AJMC's Oncology Stakeholder Summit Participants Discuss Stringencies of Data Generation and Analysis

What makes a cancer therapy effective? The answer may be in the eye of the stakeholder, even though everyone involved in healthcare decisions—from clinicians, to regulators, to payers deciding reimbursement—relies on evidence-based information.

As explored by a group of payer representatives during the Peer Exchange in September 2014, Oncology Stakeholders Summit: Evidence-Based Decisions to Improve Quality and Regulate Costs, some challenges in reaching consensus in healthcare decisions involve data. Which data are important? Is the gathered information reliable? What numbers are not gathered currently that would lead to better decisions? Taking part in the summit, which was convened by The American Journal of Managed Care, were John L. Fox, MD, MHA, senior medical director and associate vice president of Medical Affairs at Priority Health; Ira M. Klein, MD, MBA, FACP, national medical director, Clinical Thought Leadership, Office of the Chief Medical Officer, Aetna Inc; Michael Kolodziej, MD, national medical director for Oncology Strategies, Aetna Inc; Bryan Loy, MD, physician lead—cancer, Humana; and Irwin W. Tischler, DO, national medical director, Oncology, Cigna. The session was moderated by Peter Salgo, MD, professor of medicine and anesthesiology at Columbia University and associate director of surgical intensive care at New York- Presbyterian Hospital.

The panelists addressed the best ways to gather and use high-quality data to improve clinical decisions. Healthcare information should serve as a means of quality improvement that can help forge policy. Accrued data can be sourced from interventional studies—typically, an extensively regulated randomized clinical trial (RCT)—or

from noninterventional studies, such as the retrospective evaluation of available data from patient registries, patient reported outcomes, and postmarketing surveillance. However, when it comes to policy decisions by payers and regulators, which data will prove the most valuable?

Both sources are useful, the panelists agreed. According to Fox, results from RCTs help guide FDA approval for a treatment, and payers should consider trial results and the FDA approval when devising coverage decisions. However, he firmly believes that FDA approval by itself does not mean a given therapy is right for a given patient. “Optimal therapy is oftentimes not distinguished in clinical trials, and that’s where comparative data would help us define, given all the options that are available, what is best for patients, what’s best for payers, and what’s best for providers,” said Fox.

Noninterventional data, collected at different times, from different sources, can also provide useful information when integrated and could guide clinical and coverage decisions. While several data integration platforms exist, can noninterventional data replace trial data?

“I don’t think it could ever totally replace clinical trial data,” said Tischler. While he believes that a meta-analysis using various sources of data definitely broadens the horizon when it comes to personalized medicine and can expand the results of a trial in the “real world,” Tischler does not see noninterventional data sources replacing RCTs, “which still remain the gold standard.”

Kolodziej, however, believes otherwise. He argued that with treatment decisions taking a personalized approach, the complexity of the various genotypic subtypes and the phenotypic possibilities associated with each cancer would make it difficult to evaluate each condition in a clinical trial. He forecasts a change in the current approach to drug development and regulatory approval, as “personalized medicine is going to force us to reconsider what good data look like and what really actionable data look like.”

But, Fox asked, what is enough information to show that a drug targeting the BRAF V600E mutation is successful in melanoma but not in colon cancer? In response, Salgo asked the panel, if medicine lacks anything better than the current gold standard, what should the approach be? According to Kolodziej, with the cost of healthcare creating new records of sorts, redefining existing parameters and standards is essential, and he believes it will happen. Loy agreed, stating that if the healthcare dollars being spent fall short in the value equation, the point of variation needs to be unveiled.

Knowledge Gaps and Silos

Stakeholder interests vary when it comes to informed decision making—payers are primarily concerned with claims data, pharmacy data, and information on preauthorization, while regulators may be interested in biomarker status, tumor staging, and other relevant clinical information. However, any data has limitations, according to Loy. “(Claims data) is only as good as the information that’s going to be put into the system,” he said. Citing medical coding as an example, Loy pointed out that there are differences in the system between medical claims and pharmacy benefits claims. “Sometimes, what you authorize at preauthorization might not necessarily come in as a claims paid that mirrors what you agree to pay for initially. So, there’s all sorts of disconnects even within our own systems.”

Fox added that in order to be thorough in their processing of patient claims, payers want to evaluate clinical information the providers may offer. “Our goal in paying for therapies is to make sure the right patients are getting the right therapy. That’s why we have processes in place, whether it be prospectively or retrospectively, to ensure that our patients are getting the drug that’s appropriate for their biomarker status and their stage,” he said. Loy added that these processes that payers utilize help raise flags when things are off schedule, when therapy is delayed or changed, or when the patient has associated comorbidities.

Data Integration to Fill the Gaps

Would the availability of clinical information help fill data gaps and aid reimbursement decisions? Definitely, said the panelists, adding that data silos need to be overcome to improve cross-talk between various databases. Everything from employee absenteeism to clinical information should be integrated to create models that can guide effective treatment decisions, which can ultimately help contain cost.

Kolodziej believes that the provider plays the most crucial role and he needs to look at the complete picture. To prove his point, Kolodziej offered an example: A physician has 2 patients, a 40-year-old woman and a 70-year-old woman, who express the same biomarkers and need treatment for adjuvant breast cancer. When treating the older patient, it’s essential for the treating physician to understand comorbidities and other risks that may be present prior to treatment initiation. With the younger patient, however, the physician must consider how different treatments will affect her ability to return to work.

