The 3rd annual Clinical Genome Conference (CGC), held June 10–12 in San Francisco, CA, contrasted today’s limited utilization of genomic testing for disease with its expected potential. Speakers pointed out that business conflicts must be resolved if the technology is to meet society’s expectations. In contrast, technical issues, while also formidable, seem more solvable. Lecturers mapped out the battlefield with a balanced exposition and analysis of the conflicting positions of various stakeholders.
The fundamental conflict: America has a sick-care system, not a healthcare system.
The business of clinical genomics
The conflict between current practice and potential started with the opening keynote lecture by John Pfeifer, MD, Ph.D. (Washington University [WU], St. Louis, MO), who focused on the very restrictive constraints imposed by payers on the utilization of genomic testing.
Payers insist that an assay must be relevant to guiding therapy to qualify for payment. Their mantra: If there is no treatment, then why test for it? So they will not pay.
Another restriction: To qualify for payment, an assay must be based upon accepted science. It cannot be research in progress. It must be generally accepted. Even in genomics, “generally accepted” can take decades to establish, especially if the payers actively work against early adoption. In the meantime, people suffer and die. Investigational testing is also ruled out.
Asymptomatic testing of family members, even if they may be carriers, is usually excluded from payment. This includes prenatal testing for a known mutation in the family. In contrast with most clinical assays that focus on the sick, genomics can also predict predisposition spanning multiple generations. Payers exclude any other stakeholder, unless specifically mandated by law.
From Dr. Pfeifer’s lecture and others, it is clear that the payers (insurance companies) have adopted restrictive payment philosophy and rules that conflict with public health and welfare. They appear to be motivated by short-term cost reduction. Applying their rules to genetic testing precludes society from deriving most of the expected benefits of genomics, particularly when it can be applied predictively.
Dr. Jill Hagenkord1 of 23andme (Mountain View, CA) cited two examples of payer myopia where the FDA and payers are fighting on a common front: 1) Familial hypercholesterolemia (FH) affects 600,000 Americans. Current clinical testing misses about 85% of the affected patients. FH is caused by mutations in LDLR, APOB, LRLRAP1, and PCSK9 genes. Statins are effective in reducing coronary heart disease (CHD) in FH patients. Despite the fact that FH testing meets WHO criteria for screening, American payers will not consider paying for screening. 2) Hereditary cardiac conditions refer to genes that predispose a person to sudden cardiac death, often from arrhythmias. Treatment includes lifestyle changes, defibrillators, beta blockers, and other antiarrhythmic drugs.
Since the payers are not responsive, Dr. Hagenkord favors direct-to-consumer tests. Her employer offers exome sequencing for a large exome panel for under $100. Due to an ongoing conflict with the FDA over the issue of diagnostics, the test results are limited to “hereditary studies” in the U.S.A. However, the raw genomic information is provided to the consumer. She reported that several firms offer genomic interpretation programs and services independently. If patients are interested, they can send their raw data to the interpretation firms for evaluation of clinical risk factors.
How much will the payer pay?
Many of the genetic tests are so new that there is no generally accepted payment schedule. According to Dr. Pfeifer, it costs $7500 for Washington University’s gene panel. Payments vary, but WU usually receives payments in the range of $4000–$5000. To date, over 95% of the invoices are paid, but at a reduced rate. Generally, there are many inquiries from the payers about the clinical validity and utility of an assay. Payers refuse to acknowledge value or pay for “personal utility” of the information such as reproductive or life planning.
The new physician payment law, signed April 1, 2014, recognizes “Advanced Diagnostic Tests” (ADTs) for laboratory-developed assays. The payment codes and payment levels are not yet available. The expectation is that ADTs will have a three-quarter grace period where the lab can set and invoice at its list price without prior review. During the next three calendar quarters, the Centers for Medicare & Medicaid Services (CMS) will review the data on cost, complexity, value, and competition, and set a payment rate. Excess charges in the first three-quarters are expected to be subject to “claw back” provisions.
After the reimbursement rate setting by CMS, the payment law decrees a rate reduction of 15% per year in the first two years and then 10%/ year for the next three. This is a compounded reduction of 48% starting in the sixth year.
But who are the responsible payers representing? Cancer is multivariant and poorly understood. Consumers are individually insured. Many Americans pay the payers directly, or employ the insurance companies through intermediaries such as Medicare, etc. Payers should be our servants, not our masters. Restricting information-gathering and flow is inconsistent with an individual’s need to know in order to manage his or her life. Modern medicine is built around the 5 Ps—Predictive, Preventive, Personal, Participatory, and Philanthropic.1
What to do with “incidental findings”
Some genomic labs offer panels with more than 50 analytes, and others focus on offering a broader range of smaller, focused panels. Large panels increase the probability of incidental findings (IF), where one notices something that was not part of the target set or lab order. Many payers insist that test results be limited to relevancy to the presenting complaint. But frequently a focused request will be performed most economically as part of a larger panel. Or, in the case of whole genome sequencing (WGS), the interpretation report will contain “incidental findings” often for “variants of unknown significance” (VUS).
