Predictable Cost of Care Model for Treatment Decisions: Working Group Consensus Statements for Metastatic Non-Small Cell Lung Cancer

Since the advent of pathways in 2005, the fundamental criteria in evaluating value-based treatment options in cancer care have remained unchanged: efficacy, safety, then cost.^1,2 While cost is not the primary driver of pathway determination, it is essential to recognize that the expense of cancer care often exceeds hundreds of thousands of dollars each year.³ Therefore, it is critical to understand all factors that contribute to cost that can lead to better-informed treatment decisions. This can ultimately improve patient outcomes by incorporating evidence-based care that is also cost-effective.  

Historically, the cost consideration in pathways has only included the cost of the cancer drugs themselves at average sales price (ASP), leaving the non-drug cost of care out of the picture. Pathway developers and entities concerned about the total cost of care may be missing essential elements of the cost picture when assessing only drug costs under the existing framework. Additionally, with more targeted therapies and fewer toxic agents entering the market, many believe that the economic advantages a new treatment may bring are underrepresented in the current cost models pathway developers use today due to the belief that non-drug costs will be much lower.  

To better understand the economic impact of a drug’s use and broader cost comparisons to be considered in pathway inclusion, a multi-stakeholder Working Group (Group) was formed to collaboratively develop an industry-standard cost of cancer care model that would provide a standard methodology for evaluating the predictable total costs associated with therapy.  

The Predictable Cost of Care (PCC) model aims to provide an enhanced picture of the costs associated with a specific drug regimen beyond just the cost of the drug(s). The model will provide a standardized basis by which pathway developers can quickly review and compare product costs to make more informed treatment recommendations. The PCC model seeks to a) develop parameters and structures for estimating the “predictable cost of care” using data that pathway developers will find relevant and trustworthy and that pharmaceutical manufacturers can realistically provide; b) choose service utilization elements that are relevant, informative, predictable, and knowable, intentionally omitting those that are unpredictable or inherently unknowable; c) identify independently verifiable, publicly available source(s) for unit cost data; and d) develop a sample model that incorporates these utilization and cost elements. While it is necessary to consider the total cost of cancer care, the PCC model aims to arrive at the ‘predictable total cost of care.’ As such, it will exclude baseline costs (not directly associated with a regimen), comorbidity and non-cancer-related costs (not associated with the regimen), and other healthcare costs that might be included in a total cost of care calculation (hospice, trauma, etc).

The Group emphasizes that while the initial impetus of the PCC model may have been to facilitate and ensure a high-quality, reliable exchange of cost-related data between pharmaceutical companies and pathway developers, numerous other use cases have been identified, including for a) practices looking to estimate costs for value-based care; b) payers and other value-based care facilitators building bundles; c) actuaries or consultants modeling costs for employers; d) patient advocacy groups looking to estimate cancer-related costs for patients and determine financial assistance. Accordingly, on the 20th anniversary of cancer treatment pathways, the Group envisions that the PCC model will serve as a standardized set of methodologies accepted and practically utilized by various stakeholders in oncology care, providing them with additional insights that can inform patient care in value-based care settings.

Panelist Selection and Meetings

The Group consisted of representatives from 7 leading pathway developers with diverse experiences crossing academic medical centers, community oncology, payers, and practices operating in a value-based care environment. Five virtual meetings were conducted, during which the Group outlined, discussed, and refined the cost elements and parameters to be incorporated into the PCC model. In the initial phase of the process, the Group identified metastatic non-small cell lung cancer (mNSCLC) as an ideal disease state for the pilot PCC model. Based on the discussion of crucial cost elements of cancer care, the Group identified the domains that should be considered in developing the PCC model.  

Modified Delphi Method

A modified Delphi method was utilized to arrive at a consensus and understand areas of non-consensus on the cost elements and parameters to be adapted for the PCC model for mNSCLC. Three Steering Committee (SC) members with extensive experience in clinical oncology and health economics were nominated by the Group to oversee the process. The Group served as the Expert Panel (EP) to guide the development of the consensus statements. Lastly, the Consensus Group (CG), 20 experts in oncology pathway development, health economics modeling and analytics, and the pharmaceutical industry participated in the surveys.  

