treatment by a year. This is our preferred specification to quantify the option value of innovative CML therapies. While CML mortality is significantly decreasing over time, the decline in background (ie, other-cause) mortality is less prominent.

Future improvements in CML and all-cause mortality were forecasted using the Lee-Carter method.5 This method has been widely used as a benchmark for long-term mortality projections, including by the US Census Bureau in creating population forecasts and by the Social Security Administration in forecasting the number of future beneficiaries.6 The Lee-Carter method assumes that the age-specific death rate pertaining to age x at time t can be modeled as

 describes the time-invariant shape of the age-specific death rates, k

 describes the improvement in all-cause mortality over time, b

 describes the tendency of mortality at age 

 to respond to overall improvements in mortality at time 

 is an error term. The Lee-Carter method assumes that the time trend in mortality improvements takes the form of a random walk plus drift:

 and to fit the historical data on the age-specific death rates in the SEER registry over the 2000 to 2008 period. We gauged the sensitivity of the results to our forecast of improvements in allcause mortality by re-estimating option value at the upper and lower bounds of the 95% confidence intervals from the survival forecasts.

Identifying Survival From First- and Second-Line Treatment

We identified survival from first- and second-line treatment algebraically from year-specific survival curves observed between 2000 and 2008. We set up and solved a system of equations linking observed survival in year 

) in that year, and the probability of developing resistance to first-line treatment (

). We set survival in 2000 as the control group’s survival and assumed that patients diagnosed with CML between 2001 and 2005 (ie, before the introduction of second-generation tyrosine kinase inhibitors [TKIs]) faced this survival profile if they developed resistance to first-line treatment. Hence, the observed survival curves between 2001 and 2005 satisfy:

By solving this system of equations for S

, we obtained the survival under first-line treatment for patients who have not developed resistance to it.

As second-line treatments became available in 2006, survival in years 2006 to 2008 can be expressed as:

Unlike the previous scenario, patients who developed resistance to first-line treatment could benefit from secondline treatments and achieve survival profile S

. Patients who do not fail first-line treatment do not obtain survival benefits (and hence, option value) from second-line treatment. We assigned these patients a survival of S

, the most recent survival profile under first-line treatment that we can identify with our data, and solved the system of equations for S

We computed the survival gains from first-line TKI therapy by taking the difference of the 5 survival curves under first-line treatment in years 2001 to 2005 relative to the control group’s survival and aggregating the differences over this 5-year period. Similarly, we estimated the survival gains from second-line TKI therapy by aggregating the difference between the 3 survival curves under second-line treatment in years 2006 to 2008 and the control group’s survival.

We measured the option value of innovative CmL treatments in monetary terms following becker et al

 We had an individual’s lifetime indirect utility function V(S,

 and survival profile S. We supposed that a medical innovation improved survival from S to S’. We defi ned the individual’s willingness to pay (WTP) w as the monetary value that satisfi es V(S,

We assumed that the individual’s utility function in a given time period takes the form

where c is consumption in the period, α is a parameter that normalizes the consumption value of death to zero, and γ is the inter-temporal elasticity of substitution. Given these assumptions, we computed the annual WTP for the medical innovation as

where A(S) is the value of an annuity that pays 1 dollar in perpetuity under survival curve S. The lifetime WTP is the present value of the annual WTP over the lifetime: A(S)w. We assumed that 

ARTICLES

eAppendix -- New Approaches to Measuring Value in Oncology Therapy

Appendix B -- Value of Survival Gains in Chronic Myeloid Leukemia

Methodological Framework: Value of Survival Gains

An economic model developed by Becker et al

 was employed to calculate the social value of survival gains derived from first- and second-line tyrosine kinase inhibitor (TKI) therapy. This economic model calculates patients’ willingness to pay for longevity gains, accounting for discrete (rather than marginal) increases in survival probabilities. Adapting this methodology to the present study, the utility of chronic myeloid leukemia (CML) patients was calculated given the survival curve resulting from TKI treatment and the average income level among CML patients.

Patient income was estimated using the 2005 and 2006 Medical Expenditure Panels Study. Next, the amount of additional annual income necessary for patients to achieve the same level of utility given the counterfactual survival curve in the absence of the estimated TKI effect was calculated. This annual payment was interpreted to be the annual value to a CML patient of an increase in the survival curve associated with TKI treatment.

The present value of annual survival gains was calculated to produce an estimate of the per-patient lifetime value of survival gains derived from TKI treatment. The lifetime value for individual CML patients was summed across all individuals in a CML cohort, and then summed with discounted lifetime values across all present and future cohorts to determine the total social value of survival gains associated with TKI treatment. The estimated social value of TKIs to future cohorts was discounted at 3% per annum, via a method with precedents in the literature.

		Becker G, Murphy K, Philipson T. The  value of life  near its  end and  terminal care. NBER  Working Paper No.  13333. 2007.

		Philipson TJ,  Jena AB. Who  benefits from new medical tech- nologies? estimates of consumer and  producer urpluses for HIV/AIDS drugs. 

