NEWYORK (GenomeWeb) – Personalized medicine proponents have long argued that pharmacogenomics can increase drug efficacy and reduce dangerous adverse outcomes, especially if PGx information is collected and stored preemptively so that it can be available at the point of prescribing and useful in the timeframe of routine clinical care.
Ahead of what many hope will be growing adoption of this version of PGx over the next several years, researchers from several early adopter institutions have published a review of five pioneer programs online ahead of print in next year's Annual Review of Pharmacology and Toxicology.
The review touches on common elements among these early adopters and highlights challenges and some solutions to broader adoption and implementation, including the importance of robust and systematic, but also flexible, processes for integrating genetic test results into electronic health records, and enacting clinical decision support (CDS) for relevant gene-drug pairs.
The group's report discusses five preemptive PGx programs, which have published accounts of their strategies, and in some cases early results, in the scientific literature.
St. Jude Children's Research Hospital began in May 2011 a clinical research protocol called PG4KDS, under which it has aimed to recruit all critically ill patients at the hospital to be preemptively genotyped using an array-based PGx test covering several thousand variants in 230 genes. As of January 2014, four genes and 12 drugs have been migrated into the hospitals electronic health record system, and with these gene-drug pairs implemented, 792 patients (78 percent) have had at least one actionable phenotype. The group plans to expand the system and add all gene-drug pairs included in guidelines by the Clinical Pharmacogenetics Implementation Consortium, or CPIC.
Vanderbilt University Medical Center began a preemptive PGx project, PREDICT, in September 2010, with an initial focus on antiplatelet therapy. Providers in the program can enroll any patient, but the focus is on groups with anticipated future cardiac catheterization. Patients in the program are genotyped for 184 variants in 34 genes. As of November 2013, 10,000 patients have been genotyped in the program, and CDS for several gene-drug pairs has been implemented, including CYP2C19 and clopidogrel; SLCO1B1 and simvastatin; CYP2C9, VKORC1, and warfarin; and others. Ninety one percent of patients have had at least one actionable genotype, according to the review authors.
The University of Florida launched its preemptive PGx program in 2011, using chip-based genotyping covering 256 SNPs. Test results pertaining to approved gene-drug pairs, currently only CY2C19 and clopidogrel in this particular program, are added to patients' medical records. According to the review authors, 800 patients in the program have had results moved to their EHRs as of March 2013.
At the Mayo Clinic, a program in collaboration with the Pharmacogenomic Research Network and the Electronic Medical Records and Genomics Network, eMERGE, has been using a targeted next-gen sequencing panel to preemptively collect PGx information, with the results stored and able to trigger alerts to future prescribing physicians. A total of 1,013 patients were enrolled as of July 2013, according to the review authors. Four gene-drug pairs have been approved for implementation and several more are in the CDS development process.
Mount Sinai, another member of eMERGE, initiated its CLIPMERGE PGx program in February 2013, using an existing biobank as a study cohort and developing CDS for several drug-gene pairs to deliver alerts to prescribing physicians at the point of care.
Mary Relling of the St. Jude's Department of Pharmaceutical Sciences and and Mark Dunnenberger of NorthShore University Health System in Chicago, two of the review's authors, told PGx Reporter in an interview that though the going has been rough in some ways for these early pioneers, the hope is that the tools and strategies they are developing will grease the way for other health systems to more easily implement preemptive PGx protocols of their own.
Below is an edited transcript of the interview.
The idea that at least some pharmacogenomic information needs to be collected preemptively to be useful in a clinical timeframe has been around for a while, but from your review it seems like this is only in the early stages of real adoption so far. Can you tell me a little about some of the implementation barriers you and your coauthors have identified?
Relling: The barriers for a long time have been mostly the cost and limitations of molecular testing itself. It's probably only in the last few years [that] it's become very inexpensive to accurately test the vast majority of functionally important variants in the top pharmacogenes, so that has been a huge barrier to come down. Of course there are still many, many barriers that are pretty much just inherent in most healthcare systems in the US, mainly that reimbursement is set up to reimburse procedures or take care of symptoms, not to prevent symptoms. There's no one to take responsibility for the cost of doing preemptive genotyping and carrying that information along in the healthcare system because of the fractionated lack of universal healthcare we have.
Another problem is a fragmented EHR system or no EHR system at all for many healthcare systems. With preemptive PGx, we have the beauty of doing a one-time lifelong genetic test, but in practice there's no way for that test result to actually follow the patient through all their myriad associations with healthcare providers.
Dunnenberger: I can't state enough how difficult it is to relay this information to multiple healthcare systems. There is no real systematic codified method to communicate pharmacogenomic results. Most of the time when you get them now, they are a scanned PDF and not very functional in an electronic system. That missing piece of how to do this in the EMR is really important, and I think we are making strides to do that, with CPIC and CPIC informatics, and others like the eMERGE network. We are all trying to come up with ways of doing this and communicating this information, and that's what's really been holding us back.
In your review, you look at some of the first adopters of preemptive PGx efforts. Generally, how does it look like these first programs have been going?
Relling: I think probably so far it's been more work and slower than many of us anticipated it would be, because we are early adopters and there aren't existing agreed upon terminologies, phenotype definitions, or off-the-shelf clinical decision support systems available. I would say each one of our five systems are developing this sort of de novo – trying to, as Mark said, enhance the CPIC guidelines to include more concrete example language for PGx clinical decision support and alerts – but there is no way to get that off the shelf right now.
