Personalized medicine – what does the future hold?
Rebecca Aris interviews John Hornberger
Dr. John Hornberger explains the current challenges surrounding personalized medicine and how the decisions that are being made now will dictate the future direction of this area.
Personalized medicine is at a crucial standpoint right now and the future landscape is being dictated by current decisions in this space.
Ahead of his speaking engagement at Personalized Medicine World Conference 2013, we speak with Dr. John Hornberger, physician and CEO and president of Cedar Associates. He discusses the current status of personalized medicine and what the future holds in this area.
Dr. John Hornberger shares his thoughts on what challenges surround personalized medicine right now and how they will impact on how personalized medicine will develop.
RA: Dr. Hornberger, thanks for taking part. Could you please start by explaining the challenges associated with reference-based pricing when it comes to molecular diagnostics?
JH: It is generally accepted economic theory that a fully competitive marketplace leads to the optimal allocation of resources, and prices, that reflect the value of products and services. In 1963, Kenneth Arrow, a Nobel prize laureate in economics, wrote that achieving a competitive marketplace for healthcare services and products is inherently difficult. In lieu of absence of such a marketplace for healthcare products and services, there evolved a process for basing prices on some appropriate reference.
An important issue is determining the appropriate reference. There exists a number of options. One option is reference to the level of effort required to produce a product or service, referred to as labor theory of pricing. Another option is to identify products or services that are similar such that the price could be correlated to existing prices, referred to as predicate-based pricing. Yet a third option is associating the price with respect to the value it provides to a particular stakeholder, such as a representative patient, or a group purchasing the product or service, or a hypothetical representative of society as a whole. This latter type of reference considers holistically both the benefits to patients and costs of adoption of the product or service, hence, it is referred to generally as “value estimation”.
There are a number of well-validated methods for estimating value. One common approach is cost-benefit, or cost-effectiveness, analysis. Another approach is called multicriteria decision analysis, which uses ratings across a range of highly relevant aspects of a technology, including the benefit to patients, the quality of evidence on the validity of the product or service, potential to contain costs, the unmet need, and the level of true innovation. One of the challenges is reaching consensus on which of these approaches would achieve allocation of resources, and prices, that optimizes social welfare.
“…achieving a competitive marketplace for health care services and products is inherently difficult.”
RA: How do these challenges impact on personalized medicine?
JH: This fall, Medicare released a preliminary determination on the reimbursement of multi-analyte algorithm-based assays (MAAAs). Such assays generally are designed – with use of mathematic formula of biomarkers sometimes combined with clinical parameters – to more accurately predict diagnosis, risk for recurrent disease, or response to treatment. Medicare indicated a willingness to pay for determination of analytes as it has traditionally done, but an unwillingness to reimburse for algorithm-based assays. In the past, the agency decided not to pay some types of algorithms, with a notable example being the anion gap. The anion gap is computed using a simple algebraic expression of cations and anions in serum, plasma, or urine.
However, the newer algorithms associated with the field of personalized medicine are arguably substantially different, involving a deeper understanding of complex biological pathways / networks that include many more clinical, pathological, and genomic variables. The sophistication of these algorithms – applying newer principles developed by the biostatistical and the artificial intellegence research communities – are such that even the most mathematically capable members of the human race cannot easily replicate the computations by themselves. Establishing a policy not to reimburse for all such algorithms by referencing to simple algebraic expressions of a few analytes simply lacks face validity. It is akin to assigning all the value of a computer to the hardware and assigning no value to the software. Mathematically sophistocated people use computers all the time to help them do their jobs well. Would we want pilots to fly modern commercial jets without computers, or have neurosurgeons enter a patient’s brain without use of imaging machines embedded with computers? But how many people would purchase a computer without functioning software? Without algorithms to help interpret the complexity of biology, sequencing the genome is likely to be of very little value.
Without a process to pay for quality algorithms that emerge from research and development, there will be a disincentive to invest in this field. Basing the value, and hence the reimbursed price, of these new and evolving tools of personalized medicine in reference to historically and inherently much simpler algorithms, is currently the biggest threat to this nascent field.
“Without algorithms to help interpret the complexity of biology, sequencing the genome is likely to be of very little value.”
