Quick Take on Pharmacogenomics

an interview with Kristine Ashcraft

Introduction:  YouScript Background and  Experience

Background: YouScript is a clinical decision support platform that synthesizes evidence impacting drug response, including pharmacogenomics, to support doctors and pharmacists in making faster, more proactive decisions at the point of care or dispensing. YouScript has shown clinical validation in multiple peer reviewed studies confirming improved outcomes, reduced costs, and high patient and provider satisfaction. In addition, YouScript’s platform is integrated into the clinical workflow of Epic, Cerner, AllScripts, and Telus Health.

YouScript was recently acquired by Invitae, a company committed to making genomic information affordable and accessible to everyone. This aligns well with the YouScript mission to reduce costs and improve patient care by ending avoidable adverse drug events.

The Journal connected with Kristine Ashcraft, YouScript’s CEO prior to the recent acquisition. We discuss how precision  medicine  clinical tools can lead to fewer emergency room visits, hospitalizations, deaths and, consequently, lower health care costs by providing the safest drug and dose to patients that current evidence allows for.

Q1: YouScript’s website contains the story of a boy who died of an adverse reaction to his prescribed medication. Could you please talk about YouScript’s vision and mission in light of that story?

[see https://money.cnn.com/magazines/fortune/fortune_archive/2000/10/30/290610/.]

A. We are referring to an article that was in Fortune magazine in 2000 about a boy named Michael, who died of an overdose. As seen in one in twenty patients, Michael had a DNA variation, that in concert with his other medications, made the dose of one of his drugs unsafe for him. The drug levels were so high in his system that his foster parents were originally suspected of intentionally overdosing him.

Fortunately, a smart psychiatrist suspected that Michael had a pharmacogenetic variation and post-mortem testing led to his foster parents being cleared of homicide charges. The article talked about the dawning of the age of precision medicine when genetic information could be used to avoid worst-case scenarios like Michael’s, but also treatment failure and adverse effects that are not as dire.

At the time, I was working at Genelex, a lab originally formed as one of the first crime labs in the country with a National Institute of Justice Grant. They were looking for something that leveraged their core competency in providing high quality genetic testing. When the team heard Michael’s story, we went to PubMed and quickly learned how common these variations were and how many drugs they impacted That was the start of the journey to end avoidable adverse drug events, like Michael’s, that currently claim a life every two minutes in the U.S.

Q2. YouScript describes its Precision Prescribing software as the most advanced and innovative drug interaction software available. Could you please describe this platform, e.g., analysis methods, interface, input, output?

A. In the U.S., we have used the same binary drug interaction alerting system for decades; and the natural extension to this is binary drug-gene interaction alerting. We  now have over 50 million patients on five or more medications, a number predicted to double

by 2040, which creates a problem when the focus is on binary interactions. This outdated methodology, in the face of increasing patient complexity, is a large part of the reason we (in the US) spent $528 billion on non-optimized medications in 2016 – more than we spent on any major chronic disease.1

YouScript has multiple patents issued for multi-factor interaction detection that become more common and important as the number of medications increases. In a recent analysis of 100,000 patients with a major U.S. payer, multifactor interactions represented 12% of significant interactions in patients taking at least one medication. This is before adding the additional, quite common phenomenon of pharmacogenomic variants exacerbating

“We now have over 50 million patients on five or more medications, a number predicted to double by 2040”

the impact of drug interactions, such as phenoconversion – a complexity in which a patient converts to a different phenotype, such as a poor metabolizer even though they are  not so genetically. A study on phenoconversion run on claims in Austria found over 8% of the population was taking moderate or strong modulating drug for a metabolism pathway for another medication; again, this is prior to adding the further modulation of pharmacogenetics.2

YouScript has curated over 17,000 publications and product inserts for interaction and metabolism information warning of metabolic ‘traffic jams’ whether the cause is drugs, genes, herbals, or foods like grapefruit juice. Importantly, when issues are detected, we provide one click drug  alternates  by class or indication reducing interaction risk. We illustrate the YouScript Medication Management User Interface to our system in Figure 1 showing (l to r) Interaction report, ranked Pharmacogenetic Interaction Probability and the Risk Analysis Dashboard.

A common question is: how do I know which patients are at risk and who should be tested? YouScript also provides both patient and population level risk analytics that can be seamlessly integrated into the EMR or pharmacy workflow. The analysis shows the number of existing interactions by severity so pharmacists know where to focus attention first. Additionally, our patented pharmacogenetic interaction probability (PIP) score stack-ranks patients most likely to benefit from pharmacogenetic testing by computing the probability that testing will result in at least one, otherwise unrecognized, significant pharmacogenetic interaction.

