Changes in Health Care Costs, Survival, and Time Toxicity in the Era of Immunotherapy and Targeted Systemic Therapy for Melanoma

Key Points Question How have health care costs, survival, and time toxicity changed after the adoption of adjuvant and palliative immunotherapies and targeted therapies for melanoma? Findings This cohort study, which evaluated matched cohorts of 731 patients with melanoma, found a substantial increase in systemic therapy costs in 2018 to 2019 compared with 2007 to 2012. Survival improved for all stages in 2018 to 2019 compared with 2007 to 2012, and time toxicity was similar between eras. Meaning These data highlight the trade-off with new effective therapies, for which there are greater health care costs and time toxicity but an associated improvement in patient survival.


H
ealth care costs for cancer treatment are escalating, with approximately $173 billion in the US and more than $7 billion in Canada during 2020 alone. 1,24][5] In recent years, checkpoint inhibitors (programmed death 1 [PD-1] and cytotoxic Tlymphocyte-associated protein 4 [CTLA-4] inhibitors) and targeted therapies (BRAF and MEK inhibitors) have become the standard of care for patients with locally advanced and metastatic melanoma. 6,7Ipilimumab, a CTLA-4 inhibitor, was first approved in Canada in 2012 for the treatment of advanced melanoma based on randomized clinical trials showing improvement in overall survival (OS). 8,9][16][17] Surgical treatment for melanoma has also evolved in recent years, including the omission of completion lymph node dissections (CLNDs) for patients with sentinel lymph nodepositive disease.9][20] As such, consensus guidelines no longer recommend CLND in all patients with melanoma with sentinel lymph node-positive disease. 6,21However, follow-up continues to involve oncologist or specialist visits and dermatologic follow-up and monitoring.
Novel cancer treatments are often accompanied by burdensome health care encounters, which can eat into the increased survival associated with that treatment.This concept is now known as time toxicity in reference to the amount of time spent in physical health care system contact, such as outpatient visits for bloodwork, imaging, procedures, and consultation; emergency department visits; and facility stays. 22valuating these time burdens associated with cancer care is critical to fully understand the patient and care partner experience and contextualize possible survival gains associated with new melanoma treatments. 23he purpose of this study was to describe the changes in health care costs, time toxicity, and survival associated with the initial treatment of melanoma using population-level data, specifically administrative data from Ontario, Canada.A valuebased approach evaluates changes across multiple health outcomes achieved per dollar spent that are important to patients and care partners. 24We applied this method to outcomes of time toxicity and survival relative to health care costs.Because we were measuring the impact of numerous simultaneous changes in practice, a cost-effectiveness analysis was not appropriate.However, a patient-centered, valuebased approach provides a means to measure the overall impact of these changes on cost and across multiple outcomes relevant to payers and patients.We hypothesized there would be a significant increase in health care utilization and, hence, in time toxicity and systemic therapy costs for patients with advanced melanoma, particularly those with stages III and IV disease, because of the use of checkpoint inhibitors and targeted systemic therapies and a decrease in frequency of completion lymphadenectomy but with improved OS.

Study Population and Design
We performed a retrospective cohort study of adult patients diagnosed with invasive melanoma in Ontario, Canada, from administrative data collected by the Ministry of Health.During this study period, race-based data were not available because of government privacy legislation.The primary exposure was treatment era before the COVID-19 pandemic; we compared a distinct era during which new targeted and immune-based therapies had been adopted for melanoma (2018-2019) with an era before this adoption (2007-2012).The intervening time (2013-2017) represented a transition phase with a mixture of old and new practices, challenging the interpretation of a cost-consequence study.
Ontario is the largest Canadian province, with a population of 15.5 million, and provides single-payer, universal health care coverage for adult residents through the Ministry of Health. 25More than 99.9% of the Ontario population are considered eligible for the Ontario Health Insurance Plan.Costing was performed from the perspective of costs to the provincial government.We identified patients aged 20 years or older who were diagnosed with cutaneous melanoma based on International Classification of Diseases for Oncology, Third Edition histology (872-878, 8790) and site (C44) codes from the Ontario Cancer Registry from January 1, 2018, to March 31,  2019.During this period, adjuvant and palliative targeted and immune-based therapies were approved.The American Joint Committee on Cancer (AJCC) 7th and 8th edition staging was used.Patients with missing stage information were excluded.Stage I was excluded because increased overdiagnosis of low-risk melanoma could result in inflated survival differences between eras. 26A historical comparison cohort was identified from a population-based sample of invasive melanoma cases diagnosed from the Ontario Cancer Registry from January 1, 2007, to December 31, 2012.This interval was used because it preceded Health Canada's approval of checkpoint inhibitors and targeted therapies for melanoma.This study cohort and relevant findings were previously published. 27The current study was approved by the Queen's University Health Sciences and Affiliated Teaching Hospitals Research Ethics Board and followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) and Reporting of Studies Conducted Using Observational Routinely-Collected Data (RECORD) reporting guidelines.A waiver of informed consent was granted based on the Personal Health Information Protection Act, Section 44 (1).Flow diagrams for the current (2018-2019) and historical cohort (2007-2012) are described in eFigures 1 to 3 in Supplement 1.