RA: What changes would you like to see to reference based pricing with regards to molecular diagnostics?
JH: I prefer to see the reference for pricing be correlated with the value that the predictive tool brings to society. This type of valuation research has been evolving for several decades. For these types of assays, it means quantifying the predictive validity of the assay (also called clinical validity) and how the tests affects clinical practice and the quality of care. Moreover, it means quantifying the impact of changes in clinical practice that result from use of the test on clinical outcomes, referred to as clinical utility. Increasingly, payers also are requesting evidence on how changes in clinical practice and outcomes affect costs. The two major scientific societies whose members have been at the forefront of this work, the International Society for Pharmacoecomic and Outcomes Research and the Society of Medical Decision Making, recently issued joint guidelines on best practices for assuring the quality and validity of these methods and the results that stem from this type of work. In response to the urgency felt to build a more sustainable healthcare system, and perhaps even motivated to build stronger, more competitive businesses, some commercial payers are starting to apply these approaches when valuing innovations in products and services. This is definitely a step in the right direction.
RA: What potential do risk stratifiers hold when it comes to predicting risks of a current response to chemotherapy in early stage breast cancer?
JH: In the early 2000’s, more than 50% of women in the United States with early stage breast cancer received adjuvant chemotherapy even though the risk of recurrence at 10 years was less than 25%. Also, some women had recurrence of breast cancer having decided to forego adjuvant chemotherapy. Despite advances in tumor biomarker development and validation, such as estrogen and progesterone receptors and HER2/neu, it was clearly understood by many investigators that emerging understanding of the human genome would unveil new multi-analyte markers (algorithms) to improve prediction of recurrence risk and even response to therapy. In 2012, at least 4 multi-analyte algorithms had become commercially available, and more yet are in development. A recent, independent review published in the Journal of the National Cancer Institute assessed the evidence on clinical validity, clinical utility, and affordability of these assays using a standardized framework for judging evidence developed by Drs. Simon, Hayes, and Paik. What was revealed in this review is that there is quality (called level I evidence) that at least one of these assays predicts risk of distant recurrence, overall survival, and response to chemotherapy. In other words, it is no longer a question of potential benefit to predict these outcomes, rather, based on the best available science, newer risk stratifies using multi-analyte algorithms can predict these outcomes better than the traditional methods existing in the early 2000’s. This means fewer women who have low risk of recurrence or a low likelihood to respond to chemotherapy will undergo chemotherapy that provides them little benefit while exposing them to risk of adverse events. Conversely, for women at high risk who previously would forego chemotherapy, better awareness of this high risk will lead to more deciding that the benefits of chemotherapy outweigh the risks. Overall, the duration and quality of their lives should improve with this better risk stratification. It is expected, contingent on there being in place reimbursement incentives for investment in research, that even more accurate risk stratifiers will emerge.
“…personalized medicine could not only help improves lives, but make health care more affordable…”
RA: What are the economic implications of risk stratifiers?
JH: The economics of risk stratifiers depends on its predictive ability, how its use affects clinical practice, and changes in outcomes that are associated with costs to society. In the case of risk stratifiers for breast cancer, the most direct effect compared with practice in the early 2000s was to reduce the use and thereby cost associated with chemotherapy in women with low risk of recurrence. For one of the lead classifiers, more than 20–30% of decisions were found to change in studies across the world.
Costs of adjuvant chemotherapy are related to the drugs, administration, and management of adverse events. Overall, reducing the use of adjuvant chemotherapy by this amount resulted in cost savings to the health care system. This was demonstrated in an independent study conducted jointly with Humana, and published in the Journal of Oncology Practice. Another economic effect occurs for those women reclassified as high risk and who decide to take chemotherapy. This results initially in more costs because of adjuvant chemotherapy, but much of this cost is offset by reduction in recurrence rates and the higher costs of managing recurrence. Overall, considering both groups of women resulted in overall savings for a large US commercial payer, presumably the effects would generalize to the others who practice in a similar manner. The costs for a course of adjuvant chemotherapy is not as high outside the US, so the assay costs were not entirely offset by lower costs of adjuvant chemotherapy. The paper by Klang et al. published in Value in Health, reported the experience of a large managed care organization in Israel. They found that the test, while not necessarily cost saving in Israel, was cost effective having a cost per patient benefit in quality years of life gained well below thresholds of adopting innovative health technologies. These studies, and others that have confirmed these key findings, lead me to conclude that the hopes – as expressed by many experts in the early 2000s – that personalized medicine could not only help improve lives, but make healthcare more affordable, can and has been realized. It is important to build on these early successes.