How does YouScript integrate patient and family history in its platform?

A. YouScript currently does not integrate this information, although we track all interaction detection and medication changes over time. By so doing, we can start learning from dose modifications and discontinuations that are not explained by current knowledge. Now that we are part of Invitae, we are considering incorporating our risk analytics into our telehealth application, GIATM, as well as our patient and provider portal. This will enable patients to discuss the benefits of pharmacogenomics as well as other genomic testing for hereditary cancer, reproductive risk, or proactive screening, as appropriate, with their healthcare provider.

Q3. YouScript’s process starts by first testing patients to determine who might benefit from the YouScript platform.  Can you describe the algorithm of stack-ranking of patients based on the preliminary screen – e.g., type of data review? How was this ranking validated?

A. The pharmacogenetic interaction probability (PIP) score only requires the patient’s current drug regimen. We then determine if a pharmacogenetic variant would result in a new, evidence-based, significant interaction where drug or dose changes should be considered, and what the prevalence of that variant is. Next, we calculate the overall probability that testing will result in at least one evidence-based, moderate or major pharmacogenetic interaction typically associated with specific dosing or drug change guidance.

We validated this score in partnership with the University of Utah’s Pharmacoeconomics team. They ran our PIP score against claims and clinical data for several hundred thousand 65+ year old patients taking multiple medications. As the PIP score climbed for patients otherwise held equal by age, gender, Charlson Comorbidity Index, and known drug interactions, so did the likelihood that patients would end up in the ED or hospital.

In a Highmark BCBS VITAL innovation program, our PIP score correlated with the percentage of significant pharmacogenetic interactions that were found at testing, so we also know it works as intended! More recently, we completed an analysis of three years of claims data for 100,000 patients with a major U.S. payer, and found our PIP score was better at predicting future ED and inpatient visits than other indicators, such as diabetes with complications, renal failure, and historic count of inpatient claims as shown in Figure 2.

Is YouScript finding that the spectrum of patients can be expanded to include more who might benefit? That is, have you been able to identify more patients who might benefit from the platform?

A. We did recently expand the PIP score to include more pharmacogenetic variants and will continue to do so as the evidence evolves. I believe we will get to a point where testing is ubiquitous and available before a patient is prescribed a non-optimal drug or dose. In the meantime, we help payers and providers know which patients will benefit most so they can focus resources where they will have the most positive current impact. By way of illustration, we show in Figure 3 a graph outlining the correlation between YouScript PIP score and increased healthcare spend in claims analysis with a major U.S. payor.

“We did recently expand the PIP score to include more pharmacogenetic variants and will continue to do so as the evidence evolves”

If so, is this because more data are available for the ranking algorithm? Or because a broader patient range is tested?

A. As the evidence evolves, the number of significant pharmacogenetic interactions climbs as well, and with a larger, more diverse patient data pool, we can better assess the likely impact of immediate care with our approach.

Q4. YouScript offers a Clinical Decision Support service as part of its program to aid healthcare providers to “reduce trial-and-error prescribing, reduce expenditures, and improve patient outcomes,” all of which can result in overall lowered healthcare costs. Can you tell us about the learning curve of YouScript’s experience with healthcare providers – e.g., compare the experience with early and later adopters, respectively?

A. Once we demo the software, it is incredibly easy to understand and use. I believe the bigger hurdle is overcoming genetic exceptionalism; many providers hold a higher bar of evidence for genetic impacts than other factors. As I mentioned in reply to Q3, a high PIP score is more likely to predict a future ED visit or hospitalization than renal failure as an indicator alone, but no one questions renal function testing despite the lack of large randomized studies to determine predictive value for each medication cleared by the kidneys. Pharmacogenomics essentially looks at your inborn liver function for metabolizing medications. We offer prescription guidance assuming a patient is a normal metabolizer for different liver enzymes when this is typically not the case.


 

By way of example: Think of this like assuming patients have two lanes of a ‘highway’ open in the liver. Based on genetics, however, more than 95% of patients have only one lane, no lanes or three or more lanes of one or more highways open. Other drugs, herbals, and foods send signals to the body to close lanes, open lanes, or they may just take the entire highway for themselves!  Now, if you know that a highway you typically drive is shut down or congested, you drive a different way (if you are able). In fact, the app on your phone  provides  alternatives that help you avoid traffic jams.

YouScript brings this same technology leap to medication management. Early adopters quickly understand it is an improvement over the ‘paper maps’ they used to use to navigate the complexities of medication management. Late adopters tend to wait for more evidence to appear, and the laggards will not adopt until it is mandated.

