By Allison Proffitt
October 14, 2013 | Vanderbilt University’s PREDICT program is putting genomes to work in the clinic, using drug-gene interaction data to suggest the right drugs for cardiology patients. The biggest challenge, though, might lie in predicting people’s behavior.
The PREDICT program is one of several personalized medicine initiatives at Vanderbilt overseen by the Associate Vice Chancellor of Personalized Medicine. John Peterson, an Internist and Assistant Professor of Biomedical Informatics and Medicine at the Vanderbilt University School of Medical, has been tasked with overseeing and evaluating the impact of the PREDICT program on clinical medicine.
The PREDICT program currently covers five drugs. Patients are genotyped using a VeraCode ADME Core Panel from Illumina (184 variants in 34 genes) on the BeadXpress platform. All the testing occurs in Vanderbilt's Molecular Diagnostic Laboratory under CLIA conditions.
PREDICT attempts to identify at-risk patients before they need any drugs; testing early ensures that patients’ genomic data is on file when drugs are needed.
“In some cases, they don’t have any disease or prescription that we’re particularly concerned about, but we genotype them in anticipation that they will need a prescription that’s within the PREDICT program within the next three years,” says Peterson.
Peterson describes the process as “very IT-heavy.” A risk score is generated for each patient based on their health history. “The risk score is driven by some key diagnoses, things like, ‘Does the patient have heart disease?’ and ‘Does the patient have diabetes?’ and that is tied to the fact that three of our pipeline drugs are Warfarin, Simvastatin, and Clopidogrel. Those have a lot of cardiovascular indications. It’s very common that a patient that has cardiovascular risk factors or already has a cardiovascular disease ends up on one of those drugs.”
Patients can be genotyped when it is clear that they need a program drug, but the advantage of preemptive genotyping is that clinicians have immediate access to the genomic data as soon as they need it. The findings are stored in Vanderbilt’s electronic health record system, and the data is brought to the attention of physicians when prescribing takes place. The disadvantage, of course, is that it’s not certain which patients will need the information.
The goal of the PREDICT study is largely to consider physician engagement: how clinicians can practically use genomic data.
The question, Peterson says, is: “First and formost, are clinicians using this information to tailor the prescription? They are, but it’s certainly not for every patient. We realized in the very beginning that genetics is a component of personalized medicine, of prescribing in general. You have all these other issues of cost and tolerance for a particular dose or drug. We inform [clinicians] about the genetics and we give them specific recommendations about what they could do with the genetics.”
Peterson expects the program to expand in terms of genes and drugs considered, but said the process of new implementation is an involved one.
“Every drug-genome interaction is so unique in how it’s implemented. It involves a unique set of clinicians. There’s a different set of non-genetic factors that you’re incorporating into your algorithms, and the science can be fairly unique with each of these pairs,” he says. “Each implementation is carefully examined in terms of the evidence: who might benefit from this particular implementation, and for some of the newer ones that are more niche drugs, it’s a much smaller model.”
The program has so far genotyped 14,000 patients, and most have had actionable findings. “Once you get up to five or six genetic variants, the majority of patients—in our case about 87%—have one or more of those variants. That’s potentially actionable.”
Though the goal of PREDICT is primarily to look at clinician responses to the data, patients and their responses to the process are still considered.
“Patients are a key member in terms of whether they get genotyped in the beginning—whether they agree to it, and also whether they are willing to take some of the risks of being charged for genotyping.”
At the beginning of the PREDICT study, Peterson says, no patients paid for their own genotyping. That is about to change, though, as the program is just starting to wade into the waters of reimbursement and patient payments. It’s been hard to get straight answers from all the payor companies, Peterson says, but he expects the details to be sorted out soon.
No bills have been rejected in the past several months of billing, he says, and there has been no cost to patients. “But I’m sure at some point someone is going to get charged something. But predicting who and what—that’s going to be hard.”
The results for 28 of the genes tested are sequestered, Peterson says, because they don’t have well-defined clinical validity. Findings for six of the genes are released to patients and clinicians pertaining to five drugs or classes of drugs.
“Patients get a plain-language description about what their genetic results might mean,” says Peterson. The description is available on the patient portal: www.MyHealthAtVanderbilt.com and patients get verbal results from their clinicians. “We find that patients do struggle to understand what the genetics might mean,” Peterson says.
The portal system does not reach all patients, Peterson suspects. Patient education tools are an area Peterson hopes to develop over time.