Appendix A -- Value of Survival Gains in Chronic Myeloid Leukemia

This study was designed to assess the effect of tyrosine kinase inhibitor (TKI) use on nonpharmaceutical medical spending for patients with chronic myeloid leukemia (CML), and estimate the association between cost-sharing and the TKI medication possession ratio (MPR).

 The retrospective study covered the 13 years from 1997 to 2009.

Analyses were conducted using a large administrative health insurance claims database covering 45 large employers. From this database, 995 unique patients with CML were identified, with 3,765 patient-years; of these patients, 415 (or 1,689 patientyears) were TKI users. We estimated the association of TKI use with total pharmaceutical spending and total non-pharmaceutical medical spending. In addition, we characterized plan-level cost-sharing rules for TKIs and assessed whether these were associated with the MPR for TKI therapy among CML patients.

TKI users averaged $26,406 in annual non-pharmaceutical medical spending, compared with $38,194 for non-users; this was a difference of approximately 30%, which was statistically significant at the 5% level. The median patient out-ofpocket payment was $25, which increased to $63 at the 75th percentile and to $122 at the 95th percentile. MPRs were 94.8 at the median cost-sharing level and 100.0 at the 75th percentile and higher. There was no statistically significant association between cost-sharing and MPR.

Use of TKIs was associated with a 30% reduction in non-pharmaceutical medical spending for CML patients. This difference is approximately equal to 40% of the incremental pharmaceutical cost associated with using TKI therapy. The net annual cost of TKI therapy is roughly $15,000. An informal calculation suggests that this is well within the range of conventional cost-effectiveness thresholds. On balance, coverage of TKIs is relatively generous, with the vast majority of patients exhibiting high levels of adherence to therapy.

Advances in biotechnology have produced many effective new treatments for serious and debilitating diseases. This is evidenced by the growth in use of specialty products and by the number of specialty products in the pipeline, most strikingly in the field of oncology. The number of investigational new drug applications for cancer treatment rose from 925 in 2003 to 1,440 in 2008.

 In addition, the price of oncologics has been increasing over time: the most expensive chemotherapy agent in the early 1990s was paclitaxel, which sold for $4,000 per year, while today, the cost of bevacizumab is 10 times higher.

As utilization of and spending on cancer medications increase, both public and private insurers have adopted a more aggressive approach to managing reimbursement, distribution, and benefit designs. For example, most Medicare Part D plans place medications costing $600 or more per month (primarily cancer agents) on a special tier with higher cost-sharing requirements.

 In 2009, 96% of plans had a specialty tier, with virtually every plan imposing coinsurance rates of 25% to 33%.

While lack of insurance is almost certainly a barrier to accessing specialty cancer medications, it is unclear whether demand for these products is affected by the cost-sharing burden imposed on patients by insurance benefit design. The tradeoff between risk-sharing and appropriate incentives raises the issue of how and to what extent payers should cover these products. Benefit-design decisions regarding cancer agents are particularly challenging given the severity of the illnesses they are designed to treat, the diversity of patient responsiveness to therapy, and widespread off-label use within the field of oncology.

 The principal challenge facing society is balancing the need to ensure access to a wide range of therapeutic alternatives and the need to constrain the rapid growth in healthcare expenditures. In order to do this effectively, more information is needed on clinical efficacy, price sensitivity, and overall value.

In this article, we estimate coverage and demand for a specific class of high-cost cancer therapies; namely, tyrosine kinase inhibitors (TKIs). Imatinib was introduced in 2001 and quickly became the standard of care for patients with newly diagnosed chronic myeloid leukemia (CML).

 Since then, dasatinib and nilotinib have been introduced and are associated with significantly higher and faster rates of complete cytogenetic response, and better long-term, progression-free survival versus imatinib.

 Dasatinib and nilotinib were originally introduced as second-line therapies in 2006 and 2007, respectively; both agents were approved for first-line use in 2010, given their demonstrated superiority to imatinib in this setting.

While these medications can dramatically alter the course of disease progression, they can cost $50,000 to $100,000 per year.

 For this reason, we examined the ways in which these products are covered in employer-sponsored plans, the degree of demand sensitivity to cost-sharing, and the extent to which spending on these products affects use of other healthcare services.

Chronic Myeloid Leukemia and the Role of TKIs

CML is a slowly progressing blood and bone marrow disease that is usually diagnosed during or after middle age, and is rarely diagnosed in children. Prognosis and treatment options depend upon the patient’s age, the stage at diagnosis, the load of blasts in the blood (or bone marrow), and the patient’s general health.

TKIs work by targeting a specific protein that promotes cancer cell growth.

 While it is not possible to eliminate all leukemia cells, treatment can help achieve long-term remission of the disease. The prognosis for patients with CML has improved dramatically over the past decade. Prior to the introduction of TKIs, fewer than 20% of patients survived 10 years after diagnosis. First-line treatment with a TKI agent increased 10-year survival rates to 85% or more.

 Because complete response is not achieved or resistance is developed to first-line treatment in 20% to 30% of patients,

 the availability of multiple products in the class is clinically important.

We assembled a data set of all pharmacy and medical claims from 1997 to 2009 for 45 large, geographically diverse US employers. These data, obtained from a benefits consulting firm, have been used previously to explore the association of pharmacy benefit design with medication use by the chronically ill.