Data Sources and Linkage
Data were obtained from administrative data sets housed at ICES, an independent, nonprofit research institute funded by an annual grant from the Ontario Government.Cancerspecific data were abstracted from the Ontario Cancer Registry, a population-based tumor registry administered by Ontario Health.Health care utilization data were abstracted from multiple administrative databases described in the eAppendix in Supplement 1.These data sets were linked using unique encoded identifiers and analyzed at ICES.

Covariates
Socioeconomic status was based on community-specific or neighborhood household income quintiles.Place of residence was defined by the 14 Local Health Integration Networks, which are past geographic health care partitions in Ontario.Rurality of residence was classified using the 2008 Rurality Index for Ontario (RIO).Higher scores represent a greater degree of rurality categorized as urban (RIO < 10), suburban (RIO = 10-39), or rural (RIO ≥ 40). 28Comorbidities were measured by using the Elixhauser comorbidity index derived from hospital records with a 5-year lookback from melanoma diagnosis. 29

Outcomes
Systemic therapy was identified from the Activity Level Reporting, New Drug Funding Program, and Ontario Drug Benefit databases.Individuals who received a minimum of 1 dose of either oral or intravenous systemic treatment were counted as having received systemic therapy.Radiotherapy and its intent were based on the Activity Level Reporting database.Melanoma-specific primary and nodal operations were identified from physician billing claims data.Palliative intent or metastasis operations were based on surgical intervention codes from the Discharge Abstract Database (brain, lung or liver tumors, or spinal cord compression).
Health care costs were estimated using an established costing algorithm at ICES in which person-level costs are allocated for the various health care utilizations over time. 30erson-level direct pharmaceutical costs for publicly funded systemic therapy administered in the Activity Level Reporting, New Drug Funding Program, and Ontario Drug Benefit databases were estimated based on patient-level utilization with a well-described algorithm developed at ICES.Although pri-vate drug plans cover limited oral drug costs for some patients younger than 65 years (eAppendix in Supplement 1), for generalizability and interpretability, costs of approved oral drugs were considered to be borne completely by the public payer; costs incurred by private insurance and privately funded clinical trials were otherwise not included. 31,32Oral and intravenous drug costs were included (eg, BRAF/MEK inhibitors, immunotherapy, and chemotherapy).Costs were adjusted for inflation to 2019 Canadian dollars using health care components of the Consumer Price Index.Health care costs were limited to the first year after diagnosis to ensure that stagespecific health care utilization reflected primary treatment and to prevent the influence of health care system disruptions from the COVID-19 pandemic lockdown and melanoma recurrence on results.
Overall survival was measured from the date of melanoma diagnosis to death or last follow-up.Vital status data were censored 3.5 years from diagnosis, with the latest dates of follow-up being June 30, 2015, and September 30, 2022, for the 2007 to 2012 and 2018 to 2019 cohorts, respectively.Survival was measured for a longer period than cost given the increasing absolute benefit of initial treatment over time observed in trials. 8,13ime toxicity was defined as in-person days with health care-related visits for any reason. 22This composite measure included institution-based visits (eg, hospitalizations, day operations, emergency department-only visits, and longterm care stays) and outpatient care visits (eg, cancer clinics or ambulatory interventions, primary care, and specialist office visits).We reported composite time-toxic days in the 1-year period after the first day of treatment of melanoma-specific systemic therapy, radiotherapy, and/or surgery.Home care services, virtual or telephone consultations, and physician home visits were excluded because these services did not involve time away from home.We included these encounters in a sensitivity analysis.

Statistical Analysis
Data analysis was performed from October 17, 2022, to March 13, 2023.Propensity scores (PSs) were estimated using a logistic regression model with demographic covariates, the Elixhauser comorbidity index score, and melanoma characteristics.The PS was used to balance differences in patient characteristics between eras when comparing costs and outcomes regarding different treatment indications.Patients with melanoma diagnosed from 2007 to 2012 were matched 1:1 to patients with melanoma diagnosed from 2018 to 2019 using a greedy algorithm with calipers of 0.2 of the SD of the logit of the PS. 33The distributions of the PS matched characteristics between cohorts were evaluated using standardized differences, with a value of 0.10 or less suggesting adequate balance.
Figure 2 depicts time toxicity and quantifies specific health care services within the first year after the first cancer treatment for the matched cohort.As shown in Figure 2A, few differences in time toxicity were observed between eras.Stratified by stage in Figure 2B-D, similar findings were also observed for stages II and III disease, with patients in 2018 to 2019 hav-ing few differences in overall time toxicity.In Figure 2D, the patients with stage IV disease in 2018 to 2019 had numerically greater mean (SD) time toxicity (58.7 [43.8] vs 44.2 [26.5]  days; standardized difference, 0.40; P = .20).In a sensitivity analysis of time toxicity including home care visits and virtual visits, time toxicity substantially increased for all stages, although the proportional change in time toxicity between eras was similar (eTable 7 in Supplement 1).