What have you (and YouScript generally) learned from these experiences of introducing the YouScript  platform to your current experiences – e.g., What  kind of data do healthcare professionals find the most compelling?

A. Peer reviewed studies of outcomes showing that YouScript significantly improves patient safety and reduces healthcare costs certainly help. Studies where polypharmacy patients identified as at risk and treated according to YouScript showed:

In a 110 patient randomized controlled trial in 60-days in home health:3

  • 52% reduction in readmissions (p=.007);
  • 42% reduction in ED visits (p=.045);
  • 85% reduction in mortality (p=.05);
  • Estimated savings of over $4000 per patient.

In a 1025 patient trial in 120-days in ambulatory care:4

  • 71% reduction in ED visits (p=.0002);
  • 39% reduction in hospitalizations (p=.0273);
  • Estimated savings of over $1000 per patient.

In a 342-patient trial in telephone Medication Therapy Management, in the pharmacogenetics and YouScript arm:5

  • A 6.3X increase in detection of serious drug therapy problems (p>.001);
  • A 2.4X increase in prescriber acceptance of changes for these serious drug therapy problems.

I think the realization occurs when a healthcare provider sees dramatic improvements in patients whose drugs are optimized with YouScript and pharmacogenomic testing – a patient whose Alzheimers-like symptoms resolve, a patient who had a potentially avoidable 2nd stroke on clopidogrel, or a patient who finally finds the right antidepressant. I’m a big fan of data, but the “stickiness” happens when a healthcare provider sees the real impact on the patients they care for directly.

Has YouScript gained any experience working with genetic counselors to use this platform for their client patients?


A.
Our super-users are pharmacists, but genetic counselors can certainly use YouScript to determine when pharmacogenomic testing may be advised. With our recent acquisition by Invitae, we plan to make these tools more available to the genetic counseling community as understanding genetic variation and how it can impact optimal drug therapy is key in managing all disease progression, especially cancer and genetic disorders.

Data Sources

Q5. Patient data (as well as other data sources) are often sealed away in so- called siloes. Has YouScript been able to negotiate access to those siloes?

A. With the advent of SMART on FHIR, YouScript has been able to seamlessly access the patient data needed. We look forward to CDS (clinical decision support) hooks being more widely available, so we can migrate away from the custom APIs currently needed to guide decision-making at the right time in the workflow. We are already overcoming the loss of pharmacogenetic results in a PDF by storing the data discreetly; like allergies, pharmacogenetics needs to be reevaluated every time a medication decision is made the rest of a patient’s life.

We currently securely store over half a million pharmacogenetic test results, with patient permission, we can share these results through the integrated or web-based version of YouScript. We currently integrate with Epic, Cerner, and Allscripts and have the scalable capabilities to integrate with any system storing a patient’s medication regimen.

Has YouScript generated its own data through collaborations? Partnering? Licensing?

A. YouScript has a substantial amount of internal data and analytics capabilities and is always open to collaborations and partnerships seeking to further understand pharmacogenomics and its role in precision medicine. For example, we are currently working with the University of Florida and GatorCare on  a comparison of our advanced analytics versus more simplified approaches. Both YouScript and UF are members of the IGNITE network and help develop and share methods for incorporating genomic information as a routine part of patient care in diverse clinical settings.

Regulations governed by Privacy Rules (as contained in the Health Insurance Portability and Accountability Act (HIPAA) may limit access to data. How does YouScript manage its access to data? Are all data de-identified? Or otherwise secured?

A. We are HIPAA-certified and adhere to best practices for encryption and de-identification.

“We typically see an increase in Per Member Per Month (PMPM) costs overall and for ED visits and hospitalizations as our analytics show an increase in PIP and interaction detection”

Q6. Does YouScript plan to incorporate other data types, e.g., other ‘omics data?

A. We plan to incorporate data elements that will enable us to better optimize drug and dose selection if and when it becomes clear that the evidence makes the data clinically actionable.

Will YouScript expand the platform to other therapeutic areas?

A. YouScript covers all FDA-approved medications across all therapeutic areas already.

Q7. Has YouScript analyzed the patient metadata (beyond strictly medical data) to see how different population segments might best benefit from the platform – e.g., age, ethnicity, geographic location? Are there “low-hanging” fruits? That is, are there patients who benefit significantly at a reasonable cost of the YouScript service?

A. Patients taking multiple medications with pharmacogenetic risk benefit the most. You tend to see this more often in patients over 65 years of age or geographical areas with a high number of patients with multiple comorbidities. Psychiatric, cardiovascular, pain, and oncology medications are typically included in these drug regimens and many are impacted by pharmacogenetic variability.