 Each employer offered 1 or more health plans to its active or retired employees and their dependents. Each pharmacy claim included the medication name, dosage, days supplied, date and place of purchase (retail or mail order), patient out-of-pocket (OOP) expense, and health plan payment. Medical claims included information on the service type (eg, inpatient or emergency department [ED]) and date, billed charges, patient OOP expense, health plan payments, and associated diagnosis codes. Demographic variables included age, gender, the first 3 digits of the individual’s zip code of residence, relationship to the primary beneficiary, and employment status of the primary beneficiary.

The study sample included a cohort of adult patients (aged >18 years) with an initial diagnosis of CML during 1997 to 2009. We identified individuals with CML based on the presence of 2 or more

 International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) 

diagnosis codes (primary, secondary, or tertiary) of 205.1x in the medical claims, or 1 

 code and use of TKI therapy after diagnosis. The requirement for a second CML diagnosis or the use of a TKI serves to filter out patients with “rule-out” diagnoses. We excluded patients with less than 1 year of follow- up data and with just 1

 diagnosis for CML, as well as those with no claims for TKI therapy over their entire claims history. Because firms entered and exited the database over time, we did not have a complete panel on all individuals. The majority of members were observed for 2 to 4 years, with fewer than 10% of the sample followed for 7 years.

We defined 2 distinct study samples: 1) those newly diagnosed with CML, and 2) CML patients in whom use of a TKI was initiated. Patients were considered newly diagnosed if they had at least 1 year of data prior to the index date (date of their first 

 code) without a claim for the condition; for example, individuals with 2 

 codes for CML in 2004 would be considered newly diagnosed if they had no other 

 codes for the condition in prior years (eg, 1997-2003) if continuously enrolled. Similarly, we defined the TKI-initiation sample based on the absence of any TKI use in prior years. We focused on newly diagnosed patients because we had full information on the history of their condition and the treatments they had received for it.

The aims of the study were to examine the ways in which TKIs are covered in private health plans, how demand responds to cost-sharing, and whether use of TKIs affects the use of other healthcare services. We measured adherence to TKI treatment as the proportion of days in which therapy was supplied; for example, if a patient filled 9 TKI prescriptions, each providing a 30-day supply, utilization was determined to be 74% (270/365). Not all CML patients who began a TKI regimen remained on therapy; nearly 3% of TKI users stopped therapy in the first year of treatment and more than 5% stopped at some point during the study period. For these patients utilization was measured from the date of initiation to their last day of supply. This approach yields a higher estimate of adherence but minimizes measurement error due to discontinuation of therapy.

Because the data were derived from health claims and enrollment data, they contained a limited set of demographic covariates, including age, gender, marital status, and sponsorship status (employee or dependent type). Measures of health status were derived from the claims directly as a sequence of binary indicators for whether the patient had 2 or more claims for a specific condition during the year; 2 or more claims were required to avoid rule-out diagnoses. One virtue of such large data sets is that many comorbid conditions can be entered simultaneously as control variables. We included the following 30 comorbid conditions in the analyses: essential hypertension, congestive heart failure, diabetes, asthma, hypercholesterolemia, ulcer, depression, chronic obstructive pulmonary disease, allergic rhinitis, migraine, arthritis, chronic sinusitis, anxiety disorder, cardiac disease, vascular disease, epilepsy, gastric acid disorder, glaucoma, gout, hyperlipidemia, irritable bowel syndrome, malignancies other than CME, psychotic illness, thyroid disorder, rheumatoid arthritis, tuberculosis, angina, human immunodeficiency virus, anemia, and stroke.

Estimating the impact of patient cost-sharing requires measurement of benefit generosity at the plan level. Based on prior work

 we characterized the OOP cost of TKI therapy by computing the mean OOP cost within a plan for the weighted basket of TKI medications available in that year, where the weights reflect the market share of TKIs across all plans in the sample year. This allowed for a direct comparison of the OOP costs associated with purchase of these medications across plans. We excluded plans with small numbers of TKI users and adjusted prices to constant dollars using the medical services consumer price index.

Empirical Approach: Estimating the Demand for TKI Therapy

We estimated the price sensitivity of TKI therapy using multivariate regression to correlate the utilization of therapy to the price of drug therapy in the plan. Specifically, we used the following linear probability model for individual 

 is the number of days’ supply of TKIs by patient 

 is the average price per-claim for TKI therapies, which is calculated as the average OOP amount per claim among all persons using a TKI in individual 

• Γ is a vector of sensitivity parameters to individual characteristics

 is a vector of individual characteristics (including age, comorbidities, etc)

 is the error term, which we cluster at the employer level, since the variation in price is at the plan level and plans are nested within employers.