Consensus Statements and Surveys

Consensus statements were drafted based on the EP discussions of cost elements and care parameters for patients with mNSCLC that are relevant and appropriate to be incorporated into the PCC model. Two rounds of electronic surveys were conducted to gather the degree of consensus among the participants. During the initial survey round, the participants indicated their agreement or disagreement with each statement and provided a rationale and suggestions for modifications in cases of dispute. Consensus for an item was defined as agreement by at least 80% of the respondents. Non-consensus items were identified from the first survey and modified by the EP following a review and discussion of the rationale for disagreement. The second-round survey, consisting of the modified non-consensus items from the first round, was distributed electronically. The results of the surveys were reviewed, and the consensus items were utilized in developing the PCC model.  

For the treatment of patients with mNSCLC, the Group identified 4 cost categories and 3 framework parameters that need to be defined for the PCC model: a) diagnostic costs, b) treatment costs, c) treatment administration-related costs, d) costs related to treatment complications, e) data sources and framework, f ) timeframe, and g) statistical parameters (Table 1). For each domain, the Group made recommendations for the data sources. 

Cost Elements

Diagnostic Costs

While diagnostic workups can be a significant component of the cost of care for patients with mNSCLC, the Group argued that pretreatment diagnostic workups are increasingly becoming a recommended standard for all patients. Accordingly, the Group asserted that costs associated with diagnostic workups should be considered preliminary diagnostics that all patients should receive on a standard basis, not specific to a treatment. The Group agreed that the PCC model for mNSCLC should exclude the costs of the diagnostic workup, including diagnostic scans, biopsies, cancer-related surgeries, companion diagnostics, and comprehensive genomic panels unless specifically related to the treatment being modeled. 

Treatment Costs

The cost of drugs is often the most significant and predictable cost component in the care of oncology patients. The Group recommended that the PCC model for mNSCLC should include the costs of drugs, such as chemotherapy, targeted therapies, and immunotherapies. Additionally, they suggested that the model output of the drug or drug regimen cost should be specific to the indication and the dose and schedule specified in the US Food and Drug Administration (FDA) package insert, included in nationally recognized clinical guidelines/consensus statements, and from the medical literature (eg, for combination therapies or regimen). Regarding the data source, the Group recommended that the drug costs be defined using the ASP or wholesale acquisition cost (WAC) for orally administered therapies. For a new drug being launched, they recommended using WAC or average wholesale price (AWP) until a Medicare value is published, which typically takes 6 months after launch.    

Treatment Administration-Related Costs

In addition to the cost of drugs, the Group concluded that treatment administration-related costs should be included in the PCC model for mNSCLC. The Group specified treatment administration-related costs as the costs of a) preparing and administering therapies (eg, infusion costs for intravenous [IV] therapies); b) pretreatment assessment for safety (eg, baseline electrocardiogram, liver function test); c) pretreatment, concurrent treatment, or posttreatment support medications for known symptoms or known adverse effects (eg, antinausea or antidiarrheal medications, steroids, granulocyte-colony stimulating factors); d) hospital or outpatient center fees for medication administration and posttreatment monitoring; and e) follow-up visits and surveillance (eg, imaging and lab tests for monitoring response/progression/adverse events [AEs], appointments with the oncologist/primary care physician/case manager). In addition, the Group recommended the Centers for Medicare & Medicaid Services (CMS) Medicare Physician Fee Schedule (MPFS) fee schedule and ASP as sources of cost data for administration procedures and symptomatic treatment products, respectively.

Costs Related to Treatment Complications

The cost associated with treatment complications, such as side effects management, is another significant component of the cost of care for patients with mNSCLC. The Group recommended that costs associated with known treatment-related complications based on clinical study data should be included in the PCC model for mNSCLC to evaluate the cost implications of complications arising from treatment. The Group specified the following cost elements for inclusion in the model: the costs of a) procedures for managing severe and known side effects (grade ≥ 3) of cancer treatment, b) emergency department (ED) visits resulting from complications, c) hospitalization for treatment complications or symptom management, and d) other provider-related costs including healthcare provider appointment or intervention resulting from the complications.  