RA: What do you think personalized medicine will look like in the future?
JH: The scientific rationale for the field is compelling. The innovation industry is showing that it can deliver and translate research at the bench into commercial assays that are more predictive than current methods alone, that these products can impact clinical practice decisions, improve the lives of patients, and even help in some instances to reduce medical care costs. But this is still a nascent field. Assuming the challenges of reimbursement can be overcome so as to motivate further developments in the field, I see several trends for personalized medicine. First, advances in sequencing technologies will allow even greater awareness of the extent of variations in fundamental biology at the individual level, and not just for strata of patients. In other words, the field will move from what is now referred to as mostly ‘stratified medicine’ to truly ‘personalized’ medicine. Second, advances in biostatistics and artificial intelligence will continue to reveal the sophisticated and complex interactions that translate from the genome to the phenotype. This will lead to even more predictive – yet complex – algorithms for use in the clinic. Early examples of both these trends are well on their way, in which researchers are finding it possible to identify specific mutations in drug target and metabolism pathways that will allow physicians to create ‘cocktails’ of drugs that are specifically individualized to a particular patient. In the long term – perhaps involving decades – algorithms based on solid understanding of biology will be so predictive that it will transform the role of the physician. This is undoubtedly a more uncertain and controversial prediction, but it is one that I communicate regularly to young physicians in training in my Stanford clinic so that they may start to understand and plan their careers in medicine accordingly. This is less of a prediction, as it is already happening, but these advances in the field have, and will continue to, challenged traditional regulatory approval processes, and assumptions about how to assess the value, and appropriate pricing, of health innovations.
RA: Great, well Dr. Hornberger thank you very much for your time and for your insights.
JH: Thank you. It has been my pleasure.
About the interviewee:
Dr. Hornberger (CEO and President, CEDAR Associates LLC) is a health services and policy researcher focused on comprehensive assessments – i.e., clinical outcomes, costs, and cost-effectiveness – of emerging pharmaceutical, diagnostics, and device technologies in cancer, infectious diseases (HIV, hepatitis, and herpes zoster), renal disease, eye and skin disease, and others. In cancer, he has published seminal papers on the evaluation of genomic assays, clinical utility studies, and on the cost-effectiveness of specific assays. Other assessments he has conducted include: rituximab for treatment of diffuse large B-cell lymphoma in older adults, capecitabine for anthracycline-pretreated patients with advanced breast cancer, cervical vaginal smear interpretation of atypical squamous cells of undetermined significance (ASCUS), and testing strategies in the diagnosis of lung and breast cancer. His work is cited in international appraisals of new cancer technologies, such as the UK National Institute of Clinical Excellence and the Canadian Coordinating Office for Health Technology Assessment.
He has been involved in facilitating first-ever development of clinical guidelines for such conditions as adequacy of renal dialysis and management of hyperhidrosis. He is working on various projects concerning hurdles that prevent widespread adoption of anti-retroviral medications for persons living with HIV in severely resource-constrained parts of the world, and on how health technology coverage decisions are made by payer groups in low- and middle-income countries.
Dr. Hornberger also is an Adjunct Clinical Professor of Medicine at Stanford University School of Medicine, and is attending in a weekly clinic, supervising and teaching resident housestaff and post-doctoral fellows in the general internal medicine clinic. He received his M.D. degree from the University of Rochester, New York and M.S. degree in Health Services Research from Stanford University. He has papers published in Journal of National Cancer Institute, Journal of the American Medical Association, Annals of Internal Medicine, American Journal of Public Health, Statistics in Medicine, Cancer, Controlled Clinical Trials, JAIDS, Medical Care, Management Science, American Heart Journal, Journal of the American Society of Nephrology, and Medical Decision Making. He is Co-Editor of Value in Health, is on the Editorial Board of Medical Decision Making, and has served on various NIH scientific and economic advisory committees.
Where next for personalized medicine?