We first estimated the probability of a patient initiating TKI therapy as a function of plan generosity, as well as adherence to TKI therapy, conditional on use. We measured adherence as the number of days supplied (or proportion of days covered, or cumulative medication gap in a time frame), conditional on initiating therapy. This structure imposed the restriction that the price sensitivity of TKI therapy is similar across the different medications. This seemed reasonable given that all 3 TKIs currently on the market are used to treat advanced forms of CML, and imatinib was the sole TKI until 2006. Moreover, pooling the data across TKIs allowed us to obtain more precise estimates than if we had focused on each medication alone—an important consideration given that some of the medications in our sample were used by only a small number of patients. We also estimated the impact of TKI therapy on medical spending and use. Estimates from the multivariate models were used to predict medical and prescription spending for TKI users and TKI non-users, controlling for individual and plan heterogeneity. More specifically, estimates from the models were used to predict annual spending by type of service for each person in the study sample. The models included controls for patient demographics (age, gender, marital status, employment status), comorbid conditions, time since diagnosis, geographic and socioeconomic measures (census region or state, urban residence, median household income in the zip code), and annual time dummies. In all models we adjusted our variance estimates to allow for correlation (or “clustering”) of individuals within plans.

To provide some context for our results, we looked at the impact of cost-sharing on the use of another oral cancer agent. Erlotinib is a kinase inhibitor used to treat non-smallcell lung cancer (NSCLC) and is also used as part of a combination regimen for pancreatic cancer. Erlotinib is widely used in the United States and costs $30,000 to $40,000 for a year of treatment. We looked at the impact of copayments on utilization of erlotinib to provide a point of comparison.

Finally, we examined the sensitivity of our results against more stringent definitions of CML by excluding individuals with CML diagnosis codes who were not taking any medications for CML.

Our sample included 995 patients newly diagnosed with CML (

). Since data were available for these individuals for multiple years after diagnosis, the number of person-years (number of individuals multiplied by number of years in the data post-diagnosis) is considerably larger (N = 3,765). More than 80% (812) of patients in the study sample were diagnosed with CML after the introduction of imatinib in 2001, with 42% (415) having filled at least 1 prescription for a TKI over the study period. The majority of patients were observed for 3 or more years. The demographic characteristics of CML patients who initiated use of a TKI were similar to those of non-users, although non-users were more likely to be working than were users (34.7% vs 29.0%). This suggests that differences in health outcomes between the 2 groups are more likely to result from TKI use than from preexisting demographic differences. Both groups frequently utilized healthcare services, with the average CML patient making 26 outpatient visits per year and accruing total annual expenditures in excess of $50,000. Even so, the distribution of spending varied substantially. TKI users incurred higher total healthcare costs than non-users; specifically, these patients spent considerably more on pharmaceuticals and less on medical services than non-users. Non-users were enrolled in health plans with less generous coverage of TKIs but, overall, average coinsurance rates were low.

Although TKI therapies are expensive, they were generously covered in most of the employer-sponsored plans in our sample (

). A 30-day supply of a TKI costs $4,000 to $8,000, yet the median copayment in our sample was just $25. The least generous plans, with copayments of $100 or more per 30-day supply, had an average coinsurance rate of less than 5%. However, even generous cost-sharing at the plan level can translate into significant OOP expenses for CML patients, given the cost of these treatments. For example, mean annual OOP costs in the sample were $2,779, with $785 attributable to medications. Further, 1 in 4 CML patients had total OOP costs in excess of $7,000.

Given the efficacy of these therapies and the low level of patient cost-sharing, it is not surprising that adherence was high. The average medication possession ratio for TKIs was 85.7%, which suggests that CML patients had the TKI on hand roughly 7 out of 8 days. More importantly, we found no statistically significant relationship between generosity of coverage and use of TKI therapies, measured either as initiation of TKI therapy or adherence conditional on use. CML patients with less generous coverage were more likely to stop therapy. This effect was statistically significant, although quite modest in size. It is notable here that no plans charged copayments in excess of $200, and only a small number charged copays higher than $100. The lack of extreme costsharing is likely to mute the effect of cost-sharing on initiation of treatment.

 shows adjusted estimates of annual spending for TKI users and non-users based on multivariate models that control for individual and plan heterogeneity. Pharmaceutical spending was considerably higher among TKI users. CML patients using TKIs spent $26,452 more on medications than similar patients who did not begin treatment with a TKI (in 2009 dollars). However, average annual medical expenditures were $11,788 lower, offsetting nearly 45% of the incremental costs in medication spending. In sum, the use of TKIs increased total expenditures at a relatively modest cost per life-year and shifted the distribution of healthcare spending. Specifically, as discussed in the second article in this supplement by Yin et al,

 the introduction of TKIs led to a 10-year gain in median survival, from 7.5 years to 17.5 years. The incremental annual cost of TKI use is just under $15,000. Over a 17.5-year horizon, with a 3% real rate of discount, an annual flow of $15,000 in net costs has a net present value of $206,000. Thus, this rough “back-of-theenvelope” calculation suggests that the incremental costeffectiveness ratio is less than $20,600 per life-year gained, which would be considered cost-effective by conventional standards.

Given the generosity of coverage for TKIs and the lack of variation across plans, we looked at the coverage of other oral cancer agents to assess the impact of high patient costsharing on use. As previously noted, the kinase inhibitor erlotinib is used for the treatment of NSCLC and as part of a combination regimen for the treatment of pancreatic cancer and costs $30,000 to $40,000 for a year of treatment. The mean copayment in our sample was $124, with the least generous plans (top decile) charging members $224 or more for a 30-day supply. 

 shows the adjusted impact of copayments on the annual days of supply for erlotinib. For lung cancer patients using erlotinib, annual days supplied was 15% lower for those paying $200 or more, relative to similar patients with copayments less than $100 per 30-day prescription.