However, the Group reached a consensus to exclude the costs of treatment failure (eg, radiotherapy for brain metastasis) and terminal care costs (eg, palliative care and hospice services), as this consideration would be tumor-specific and would have been factored in during the pathway review of the therapy’s efficacy.  

The Group determined that to enable comparisons between therapies, some of which may be newly approved and only have clinical trial data available, clinical trial data and peer-reviewed published data should be used to calculate the costs associated with managing AEs. Further, they noted that incidence data for ED visits and hospitalizations related to AE management may be obtained from pharmaceutical companies, as the FDA requires them to report. Lastly, for therapies with new side effect profiles, the Group recommended consulting a sub-panel of clinicians on the management strategy and associated cost.

As recommended by the American Society for Clinical Oncology (ASCO) 2016 Policy Statement on Clinical Pathways in Oncology, the pathway development process should be clear, consistent, and transparent to all stakeholders.⁴ An essential requirement for a model that enables cost-of-care comparisons between treatment options is the transparency of the data sources. To ensure this, the Group recommended utilizing the following publicly available sources for cost data to be included in the PCC model for mNSCLC: a) the CMS ASP for unit costs for drugs when they become available, b) Medicare fee schedule for the drug administration cost data, and c) Medicare reimbursements for costs of AE management procedures. The Group noted that additional service utilization data may need to be collected from trial centers for data not published at the time of trial data disclosure.

The Group also agreed that pharmaceutical companies should provide the utilization (unit of service) recommended per indication over the defined timeframe based on clinical trial experience. They stipulated that the PCC model should calculate costs associated with the clinical trial benchmarks based on pivotal trials that led to drug approval. However, they noted that the inputs and variables might be modified as real-world evidence, including articles and case reports/studies published in peer-reviewed journals, FDA surveillance reports/advisories, and other post-marketing surveillance reports becomes available. The Group agreed that only clinical trial data would be utilized for the initial use case of the PCC model, which would be the submission of enhanced cost data beyond the cost of drugs alone for newly approved drugs or indications. They noted that if a reconsideration is warranted, real-world data (RWD) may be included and submitted as a blended or RWD-only model. While the overarching model may provide the combined cost, the Group noted that a supplementary file may be created with the availability of data input fields for individual center use.  

By relying on independent sources of information, the PCC model aims to be “self-validating.” These sources include package inserts published by the FDA for dosing and side effect severity and frequency; published ASP or WAC data for drug pricing (or list price for drugs not yet on the ASP file); Medicare allowable costs for fees for professional services, labs, imaging, monitoring, etc; and the number of units based on published clinical trial data. The Group noted that while the model values are based on clinical trial outputs, additional validation may be needed in the future as real-world data accumulates.

In terms of parameters for the PCC model for mNSCLC, the Group recommended basing calculations on the regimenspecific costs of care for an average patient (as defined by the Centers for Disease Control and Prevention [CDC]) receiving a specific regimen: 80 kg (weight), 5’7” (height), creatinine  0.8 mg/dL (normal).

Timeframe

To enable a cost comparison of treatment regimens, it is necessary to establish a timeframe during which care costs should be considered. The Group recommended utilizing 6 months and 1 year as the benchmark timeframes for the PCC model for mNSCLC. They specified that the costs of drugs and administration should be calculated based on the indicated dosing and schedule for the regimen and projected out to 6 months and 1 year. Some members advocated for a longer timeframe; however, costs exceeding more than 1 year may be more difficult to predict and less relevant for pathway determination. Therefore, the Group agreed that the costs of AE management should be added to arrive at the total predictable cost for 6 months and 1 year, which can be calculated as monthly cost if needed.  

