When Raju Kucherlapati, the Paul C. Cabot Professor of Genetics at the Harvard Medical School and chair of the organizing committee for the conference, “Personalized Medicine: The Time is Now,” summed up the defining annual event on November 18-19 in Boston, he noted that the time for personalized medicine is not only now but that personalized medicine is here.
Five years ago the Personalized Medicine Coalition helped organize the first international conference at Harvard entitled, “Personalized Medicine: Promises and Prospects.” One third of the size of this year’s conference, which brought together over 600 high level executives from across the healthcare spectrum, it made the case that the accelerating pace of discovery was rapidly making one-size-fits-all, trial and error medicine obsolete. We argued that it was important, if not critical, to realign our regulatory, reimbursement, and educational systems to harness the power of new insights and new technologies for the benefit of patients who want safer and better drugs as well as for society, which can no longer afford the inefficiencies of 20th century medicine.
Five years ago we were struggling to define personalized medicine while the Royal Society, the United Kingdom’s independent academy of science and technology, issued a troubling report on the future of personalized medicine that may be summed up in a single phrase: Don’t hold your breath.
Today, as Dr. Kucherlapati pointed out, we see evidence of personalized medicine’s potential impact on millions of people. Not only is the list of widely prescribed drugs targeted for specific populations growing for many different kinds of cancer, heart disease and other disorders, it is being adopted in the United States by large institutions whose programs have the capacity to touch millions of people.
DNA Direct, a company that provides guidance and decision support for genomic medicine, and Humana, an insurance company that covers 11 million people, have announced a program to help physicians understand how genetic counseling can assure better outcomes and avoid unnecessary expenses. Generation Health, a new genetic testing health benefit management company, is partnering with CVS Caremark, a pharmacy benefit manager for 50 million subscribers, to analyze more than a dozen drugs with associated diagnostic tests to determine the best prescriptions for patients with selected diseases including cardiovascular disease, cancer and HIV. The two companies think that they can also improve patients’ health, reduce adverse events, and save costs by not prescribing drugs to non-responders. And Medco, the largest pharmacy benefit manager in the United States, has created a program to help advise physicians when to prescribe a diagnostic test to improve patients’ health and lower costs of care for their members. Medco has invested heavily in its own research, for example, to test the hypothesis that genetic testing can improve warfarin dosing. Warfarin is the second most widely prescribed drug in the United States and also, according to FDA, the second leading cause, after insulin, of drug-related emergency room visits.
In a sign of the times, Pfizer and Abbott, two large pharmaceutical companies, have formed a partnership to develop a drug and companion diagnostic test for non-small cell lung cancer, based on the knowledge that some 6 to 7 percent of patients with these tumors have the genetic markers that make them good candidates for the drug in development. According to Stafford O’Kelly, president of Abbott Molecular, “What’s happening is that [once] the understanding of the molecular basis for disease is validated, it’s increasingly important to get these tests.”
To be sure, we have a long way to go before all therapies are linked to molecular diagnostic tests of one sort or another. But, as Mr. O’Kelly’s comment suggests, we are on the right road because the science points in that direction.
By Edward Abrahams, Executive Director, Personalized Medicine Coalition.
The Age of Personalized Medicine Blog 
December 2, 2009 at 11:29 AM |
“The time is now” for personalized medicine. However, trying to do personalized medicine with science that uses group averages to assess causality as with current gold-standard randomized controlled trial (RCT)designs can be likened to trying to do nuclear science with Newtonian physics. We need a new paradigm. Personalized medicine calls for a new scientific standard for causality assessment in which the effects of individual differences, which includes effects of genetic differences on disease and treatment response, can be accounted for rather than being lost in group averages.
This specific problem often does have a specific technical solution. This solution is suggested by the following observation in the context of chronic health problems. Clinicians often assess whether drugs cause harmful and/or beneficial effects over time for individual patients by making subjective judgments about responses to drug challenge, de-challenge, re-challenge and other more incremental changes in dose. Such evidence collected over time for individuals can trump evidence collected across individuals in groups as when clinicians decide to stop using approved drugs for particular patients because liver enzyme results suggest that particular patients are at risk of liver failure and death.
We can help enable personalized medicine by assessing causality scientifically over time for individual patients. The scientific version of assessing causality over time has four major steps. These steps are presented in the context of drug treatment evaluation first for individual patients and then for groups and populations of patients with chronic health problems.
1. Randomize two or more doses of a particular type of drug to different periods of time for each patient. Placebo is zero-dose. Randomization is best done with blinding or masking.
2. Collect data periodically about both independent variables (e.g., doses) and dependent variables (e.g., signs, symptoms, risk factors, laboratory measures, biomarkers) for each patient in the form of an advanced electronic health record. Include as many dependent variables as feasible to obtain more detailed and comprehensive scientific assessments of safety and effectiveness.
3. Process the data with a computational algorithm that includes features to account for the following challenges and to yield the following types of results:
• Account for trends that are long-term relative to the length of the time periods used for the randomization. Trends include disease progression and recovery.
• Compute arrays of standardized (mean 0, standard deviation 1) benefit/harm scores that quantify the amount and strength of evidence for apparent treatment effects as functions of dose as well as for temporal parameters such as any delay and/or persistence of treatment effect.
• Summarize apparent treatment effect with respect to each dependent variable by selecting the most extreme positive or negative standardized benefit/harm score in an array. This yields a benefit/harm profile across dependent variables for each patient.
• Compute an overall benefit/harm score for each patient by averaging the dependent variable specific benefit/harm scores, which can be differentially weighted to account for differences in clinical significance, patient personal preferences, and social import of treatment effects.
4. Analyze patient-specific benefit/harm scores from two or more patients statistically to describe groups and make inferences from samples to populations. One single-group t-test on mean benefit/harm can be used to evaluate safety and effectiveness with respect to many dependent variables. Rejection in the positive direction would indicate benefit. Rejection in the negative direction would indicate harm.
Please note that the strategy just introduced enables single-group RCTs. Here are some advantages of single-group RCTs. They can:
• Save time, money, and patient resources during drug development by using more data for each patient compared, for example, to before and after treatment change scores. We can achieve higher statistical power with smaller numbers of patients. Diagnoses of health problems are becoming rarer and patients are becoming harder to recruit as diagnoses become more specific.
• Improve research ethics. No patient would be randomized to only placebo, only an inadequate dose, or only an excessive or harmful dose. We can identify optimal doses starting at the level of each individual. An optimal dose of zero means that the patient should not take the drug.
• Obtain reliable and valid measures of how individual patients respond. Reliable and valid benefit/harm scores can speed identification of genetic and other predictors of differential responses and optimal doses. Conventional RCT designs do not provide measures of treatment effect that are reliable and valid for individual patients.
• Integrate new gold standard methods of clinical practice with new gold standard methods of clinical research. This would help obviate the translation problem.
• Treatments for individual patients often can be justified or refuted scientifically.
Please note that the technical solution just outlined can help harmonize personalized medicine with comparative effectiveness research and help enable healtcare reform.
Personalized medicine is no longer primarily a technical problem.