Some individuals in our sample were classified as having CML based on 2 or more ICD-9-CM diagnosis codes, but were not taking TKIs or other medication for the condition. We re-estimated the models excluding these patients to assess the extent of this misclassification bias. Although the overall results were substantively unchanged, the magnitude of effect increased in some specifications. For example, 58% of CML patients were taking a TKI, which was more in line with current levels of treatment. In addition, TKI use was associated with even larger reductions in medical spending (38%) when potentially misclassified CML patients were excluded.

Cost-sharing has been viewed as a way to protect against moral hazard and discourage overconsumption of services with little social value. One difficulty, of course, is identifying the services that are cost-saving or cost-effective and then ensuring that cost-sharing policies are sufficiently generous to encourage their use.

We found that use of TKIs was associated with lower spending on other types of healthcare services. CML patients on TKI therapy had roughly $12,000 less in nonpharmaceutical medical costs than did patients on alternative forms of therapy. This translated into a decline of more than 30% in medical spending and offset roughly 40% of the cost of TKIs. This result is consistent with prior work that suggested changing generosity for one healthcare service has both short- and long-term implications for spending in other areas. Chandra et al

 found substantial off-setting increases in hospitalizations when copayments for outpatient and pharmaceutical use were increased. In this case, the cost of hospitalizations was borne by Medicare while the benefits of the reduction in prescription use accrued to the state’s supplemental retiree plan. Similarly, Gaynor et al

 found that while increased copayments for prescription medications reduced prescription utilization and spending, consumers spent more on other outpatient care.

Current TKI coverage policies appear to support the continued diffusion and use of these products. While the medications cost approximately $4,000 per month, the median copayment for a 30-day supply of TKIs was $25, and only a few plans charged members more than $100 per prescription. Further, neither initiation nor use of TKIs was related to cost-sharing. Patient adherence to prescribed medication regimens was high, perhaps reflecting the generosity of insurance coverage and the efficacy of these agents. By contrast, other oral cancer agents are not covered as generously and can serve as cautionary examples of the demand response if cost-sharing increases beyond certain thresholds. Copayments of $200 or more per month for erlotinib, an analogous oral agent used to treat lung cancer, were associated with a 15% reduction in use relative to those with copayments less than $100. Thus far, copay burdens of this magnitude are absent for TKIs. If they were to appear and exert downward pressure on adherence of the sort observed in the case of erlotinib, the evidence suggests that adverse health outcomes would result.

A key challenge in this type of analysis is that disease severity cannot be measured directly, and patients who are more severely ill may use different types of therapies. Ideally, we would want to instrument for TKI use with the generosity of plan coverage, but, as noted earlier, there was limited variation in cost-sharing across plans in our sample. We attempted to minimize bias by selecting a more homogenous cohort of patients newly diagnosed with CML. Although this reduced our sample size by nearly 40%, we had full information on the history of the condition and the treatments received by these patients. Further, we found no significant differences between TKI users and non-users with regard to number of comorbid conditions or the probability of being hospitalized or admitted to the ED during their first year in the database. An additional limitation of our study was that the sample of CML patients, although large for this condition, was not representative of all privately insured patients or CML patients nationally. Lastly, our measure of medication utilization, days supplied, may have overstated actual adherence to TKI therapies, although this measurement error should not vary systematically across health plans.

The continued introduction of effective, high-cost, novel cancer therapeutics and diagnostics is likely to exert increasing financial pressure on patients, oncologists, payers, businesses, and society, while simultaneously improving outcomes and expanding treatment options. Payers will have more latitude with which to vary the benefit designs offered to patients with cancer, and these decisions are likely to have significant health consequences. The solution is a more evidence-based approach to the design of insurance benefits. It will become critical to identify the value to patients of new therapies themselves, and of generous coverage for these therapies. Greater importance will thus be placed on evidence of value, and pressure will grow for payers to connect reimbursement and cost-sharing more directly to this evidence.

TKIs in particular have generated substantial health improvements for patients with CML; for instance, the evidence presented in this study suggests that their use has decreased medical care utilization by more than 30%. Payers appear to have recognized this value insofar as they have insulated TKIs from the more extreme cost-sharing requirements that have been placed on other oncology agents. This decision likely explains why adherence to TKIs is relatively high compared with other oncology agents. Taken together, these facts suggest that TKI coverage represents an important and instructive success story, in which benefit design is connected to value. 

Bristol-Myers Squibb, Plainsboro, NJ (TD, JRM); Leonard D. Schaeffer Center for Health Policy and Economics, and School of Pharmacy, University of Southern California, Los Angeles, CA (GFJ); Leonard D. Schaeffer Center for Health Policy and Economics, Sol Price School of Public Policy, and School of Pharmacy, University of Southern California, Los Angeles, CA (DG); Leonard D. Schaeffer Center for Health Policy and Economics, University of Southern California, Los Angeles, CA (DNL).

 This supplement was supported by Bristol-Myers Squibb.