Discussion
In this population-level analysis of patients diagnosed with melanoma in Ontario, we observed a significant increase in mean health care costs for systemic therapy and health care utilization in patients with stage IV disease, with respective differences in OS and high time toxicity for stage IV disease in both eras.We observed a 6-to 9-fold increase in mean per-person costs for systemic therapy among patients with stages III and IV disease between 2007 to 2012 and 2018 to 2019.In patients with stage IV disease, there were also greater mean health care utilization per-person costs.Notably for patients, there was little association with mean health care utilization per-person costs and, relatedly, time toxicity for patients with stages II and III disease.Importantly, there was a significant improvement in OS in a PS-matched cohort of patients with stages II to IV melanoma, which likely is secondary to the greater effectiveness of checkpoint inhibitors and targeted therapies in routine practice. 10,11Patients with stage I disease were excluded because of the high potential for overdiagnosis.These data highlight the value trade-off with these new effective therapies in which there is a significant increase in the economic burden to the payer and continued high time burden to patients with stage IV disease, albeit with an associated improvement in OS.These systemwide trends in value of melanoma care in routine practice are relevant to other systems, both public and private.We were able to provide a systemwide comparison of the impact of the adoption of new immunotherapies and targeted therapies for melanoma treatment.][43][44][45] Most of these economic analyses are model based, with limited analysis of patient-level data from routine practice. 45Our study thus complements the existing literature by providing a population-level perspective for a whole health care system before and after adoption of multiple novel targeted and immune-based therapies for melanoma.
The significant costs of systemic therapies for the treatment of stages III and IV disease highlight the importance of early detection of melanoma, as effective early detection and referral by dermatologic specialists may decrease the burden of advanced disease and decrease costs and improve survival with melanoma. 46Delays in melanoma diagnosis have been shown to be associated with stage progression and worse survival. 47Our data suggest that by diagnosing melanoma earlier, health care costs and time toxicity may also be less, improving the health care value.Patients with stage IV disease in 2018 to 2019 experienced a mean of 58.7 time-toxic days in the year after initiating treatment; thus, patients spent almost 2 of the first 12 months, or more than 1 day per week, in a health care facility or attending outpatient appointments.These data can be used to communicate expected burdens of In treatment to patients as well as guide improvement efforts, such as through care coordination.
When excluding the costs of systemic therapies, there were greater mean health care per-person costs in stage IV disease.This outcome is likely multifactorial.]48 Limitations Despite the strengths of this study, we acknowledge its limitations.We limited our analysis to costs within the first year after diagnosis to examine results for initial treatment.We therefore did not include costs associated with ongoing treatment, recurrence, and surveillance beyond this point.Additionally, we did not have information with respect to the individual costs incurred to patients (indirect costs), such as the cost of transportation and lost wages, which may also have affected our time toxicity estimations.We also could not account for undisclosed discounts on drug prices negotiated by the health ministry.Our measure of time toxicity may be both an underestimate of true total time burdens faced by patients (it did not include home-based care, such as telemedicine) and an overestimate of melanoma-specific time burdens (included all health care days, not only those associated with melanoma treatment, adverse effects, or recovery).However, we expect the misclassification bias to be close to identical between the 2 treatment eras, so the differences should be proportional.Our sensitivity analysis of time toxicity, including virtual and home care visits, supports this view.We were also restricted by the administrative databases' limitations regarding missing data with respect to disease stage.We note that the 2018 to 2019 cohort was based on cases with disease stage reported from cancer centers, effectively capturing nearly all patients treated for advanced cancers.Our matched analysis mitigated effect.To allow inclusion of patients with AJCC 7th and 8th edition staging, stage subgroups could not be used for matching.We note that factors associated with advanced melanoma thickness and stage were reasonably balanced in the stages II to IV cohorts, mitigating against confounding due to possible changes over time in stage subgroup case mix or sentinel lymph node biopsy use according to melanoma thickness. 27However, we acknowledge that despite the strength of using PS matching, there may still be unmeasured confounders affecting our analysis.

Conclusions
In this population-level, value-based cohort study, there was a significant increase in mean per-person health care costs over time, largely due to the high costs of immunotherapies and targeted systemic therapies for the treatment of advanced and metastatic melanoma.Additionally, there was continued high time toxicity in patients with stage IV melanoma.These results suggest the importance of early detection, as earlystage (stage II) melanoma was associated with much lower per-person health care utilization costs as well as improved prognosis.Furthermore, these results highlight the value tradeoff of these new effective systemic therapies in which there is a greater economic burden to the health care system and time burden to patients with stage IV disease but with associated improvements in patient survival.][51][52]

Figure 1 .D
Figure 1.Kaplan-Meier Overall Survival Curves for the Matched Cohorts Stratified by Stage

Figure 2 .
Figure 2. Matched Cohort of Time Toxicity and Health Utilization Within 1 Year of Treatment Initiation Among Patients With Stages II to IV Disease

Table 2 .
Mean Health Care Utilization Per-Person Costs in Canadian Dollars for the Matched Cohorts Stratified by Stage Within the First Year After Melanoma Diagnosis Mean (SD) costs are suppressed because number of patients can be backward calculated based on the SD costs.
a Significant effect size between groups based on a standardized difference of 0.20 or greater.b