Statistical Parameters

Regarding the model output, the Group recommended that the PCC model provide both mean and median costs, with the range, standard deviation, and confidence interval provided if data granularity allows. They noted that the model should calculate costs inclusive of all outliers; however, it should indicate the inclusion of outliers and refer to the differences in mean and median costs to see the outlier impact. The Group also recommended that the PCC model include information on data sources, handling, preparation, and analysis for review as best practice.

Using a modified Delphi method, the Group defined key predictable cost elements to be considered in developing the PCC model for mNSCLC. The Group decided to exclude diagnostic costs while including drug costs, costs of administration, and some costs associated with treatment complications in the PCC model. They also defined data sources for each cost element, the timeframe for cost calculation, and statistical parameters for the output.  

The Group would like to emphasize that the initial PCC model was formulated with pharmaceutical companies as the submitters and pathway organizations as the data recipients. To that end, pharmaceutical companies and major pathway developers were involved in developing the PCC model. In the next phase, the Group intends to broaden participation by including community practice pathways and additional pharmaceutical companies. 

The initial PCC model does have some limitations. While it was developed with NSCLC as the model disease state, it was intended to be comprehensive and broadly applicable. It is worth noting that the PCC model would be applied for each indication to assist decision-making by keeping the variables controlled. However, as treatment paradigms vary, developing different versions of the model between solid tumors and hematologic malignancies may be necessary.  

During the discussion, patient-relevant data points, such as time toxicity, were considered for inclusion in the PCC model. Although they expressed a desire to include a component addressing time toxicity, the Group concluded that it is highly variable and challenging to define. They recommended this element be included in a future version of the PCC model.  As a cost model for a specific drug, the PCC model is intended to accurately reflect the cost associated with an average patient for that therapy. While no weighting is built into the model, the Group recommends that the PCC model report individual cost segments (drug, administration, monitoring, side effect management, etc) for the users to determine whether weighting of the individual cost elements may be warranted.

The Group intends for this initial PCC model to evolve into an industry standard. It has the potential to inform more effective, value-based treatment recommendations for multiple stakeholders involved in oncology care, ultimately improving outcomes for cancer patients.

1. Ginsburg A, Neubauer M, Wilfong L. Update on the evolution of the US Oncology Network’s clinical pathways program. J Clin Pathways. 2020;6(9):38-41. doi:10.25270/jcp.2020.11.00001
2. Mullangi S, Chen X, Pham T, et al. Association of patient, physician, and practice-level factors with uptake of payer-led oncology clinical pathways. JAMA Netw Open. 2023;6(5):e2312461. doi:10.1001/jamanetworkopen.2023.12461
3. National Cancer Institute. Financial burden of cancer care. Updated March 2024. Accessed November 7, 2024. https://progressreport.cancer.gov/after/economic_ burden  
4. Zon RT, Edge SB, Page RD, et al. American Society of Clinical Oncology criteria for high-quality clinical pathways in oncology. J Oncol Pract. 2017;13:207-210. doi:10.1200/JOP.2016.019836 

Affiliations:

¹Tennessee Oncology, Nashville, TN, USA; ²Evolent, Arlington, VA, USA; ³McKesson, Irving, TX, USA; ⁴Dana-Farber Cancer Institute, Boston, MA, USA; ⁵Kuntz Consulting,  Minneapolis, MN, USA; ⁶Elsevier, Munich, Germany; ⁷ConcertAI, Cambridge, MA, USA; ⁸Thyme Care, Nashville, TN, USA

Funding:

Johnson & Johnson and Eli Lilly and Company sponsored the Predictable Cost of Care Working Group but did not influence the participants’ findings.

Correspondence:

Carole Tremonti, RN, MBA

ctremonti@concertai.com

Acknowledgments:

Medical writing assistance was provided by HMP Collective.

Disclosures:

V.G. is employed by Evolent. R.H. is employed by McKesson. A.H. is employed by and owns stock in Evolent. O.L. is employed by Thyme Care, serves on an advisory board for Pfizer, and is a speaker for the National Comprehensive Cancer Network, American Journal of Managed Care, and HMP Global.

All other authors have no disclosures to report.