Dr Darkow and Dr Maclean report employment with and stock ownership in Bristol-Myers Squibb. Dr Goldman and Dr Lakdawalla report consultancy with Bristol-Myers Squibb and partnership in Precision Health Economics. Dr Joyce reports consultancy with Precision Health Economics.

Concept and design (TD, DG, GFJ, DNL, JRM); acquisition of data (DG, GFJ, DNL); analysis and interpretation of data (TD, DG, GFJ, DNL, JRM); drafting of the manuscript (TD, DG, GFJ, DNL, JRM); critical revision of the manuscript for important intellectual content (TD, DG, GFJ, DNL, JRM); statistical analysis (GFJ); obtaining funding (TD, GFJ, JRM); and supervision (JRM).

 Dana Goldman, PhD, Leonard D. Schaeffer Center for Health Policy and Economics, University of Southern California, 3335 S Figueroa St, Unit A, Los Angeles, CA 90089-7273. E-mail: 

		Schickedanz A. Of value: a discussion of cost, communication, and  evidence to improve cancer care. 

Coverage and Use of Cancer Therapies in the Treatment of Chronic Myeloid Leukemia

To develop a methodology for computing cost-effectiveness measures of a drug throughout its life cycle.

 We developed a set of models that measure the long-term cost-effectiveness of 2 oncology drugs, paclitaxel and docetaxel, throughout their life cycles.

 The study combined pricing history of the drugs, US Food and Drug Administration approval dates, drug utilization from Medicare claims, and clinical effectiveness information from phase III studies reported in the scientific literature. These data were used to estimate the incremental cost-effectiveness ratio (ICER) at the time of market entry and by year thereafter. The study population included patients with cancer who were treated with paclitaxel or docetaxel.

 The prices of paclitaxel and docetaxel dropped substantially due to patent expirations, while the number of users increased several fold because of subsequent empirical evidence and approval of new indications that resulted in greater efficacy. The ICER over a 10-year period was approximately 60% of the ICER at product launch for both drugs, and was further decreased when a longer-term perspective was taken.

We demonstrated that the ICER of a drug can decrease substantially over its life cycle. Thus, cost-effectiveness at drug launch might be a poor indicator of the longterm value of the drug. The results of this study are based on the analysis of 2 prominent oncology drugs, paclitaxel and docetaxel. The results may not be generalizable to other drug classes or other oncology drugs for which new indications are less common.

Rapid increases in drug-related expenditures over the last several decades have led payers to focus increased attention on cost containment and the trade-off between the benefits of a drug and the drug’s budget impact. An increasingly common method to estimate this trade-off is to perform cost-effectiveness analysis (CEA).

 CEA provides a summary of value that is considered by payers to help inform their reimbursement policy for a drug.

 The use of CEA by payers also influences drug expenditures and pricing,

 which can subsequently have a significant impact on the pace and arrival of drug innovation.

 One important limitation of the current approach to conducting CEAs is that they are typically performed and attract the most attention at the time of initial drug launch or for individual indications; therefore, such analyses often disregard the overall cost-effectiveness of a single agent in the longer term, which can change dramatically over the life cycle of a drug.

The change in cost-effectiveness over time is especially important in the context of oncology drugs. The initial indication for a new oncology therapy is typically treatment of patients with late-stage disease for whom all approved treatment options have failed. It is challenging to demonstrate a large clinical benefit in patients with refractory disease who may have received multiple lines of therapy and may also have experienced toxicity from prior treatments. Subsequent indications tend to be for earlier lines of therapy, when patients are more robust and a greater clinical response is more likely. These populations also tend to be larger and, therefore, reflect a more common experience of patients treated with the drug. Therefore, CEA analyses focusing on the initial indication might not accurately capture the value of a drug over its life cycle for 2 important reasons. First, the patient populations receiving the drug can change substantially as new empirical evidence of the drug’s effectiveness is published or new indications for the drug are approved. A second and important factor is that the costs of the drug for payers also change significantly over time due to loss of exclusivity and patent expirations, even though many oncology agents, such as paclitaxel, remain important elements of standard of care well after they have lost exclusivity.

The cost-effectiveness of a drug at initial launch, then, might be a poor approximation of the drug’s cost-effectiveness over its life cycle. However, summary measures of a drug’s cost-effectiveness that include the increasing efficacy in subsequent indications and as price declines are rarely presented. To our knowledge, only 1 other article, by Garrison and Veenstra,6 has included computations of the cost-effectiveness of a drug from a dynamic or life-cycle perspective. To do this, the authors simulated the change in cost-effectiveness of trastuzumab from the time of drug launch due to the introduction of a new indication.

In this article we add to the literature by estimating the cost-effectiveness of 2 oncology drugs, paclitaxel and docetaxel, over their life cycles. These 2 drugs were chosen because they have played an important role in improving cancer survival and because we are able to observe changes in prices and utilization of these drugs from product launch to the present. We extended prior research along several dimensions. First, we used data on drug utilization and prices to show how cost-effectiveness changes over a drug’s life cycle from product launch to post—patent expiration. Second, we documented changes in cost-effectiveness due to new clinical evidence of efficacy, US Food and Drug Administration (FDA) approval of new indications, the size of the patient population treated, and changes in drug prices. Finally, we used Medicare claims data to construct a nationally representative sample to depict the ways in which the utilization of these drugs evolved over time by cancer type and year.