What cases have you found to be the most difficult to rank?


A.
We typically see an increase in Per Member Per Month (PMPM) costs overall and for ED visits and hospitalizations as our analytics show an increase in PIP and interaction detection. The changes are not as pronounced in patients with chronic heart failure.

Summary/Lessons Learned

Q8. Lessons learned/Summary remarks?

Lessons learned from twenty years in pharmacogenomics:

  1. Lack of widespread pharmacogenetic reimbursement is the primary issue preventing many health systems from deploying a pharmacogenetic and clinical decision support.
  2. Static pharmacogenomics PDF testing reports are quickly outdated and results need to be stored discreetly so dynamic CDS can be provided based on a quickly evolving knowledge base.
  3. Polypharmacy patients with high PIP scores benefit the most, especially Medicare
  4. Medications most commonly impacted are those used in cardiology, psychiatry, oncology, and pain Expanding coverage to gene panels rather than drug-gene pairs or disease states will decrease cost and reduce avoidable patient harm. Laboratories are running gene panels for efficiency and overall cost savings already.
  5. Many patients are interested and willing to pay for testing, but to provide clinical utility, their healthcare providers need ongoing clinical decision support access and resources they can trust.
  6. Pharmacist-led initiatives yield best results; this is highest when there is an existing pharmacist/prescriber/patient relationship.
  7. Pharmacists are ideal for initial medication regimen “right-sizing,” but for ongoing optimization, EHR integration is preferred so that YouScript can intervene before the prescription is finalized.
  8. When creating evidence, comparison groups need to be risk-matched whether comparing the same patients before and after intervention or when using a treatment as usual control.
  9. Our data suggests that de-identified data can help us find and prevent currently unknown causes of adverse drug events.

 

 

 

Kristine Ashcraft, BS, MBA is a molecular biologist by training and is the former CEO and founder of YouScript, recently acquired by Invitae. She has worked in pharmacogenomics since 2000 and was recently named one of the 25 leading voices in precision medicine. Kristine has authored multiple publications on the clinical and economic benefits of pharmacogenomic testing including one lauded as one of the most  influential publications at an American Medical Informatics Association meeting. She has been interviewed by numerous media outlets including the New York Times, the Wall Street Journal, and NBC Nightly News and has spoken at SXSW, American Society of Human Genetics, and numerous precision medicine conferences. She is committed to catalyzing the adoption of precision medicine.

 

 


References

  1. Cost of Prescription Drug-Related Morbidity and Mortality, Watanabe JH, McInnis T, Hirsch Ann Pharmacother. 2018 Sep;52(9):829-837. doi: 10.1177/1060028018765159. Epub 2018 Mar 26, https://www.ncbi.nlm.nih.gov/pubmed/29577766
  2. The Importance of Gene-Drug-Drug-Interactions in Pharmacogenomics Decision Support: An Analysis Based on Austrian Claims Data, Blagec K, Kuch W, Samwald , Stud Health Technol Inform. 2017; 236:121-127, https://www.ncbi.nlm.nih.gov/pubmed/28508787
  3. Clinical impact of pharmacogenetic profiling with a clinical decision support tool in polypharmacy home health patients: A prospective pilot randomized controlled trial, Elliott LS, Henderson JC, Neradilek MB, Moyer NA, Ashcraft KC, Thirumaran , PLoS One. 2017 Feb 2;12(2):e0170905. doi: 10.1371/journal.pone.0170905. eCollection 2017, https://www.ncbi.nlm.nih.gov/pubmed/28151991
  1. The effect of pharmacogenetic profiling with a clinical decision support tool on healthcare resource utilization and estimated costs in the elderly exposed to polypharmacy, Brixner D, Biltaji E, Bress A, Unni S, Ye X, Mamiya T, Ashcraft K, Biskupiak J, J Med Econ. 2016;19(3):213-28. doi: 10.3111/13696998.2015.1110160. Epub 2015 Nov 11, https://ncbi.nlm.nih.gov/pubmed/26478982
  2. Clinical Utility of Pharmacogenetic Testing and a Clinical Decision Support Tool to Enhance the Identification of Drug Therapy Problems Through Medication Therapy Management in Polypharmacy Patients, Kim K, Magness JW, Nelson R, Baron V, Brixner DI, J Manag Care Spec 2018 Dec;24(12):12501259.doi:10.18553/jmcp.2018.24.12.1250, https://www.ncbi.nlm.nih.gov/pubmed/30479202