IF and VUS are not welcome by payers since they may lead to added expense. Indeed, several reported that payers favor labs that ignore IF and VUS. Others report that some oncologists desire reports responding to only what was specifically ordered, even if the assays leading to IF are run as a subset of a larger panel. Some clinicians want even a narrower report of only actionable diagnoses. Only a small fraction of clinicians self-report as being comfortable and confident with broad results from genetic assays, particularly whole genome sequencing.
Prof. David Magnus, Director of the Stanford Center for Biomedical Ethics (Stanford, CA), offered that failure to report IF could be a basis for medical malpractice litigation. As a patient I want to know about IF since some could emerge as serious as our knowledge of genomics expands.
However, not all patients feel the same way. Some want to protect other family members, especially progeny, from having genomic data freely available. Yes, discrimination on the basis of genomics is prohibited, but the nature of discrimination often makes detection nearly impossible. And, yes, data are often “de-identified,” but the protection is far from perfect.
The business segment of clinical genomics is in conflict. So how about the technical side? Conflicts abound here, too. But these are more interesting to chemists, and probably more solvable.
The technical side of clinical genomics
Exome vs whole genome sequencing
Dr. Pfeifer’s keynote was the first of many lectures to compare the cost and merits of exome sequencing, primarily with arrays on chips, with sequencing the entire genome (WGS). New technologies from several firms have reduced the cost of the analytics to $5000 or less, some much less. Interpretation is not included and can be much more expensive. Dr. Pfeifer explained that WGS has a cost advantage in his lab when the question at hand involves more than 2 kilobases. Below that, focused exome sequencing (ES) with arrays is lower in cost.
Dr. Elizabeth Worthey’s (Medical College of Wisconsin, Milwaukee) lecture presented a powerful, data-filled rebuttal. For her lab, they do WGS simply because it avoids the issue of “Are we missing something by using focused arrays?” She reported that with WGS, you will consistently find more, but this raises the issue of incidental findings above. She also explained that about 2% of patients at the Medical College present with more than one disorder. Without WGS, the clinical team would stop with the first positive result, and probably miss the second and subsequent. Dr. Worthey feels that patients and society are best served when full information is at hand. Some information may not be useful now, but she is confident it will be more useful as our experience grows.
She also reported that WGS has many advantages over whole exome sequencing (WES) in a clinical setting. WGS provides better coverage (~96%). This leads to discovery and explanation of rare diseases. One example involved a splice mutation that would have been invisible by WES. Plus, WGS is one protocol that standardizes laboratory workflow and avoids interruptions for updates, which was common with WES protocols. Turnaround time is about the same. Costs of WGS and WES are coming down rapidly. A group in Australia is now offering clinical-grade WGS for $1500 (including interpretation).
Prof. Jennifer Erwin of the Salk Institute (La Jolla, CA) presented a supporting lecture on mosaicism of single neural cells. Mosaicism is caused by rearrangement (transpositions) or later in development intracellular somatic insertion of DNA changes the sequence in cells. The rearrangement or insertion leads to diversity in propagating cells. Dr. Erwin used deep sequencing of single cells to study the location and structure of transposons. A study of five individuals showed the number of insertions varied from 10 to 40. Somatic insertions are rare, usually with one copy number. However, healthy individuals have a large number of somatic mutations, most of which are silent. It seems that transposons need supportive flanking sequences for activation. Since responsible variants can be positional or sequence, whole genome sequencing is essential.
Copy number variants of DNA segments are frequent in humans. In the past, it was suspected they were strongly associated with various diseases including some cancers, but other correlations with diseases have not been confirmed. Dr. Alex Kaplun of BIOBASE GmbH (Wolfenbüttel, Germany) lectured on detecting copy number variants (CNVs) with WGS. CNVs have been implicated in about 14,000 diseased patients. Compared to various probes, CNV with WGS is more reliable. Small CNVs stand out. Translocations and inversions are not ambiguous as they are with exome sequencing. BIOBASE curates data on CNV. Dr. Kaplun reported that about 50 software tools are available for CNV analysis.
Indeed, there is a general consensus that WGS is the best technical choice for clinical genomics. Cost is the only issue, and the differential is declining rapidly.