And with electronic health records so fragmented and disparate among sites, it takes person power to implement the clinical decision support you need into the record and alert physicians appropriately at the time of prescribing when there are relevant genetic test results. We were excited when the cost of testing went down to essentially negligible, but the cost associated with all the informatics is higher than we anticipated. However, this is being borne mostly by early adopters, and as time goes on hopefully there will be more off-the-shelf systems. Not every healthcare system is going to have to reinvent the wheel, reinvent the language, reinvent the systems, and its going to be less and less expensive and easier for other sites to do this in the future.
Tell me a little more about the effort at St. Jude specifically. How did you transition to this model? How many genes are included? And how are kids/families recruited to participate?
Relling: St Jude Children's Research Hospital is very special in that we are a center where we treat children with catastrophic diseases and assume complete responsibility, including financial responsibility, for the healthcare of those patients. That allows us to have a comprehensive healthcare record that covers almost all problems for each of our patients, and that is penetrant through inpatient, outpatient, and clinic care through their lives. Our goal is to enroll every patient with a catastrophic disease at St. Jude into our preemptive PGx research protocol, PG4KDS. We screen using the DMET array plus some added gene-specific tests for about 230 pharmacogenes, all generated in a CLIA lab setting. The goal is to migrate results from the array into the healthcare record for every patient as we develop clinical decision support to help physicians use the genetic test result for prescribing decisions. Our goal is to implement all the CPIC genes – currently there are 13 – and we have implemented clinical decision support for five genes so far affecting 14 drugs.
Is this all conducted in a research or fully clinical setting?
Relling: It's fully patient care, but we are doing this implementation via a research protocol. The goal is someday for it to just be part of everyday practice, but because of lingering concerns about genetic testing and our desire to make sure we had a very organized process to the procedure we elected to do it via a research protocol. Not all the five sites in our review are considering this type of implementation; some are doing it just as best clinical management.
Back to the broader picture — were you able to glean from your analysis of these five early adopter programs any important lessons in terms of best practices, or aspects that have been particularly successful? What about cautions or lessons on what to avoid?
Dunnenberger: I think there are two things that are really, really important. One is having good institutional structure to govern pharmacogenomic testing however you are doing it. All these sites we reviewed, or most at least, have had pharmacogenomics committees of some sort that report on up through the chain of the hospital system. I think that's really crucial, to provide legitimacy to your program and to get buy-in from everyone who has a say so in practice at your hospital, so that when a clinician sees this, they know that the whole hospital agrees. Even if they had resistance on their own level, they know it's institutionally supported.
The other thing that is important is selecting an appropriate patient population for your hospital. As we are doing this, it would be great if everywhere was St. Jude, and we could genotype every patient that came in our doors, and we knew that we could use this information over and over again, but unfortunately in the real world that's just not the case. So you need to be picking out patients that are going to be receiving multiple of these drugs over a long time period and coming back to your health system. At North Shore, we have both inpatients and outpatients, so we can provide this information both for acute care and longitudinal care with their primary care physician's office. Being able to use one test in a preemptive manner to change multiple therapeutic decisions over a patient lifetime is what makes this type of approach economical.
I know that there are projects looking now at collecting preemptive PGx information from even broader technologies, perhaps even whole-exome or whole-genome sequencing. Do you see that as part of the near future for these efforts or not?
Relling: At St. Jude we are embarking on a pilot project to see how whole-genome sequencing can or cannot be used for clinical pharmacogenomics purposes. There are a lot of centers, including some of these five we reviewed, using a platform, PGRN-seq, which is NGS-based and targets 82 or 84 genes. But I think it's still a little early to say how NGS is going to compare to more variant-specific genotyping approaches.
Where do we go from here? Are we likely to see the efforts of these early adopters replicated more broadly in the near future?
Relling: I think this is spreading, but still kind of slowly because we are still at a stage where it takes quite a bit of effort to do the implementation and many sites struggle with the financial complications of how to do this.
Dunneberger: Every meeting I go to there are more and more people interested in doing this and on the cusp of trying to get something going. Not many are ready to make the leap into the pool, but I think that number is growing. Cost is a big concern; how to do this inside the EHR is a big concern. But five years from now, I really do think most large institutions will have some support for preemptive PGx in some way. I also do think we will be headed toward NGS as the avenue for where we get our PGx information — getting the genetic information for some other healthcare reason and then having the PGx genes there ready to be used is how this is going to make sense in the future.
Relling: I'll add that I think one thing this paper also contributes that is a little unique is our data on prescription drug use in the US. With our collaborator Dr. Hunkler, we had access to a unique resource for numbers on outpatient prescriptions in the US. If we take the top 30 drugs with a high PGx risk, it's over 738 million prescriptions per year in the US. That's not the majority of drug use, but it's a pretty big chunk that is potentially affected by CPIC pharmacogenes that we can test for right now.
If how to pay for these programs and reimbursement are still lingering questions, is there some expectation that some of these early adopter programs might help demonstrate the clinical utility of cost effectiveness of preemptive PGx toward that end?
Relling: My opinion is that for these CPIC genes and drugs, the data is already there in terms of clinical utility, we are just not acting on them. Cost effectiveness is a whole different argument that you can talk till you're blue in the face because costs are constantly changing, but there is no question that there is clinical utility for testing for these genes and using the results to prescribe drugs more wisely.