We developed the incremental cost-effectiveness ratios (ICERs) of 2 oncology drugs from a life-cycle perspective. These measures describe the net value of products over their life cycles rather than just at the time of product launch. The underlying model framework that we propose builds on the model developed by Garrison and Veenstra.

To formally illustrate the difference between product life cycle and standard CEA measures, we first consider the ICER at product launch (t = 0), with the equation 

 is the incremental cost of the new drug at product launch and E

 is the measure of incremental effectiveness of the drug at product launch.

Next, we consider the ICER T time periods after launch:

 are the incremental cost and effectiveness of the drug T time periods after product launch, 

 is the population of users T periods after product launch. Given the definitions above, the ICER T time periods after product launch is simply the ratio of the present value of incremental costs associated with the drug until period T (numerator) and the present value of the incremental effectiveness associated with the drug until period T. The ICER from a product life-cycle perspective is simply ICER

, where T equals the end of the product life cycle.

Because drugs can last on the market for several decades, we calculate ICER

 for 10 and 20 years after product launch.

 only if the following 2 conditions are satisfied:

1) the incremental costs do not change over the life cycle IC

2) the incremental efficacy does not change over the life cycle E

If either of the above conditions are violated, then ICER

 would be a poor approximation of the true long-term cost-effectiveness of drugs.

As discussed earlier, the first condition is likely violated because drug prices can decrease dramatically over the life cycle, with prices falling to as low as 10% of the product launch price after patent expiration. Notice that the effects of reduction in prices after product launch will be magnified by the increased diffusion of the drug over time; in other words, lower prices mean that a larger population will use the drug. This would increase the weight attached to costs in later years, when prices are low, and would correspondingly decrease the weight attached to costs in earlier years, when prices are high.

The second condition is also likely violated as the efficacy and effectiveness of the drug might change over its life cycle. As mentioned above, this is especially likely for oncology products, which are initially used as second- or third-line treatment for patients with late-stage cancers; after product launch many oncology drugs receive approval for additional indications, such as treatment of early-stage disease and/or other types of cancers. As was the case with changes in price, the effects of changes in effectiveness will be magnified by the increased diffusion of the drug over time; in other words, increased effectiveness translates into greater diffusion of the drug. This would increase the weight attached to effectiveness in periods when efficacy is increased. To compute the ICER from a life cycle perspective we need information on the measures described below.

We estimated incremental costs by indication by using cost-effectiveness data published in peer-reviewed journal articles whenever possible. For some indications approved in the early 1990s, when costeffectiveness data were often not available, we estimated the incremental cost of treatment using information on unit prices and treatment regimens from published data and/or FDA product labels, including dosage and average number of courses.

 To estimate the incremental costs for a given indication change over the life cycle of the drug due to changes in real drug prices, we obtained the wholesale acquisition cost (WAC) of paclitaxel and docetaxel in branded and generic forms (if applicable) from Analy$ourceOnline. These data were obtained from a national drug database, and are included with permission from and copyrighted by First DataBank, Inc. Information on the methodology of collecting price data is described at the First DataBank website: 

http://www.firstdatabank.com/Support/drug-pricing-policy.aspx

. When multiple prices were posted for a single year, the lowest WAC available for that year was used. All costs and prices shown in this article are inflation-adjusted to 2010 US dollars, assuming an annual inflation rate of 3%.

 For each drug we constructed nationally representative utilization rates by cancer type or indication from 1991 to 2007. The utilization data were derived from the surveillance, epidemiology, and end results (SEER) Medicare data and include the number of Medicare patients using each drug by major cancer categories, including breast, lung, prostate, and colorectal cancer. Because no utilization data are available for the time period after 2007, we assumed that utilization rates stayed at the 2007 levels for future years. The data were weighted to be nationally representative using information on national estimates of prevalence of different types of cancer by year. In particular, we assumed that the utilization rate of different drugs by cancer type is the same for Medicare and non-Medicare beneficiaries; therefore, national estimates of prevalence of different cancers multiplied by the utilization rates of different drugs by cancer type from SEER Medicare provide an estimate of the national utilization of drugs by cancer type.

 We also quantified life-cycle changes in efficacy. For each drug we identified the approval of new indications over the product life cycle from the FDA product labels, and the life-years gained for each new indication from peer-reviewed journal articles. We omitted certain indications (such as paclitaxel for second-line treatment of ovarian cancer and Kaposi’s sarcoma, and docetaxel for first-line treatment of non-small-cell lung cancer and gastric cancer), as no reliable data on cost-effectiveness were available or could be constructed for these indications. We combined these data with claims-based measures of the indication-specific utilization described above, in order to produce timevarying estimates of mean life-years gained. These estimates change with the introduction of new indications, as well as with shifts in the proportion of patients treated for newer or older indications over time.