Benchmarking the WGS laboratory
Dr. Worthey provided some benchmarks for laboratory performance in whole genome sequencing. The lab started with exome sequencing, but in 2009, began shifting to WGS. Today, the lab “sequences, analyzes, and interprets a whole genome in under a week.” Typically, for a sample that arrives on Monday, the report is out on Thursday. The report included IF. The key is the HiSeq2500 (Illumina, San Diego, CA), which reduced sequencing, including verification, from 245 hr to 20. Interpretation is now complete in 3.5 hr, down from 7.5.
Reporting actionable incidental findings
Prof. Julianne O’Daniel described the policy of treating IF at UNC Lineberger (University of North Carolina at Chapel Hill). First, the lab uses a list of types of IF that may be reported to the physician or patient. Then, the age of onset is compared to the age at actionability. Once eligibility based on age is established, actionability is evaluated. The goal is to define the options for patients and parents. “Medically actionable” is defined as a disease that 1) exhibits a strong genotype–phenotype causative relationship, 2) has a serious risk of morbidity, 3) has existing guidelines for prevention and treatment, and 4) demonstrates success in improving patient outcome. Using the actionable criteria, one list has only 120 reportable variants.
Dr. Worthey reported that, when using WGS, the chance of IF is nearly 100%.2 But how many are actionable? Dr. Gail Jarvik (University of Washington, Seattle) pointed out that only 3.4% of European descendants had actionable pathogenic variants. The percentage for African Americans dropped to 1.2%. Both are small numbers, but this also probably reflects the state-of-the-art in oncology.
Data and diagnosis quality
As you might expect, data quality was a major concern, since genomics relies upon many external databases, generated by even more laboratories.
Dr. Jarvik addressed the conflicted clinical utility of genomics. Classification of tumors based on genomics was inconsistent, especially between cells of severity, particularly “pathogenic” and “likely pathogenic.” One study showed that of 128 variants, 52% were misclassified. Upon rereading, 63% changed classification, usually to a lower pathogenic class.
As part of the session on references and standards, Valerie Schneider of the National Cancer Institute (Rockville, MD) described the most recent build of the human reference genome, which was issued on 12/24/2013 by the Genome Reference Consortium (GRC) as GRCh38. This adds about 2% to the prior reference genome (GRCh37). In addition to cleaning up various items spread over nearly all chromosomes, Build 38 focuses on centromeres, which are difficult to sequence since they are rich with alpha satellite repeats. These account for about 2.6% of the human genome. The repeats are organized into higher-order repeats of unknown significance, but they are conserved, and hence probably important.
Databases contribute to uncertainty
Some caveats according to Dr. Worthey: Some of the “ultrarare” disorders are actually common. The choice of reference data is critical. Databases often fail to update changes in annotations. Dr. Worthey focused on RYR2 p.G1885E, which was first published in 2002 and identified as a polymorphism. It is now recognized as common, with a prevalence of about 5%. Dr. Worthey noted that recuration of databases is a huge, never-ending task.
Missing: Patient phenotype and identity
Another conflict is that only about 0.05% of the databases contain phenotypes, apparently for HIPAA and political correctness considerations. Omissions may also include identification of the disease understudy in the data set. Slow data sharing is a related problem.
Databases usually claim that the patient data are “de-identified,” but several lectures pointed out that phenotype or ethnic group is an important variable. Databases have other difficulties. Dr. Karen Eilbeck of the University of Utah (Salt Lake City) discussed problems with variant files and data. Files are not typed consistently (e.g., chromosome 1, chr1, or 1). Semantic inconsistences are common, e.g., the same word might have several meanings, or many words are used synonymously. Research databases are not compatible with health provider IT. Different assembly versions of the genome often correct errors in position, so one should be clear about which database version is used. These and other potential and often invisible inconsistencies make it difficult for diagnostic labs to exchange or compare interlaboratory results.
Credits
This report focused on the current state-of-conflict in adoption of clinical genomics. The conference cast a wider net, including exogenous RNA, differential expression and classification of cis and trans eQTLs (expression quantitative trait loci), standard reference materials, newborn screening, entrepreneurship, etc.
These topics added color to the main program, and showed that clinical genomics will be a much more complicated and information-rich topic than most envision today. The staff of Cambridge Healthtech Institute (CHI), in particular, Mary Ann Brown, deserve special credit for organizing a strong, multifaceted technical program and for managing the creature comforts that are essential for a great meeting.
References
- Hagenkord, J. “The changing definition of utility in the era of genetic abundance”; Lecture, Clinical Genome Conference, June 10–12, 2014, San Francisco, CA.
- O’Daniel, J. “Making meaning from genome sequencing”; Lecture, Clinical Genome Conference, June 10–12, 2014, San Francisco, CA.
Robert L. Stevenson, Ph.D., is Editor, American Laboratory/Labcompare; e-mail: [email protected].