While data sharing may be restricted by the Health Insurance Portability and Accountability Act, Klein thought some leeway is necessary for adequate data sharing. The elephant in the room, however, is the broken healthcare payment system and the lack of post-approval data collection, Fox said. “We pay hundreds of billions of dollars for chemotherapy and for cancer treatment, but once a drug is approved, we have no way of collecting, aggregating, and interpreting that data to know how it works in the real world.”

Noninterventional data, in the form of cancer registries, medical chart reviews, electronic health records, and claims data, is available. Harmonizing the data using integration platforms, such as the Mosaic Platforms developed by Remedy Informatics,¹ could prove a tremendous clinical advantage. But these platforms are not yet in widespread use by the healthcare world, said Loy.

Connecting the Dots

The Quality Oncology Practice Initiative, started by the American Society of Clinical Oncology (ASCO),² seeks to improve and regulate the quality of data being generated. Asked to comment on data quality and attempts to govern it, Fox pointed to a bigger problem: integrating patient data generated at different centers to compose a comprehensive picture of the continuum of care. Loy agreed, saying that the cost of care for a cancer patient is not restricted to chemotherapy or radiation—that is just 1 piece of the puzzle.

Considering data attributes, whether interventional or noninterventional, would depend on the stakeholder’s interests. We need an integrated, coordinated, data platform that would allow cross-talk between the various systems to guide optimized care delivery.

Salgo then asked the panel about the value of registries and databases—such as the National Cancer Institute’s SEER database, which documents cancer incidence and survival data—in predicting cancer risk. The consensus was that, historically, SEER was not developed to guide treatment decisions or improve transparency that would allow patient understanding of care decisions. “Medicare SEER has limitations in terms of how the data is collected, how complete the data is, whether it’s an accurate representation of the country as a whole, or whether it’s urban predominated; which it is. So Medicare SEER is a great idea in principle, but in practice, it is not a useful tool,” said Kolodziej.

Fox agreed, and said that SEER data needs better national representation to be useful as a registry that can be tied to outcomes and costs. The process to connect the dots already has been initiated, said Kolodziej, via programs such as ASCO’s CancerLinQ—a platform that promises real-time quality feedback and clinical decision support to providers, along with analytical support that can analyze current treatment patterns and help improve future care rendered to patients.³

Healthcare and Big Data

Several projects have adopted the concept of data sharing to promote healthcare decision-making. The National Institutes of Health launched the Big Data to Knowledge (BD2K) initiative to incentivize data sharing among biomedical scientists who work with big data sets and generate large amounts of information through their research.

This platform aims to provide easy access to, analysis of, and integration of the information.⁴ A similar program developed by the pharmaceutical and biotechnology industry, Project Data Sphere, aims to share, integrate, and analyze phase 3 comparator-arm data from oncology trials.⁵

The panelists did see value in these programs. Tischler thought that despite being a late entrant in oncology, big data would enhance data access, while Loy saw value in comparator arm data. Fox, however, argued over the economics of these ventures. Using CancerLinQ as an example, he asked, “Who owns the data and how much is it worth? And if you were to come to my company and ask me for money for that data, is that public data or is that private data? Is this all for public good?” Klein concurred that these issues need to be clarified upfront.

Influence on Drug Development

Ultimately, how would the drug development process be influenced? Or would it? For starters, it would entail overcoming federal privacy barriers in the healthcare industry. The discussion ended with panel members disagreeing on whether the path to transparency and openness would result in the pharmaceutical industry making healthcare decisions based on patient demands.

EBO

Argued Kolodziej, “What’s going to happen is let’s say you have renal cell carcinoma and you’re trying to decide what to do. There will be information available that will be transparent. ‘This is what it costs me in terms of human misery, out-of- pocket costs, and toxicity.’ Patients will say clear as the nose on your face, ‘That’s the wrong treatment for me based on what’s on Google,’ and we’re going to have that kind of thing in healthcare. I’m quite sure of it.”

Note:

To see the videos of this Peer Exchange, please visit http://www.ajmc.com/ajmc-tv/peer-exchange/oncologystakeholder-summit-2014/Episode-1-Evidence-Based-Decisions-for-Data-Generation-Collection-and-Analysis.References

1. Mosaic Platform. http://remedyinformatics.com/mosaic-platform_/. Remedy Informatics website. Accessed October 1, 2014.

2. Dangi-Garimella, S. QOPI, the ASCO initiative, improves compliance and promotes quality of patient care. Am J Manag Care. 2014;20(5):SP153-SP155.

3. CancerLinQ. http://www.asco.org/qualityguidelines/cancerlinq. American Society of Clinical Oncology website. Accessed October 1, 2014.

4. About BD2K. http://bd2k.nih.gov/about_bd2k.html#sthash.cZgHyq0Y.dpbs. National Institutes of Health website. Accessed October 1, 2014.

5. Project Data Sphere. http://ceo-lsc.org/content/project-data-sphere-1. CEO Life Sciences Consortium website. Accessed October 1, 2014.

Download PDF: AJMC's Oncology Stakeholder Summit Participants Discuss Stringencies of Data Generation and Analysis