Imputations in the Absence of Price and Utilization Data

In order to illustrate the long-run perspective, we calculated the ICERs of each drug for 20+ years since product launch. For paclitaxel we covered the years 1994 to 2016. For the 5 years beyond 2011, for which actual price data were not available, we assumed the real price of paclitaxel stayed constant at the 2011 level until 2016. For docetaxel, we covered the time period of 1998 to 2020. Because the docetaxel patent expired recently (in 2010), and we have not had a chance to observe price drops over a long period of time after the loss of exclusivity, as in the case of paclitaxel, we assumed its price would decrease at the same rate as that of paclitaxel after its loss of exclusivity, and imputed the annual prices of docetaxel between 2012 and 2020 using a linear prediction model. The data on utilization rates by indication were not available beyond the year 2007, and we assumed that the rates stayed at the same level of year 2007 for future years in our model.

The credibility of the results depends upon their robustness to alternate assumptions about the life cycle trajectory of each of the key model inputs described above. We therefore tested the robustness of our results with multiple 1-way sensitivity analyses, by varying the values of the incremental cost and life-years gained to plus or minus 25%.

New Indications and New Data on Cost-Effectiveness

Several new indications were approved for both paclitaxel and docetaxel within 10 years of product launch (

). In general, the effectiveness of the 2 drugs for new indications was greater than the effectiveness of their original indications at product launch. Two exceptions were a new indication for paclitaxel for the treatment of lung cancer and a new indication for docetaxel for the treatment of metastatic prostate cancer; however, both conditions are associated with a poor prognosis. Table 1 also shows that the empirical evidence on effectiveness in peer-reviewed journal articles usually predates its official FDA indication approvals by a few years.

Fifteen Years of Pricing History Since Product Launch

 shows the trends in real prices for paclitaxel and docetaxel in 2010 US dollars. The prices are rescaled so that the initial price in the year of product launch equals 100; this allows for easy comparisons of product prices over the life cycle relative to price at product launch. Figure 1 demonstrates that the price of paclitaxel decreased after product launch, with real prices being approximately 30% lower 6 years after product launch. There was also a sharp decline in prices after loss of exclusivity due to patent expirations; the price of paclitaxel 12 years after product launch was only 10% of its price at product launch. In contrast, the price of docetaxel increased gradually after product launch, rising to approximately 17% higher 6 years after product launch and 24% higher 12 years after product launch. Subsequently, the price of docetaxel declined by approximately 25% in a single year after it lost exclusivity in 2011 due to patent expirations.

Utilization Rates and New Evidence of Efficacy

 shows the time trends in the utilization of paclitaxel and docetaxel by cancer type, demonstrating that the utilization of both drugs changed in response to publication of clinical evidence of new indications and regulatory approval of new indications. Figure 2 also shows that for both paclitaxel and docetaxel, increased use for an indication began quickly after clinical data showed effectiveness, usually some time before the official approval of the new indication. In 2007 there were 8 times as many paclitaxel users as in 1994, when the first cost-effectiveness data on paclitaxel became available. Also in 2007 there were 9 times as many docetaxel users as in 1998, when docetaxel was first launched. Thus, the most common use of the products is generally for the approved indications for which efficacy is greatest. This use is much more widespread after loss of exclusivity than at the time of product launch.

The ICERs of paclitaxel and docetaxel over 22 years are shown in 

. The ICERs of both drugs decrease as one takes a longer-term perspective; for example, the ICER at year 0 or at product launch was $30,723 per life-year gained for paclitaxel and $37,537 per life-year gained for docetaxel. The ICER over a 10-year time period is only a fraction of the ICER at product launch. In particular, the ICER of paclitaxel over a 10-year time period was 61% of its ICER at product launch and, similarly, the ICER of docetaxel over a 10-year time period was 60% of its ICER at product launch. The ICERs of both drugs decreased even further when viewed with a 15- and 20-year perspective; specifically, the ICER of paclitaxel over a 20-year period was $8,563 per life-year gained (28% of the ICER at launch) and the ICER of docetaxel over a 20-year period was $15,468 per life-year gained (41% of the launch ICER). The 1-way sensitivity analyses produced a range of ICER values but do not change the main qualitative results of the study (

The ongoing debate over rising healthcare spending has led to additional scrutiny of the value of healthcare interventions, including pharmaceuticals. This movement has also resulted in keen interest in CEAs, which have been adopted as a standard for reimbursement approval in countries such as the United Kingdom. Although CEA is often conducted at product launch or for specific indications in isolation, and is an important component of dossiers submitted to public and private payers,

 the initial CEA or CEAs for new indications might not reflect the true long-term value of a particular drug. For this reason, the initial CEA needs to be placed into context against a summary measure of CEAs that reflects long-term value.

New Approaches to Measuring Value in Oncology Therapy

Participating Faculty: New Approaches to Measuring Value in Oncology Therapy

eAppendix -- New Approaches to Measuring Value in Oncology Therapy

Appendix B -- Value of Survival Gains in Chronic Myeloid Leukemia

Appendix A -- Value of Survival Gains in Chronic Myeloid Leukemia

Coverage and Use of Cancer Therapies in the Treatment of Chronic Myeloid Leukemia

Dynamic Cost-Effectiveness of Oncology Drugs

The Option Value of Innovative Treatments in the Context of Chronic Myeloid Leukemia

Value of Survival Gains in Chronic Myeloid Leukemia