Reconciling Opportunistic and Population Screening in Clinical Genomics

Abstract

Opportunistic genomic screening is becoming increasingly common as laboratories adopt recommendations to report secondary genomic findings. In parallel, interest in using genome sequencing as a population screening test has grown rapidly. We consider here three potential applications of genome sequencing for preventive medicine: 1) provider-ordered predispositional testing in healthy adults, 2) indication-based testing with opportunistic screening of secondary results, and 3) population screening in the public health context. We conclude that despite superficial similarities, there are important and fundamental differences in the way medical risks and benefits can be addressed in these three contexts. Recommendations to report secondary genomic findings should not be interpreted as an endorsement of population genomic screening.

Introduction:

Interest in using genome sequencing in healthcare is growing rapidly. Genome sequencing technologies, including whole exome sequencing and whole genome sequencing, are increasingly being used in clinical settings to diagnose patients with rare and undiagnosed conditions¹ and to personalize cancer treatment.^{2, 3} As sequencing may yield incidental findings of medical importance, formal recommendations were issued in 2013 by the American College of Genetics and Genomics (ACMG) that patients who have undergone sequencing for one clinical indication should be offered secondary results for other actionable conditions.^{4, 5} These recommendations were reinforced in an update that revised the gene list, but upheld all of the central principles of these recommendations.⁶ Most clinical laboratories performing genome sequencing in the U.S. now report returning secondary findings based on these ACMG recommendations, performing opportunistic screening for actionable risk variants that may inform disease prevention.

At the same time, a number of companies, practitioners, and academic medical centers have begun to offer sequencing to healthy individuals or to research participants with the logic that screening for at least the ACMG list of actionable conditions may be beneficial.^{7, 8} This trend has been driven by an emphasis on medical care that is proactive rather than reactive, by the exuberant approach of companies seeking to expand the market for genetic testing, and by emerging evidence that some mutations in the general population can be associated with later development of related phenotypes.⁹ Nonetheless, genome sequencing has not been widely adopted as a screening test, and population screening of healthy individuals with sequencing remains controversial.^10–14 Population screening with sequencing is neither recommended nor proscribed by the ACMG,¹⁵ and it is not among the preventive screening measures recommended by bodies such as the U.S. Preventive Services Task Force.¹⁶

These conflicting perspectives have understandably created confusion. If experts recommend offering patients secondary results related to conditions for which they have low prior probability when they undergo sequencing for another purpose (opportunistic screening), then why shouldn’t healthy individuals be offered genome sequencing as part of preventive healthcare (population screening)?

The most important explanation for this distinction is the fundamental difference in the way medical risks and benefits are addressed in opportunistic versus population screening. In order to clarify this difference, we will examine three potential applications of genomic technologies to preventive medicine: 1) provider-ordered predispositional testing in healthy adults, 2) indication-based testing with opportunistic screening of secondary results, and 3) population screening in the public health context. Since these three applications share a superficial similarity, the use of genomic testing to guide preventive healthcare, it is not surprising that many have interpreted the ACMG recommendations on secondary findings as an endorsement of genome-guided screening in all of these contexts. We suggest, however, that these three applications of genomic technologies are not as similar as they first appear. Each takes place in a particular setting where context-specific strategies are used to mitigate risk and maximize benefit.

Mitigating Risk and Maximizing Benefit

Like other medical interventions, clinical tests – whether diagnostic or preventive – involve some risk of harm. The most conspicuous harms from clinical tests are false positive results that may lead to actions or procedures that may cause unintended morbidity and mortality. The risk for false positive results is particularly high in genomic testing. The extremely large number of sites interrogated with exome or genome sequencing increases the statistical likelihood that one or more analytic false positive results will be generated.¹⁷

But even if all variant calls are confirmed by an orthogonal technology as some have recommended, genomic variants may be interpreted inaccurately or they may be interpreted accurately but the disease condition may never manifest. Rare or novel genomic variants occur commonly in human genomes,¹⁸ and may initially be classified as pathogenic only to later be proven benign.^19–21 Newer large-scale population-based sequencing data are demonstrating that the penetrance of many common pathogenic variants is lower than initially estimated.²² In practice, then, many pathogenic variants will be returned to individuals who will never develop the medical condition.

False negative results are also relatively common in genomic tests and raise important concerns. Many genes contributing to risk for specific conditions have not yet been identified, and the pathogenicity of many variants within genes that are clearly associated with disease is often uncertain. In addition, current analytical pipelines can miss pathogenic variants due to structural variation or changes in regulatory regions.²³

These potential blind spots contribute to a well-known challenge in clinical genetics: false reassurance. There is a dramatic difference between telling a patient they are not at risk for a condition and telling them no genetic factors that increase their risk were identified, but the distinction between these messages may be muddled by providers or misunderstood by patients.^{24, 25} As a result, patients can be left believing that are not at increased risk for a condition, or even that they are entirely free of risk for that condition. This false reassurance may cause patients to forego other types of screening that would normally be recommended, including mammograms and colonoscopies.²⁴ If this happens, they might have been better off having never had a genomic screen.

These factors remind us that genomic tests are far less deterministic than is generally believed and therefore not exceptional, but that they reflect the common clinical and public health challenge of balancing medical harms with medical benefits.²⁶ Fortunately, experience in other domains of screening has led to the development of effective practices that can be used to maximize benefits and minimize risks when performing genomic testing.

In the previous era where genetic testing was scarce and expensive, the most important of these strategies has been the careful selection of which patients would undergo testing on the basis of signs, symptoms and family history. When clinical tests are used in patients with relevant signs and symptoms (indication-based testing), there is an increased likelihood that positive findings are accurate. In other words, when the prior probability of a positive result is high, the rate of false positive results is correspondingly low.²⁷ But in the modern era where genetic testing is becoming inexpensive and abundant, and will increasingly be used for screening, new strategies will be needed. For example, detection of risk variants may need to be followed up by iterative attention to focused physical examination or diagnostic testing (socalled deep phenotyping), or to family history, since we now know that unanticipated findings often prompt unexpected detection of both.^{28, 29}

A second strategy to help mitigate risk relies on the context where testing takes place. Indication-driven genomic testing and provider-ordered preventive testing both take place in clinical contexts. This setting is typified by the opportunity for a shared decision-making process between patients and providers. Before testing, this type of decision-making process allows patients and providers to carefully consider how testing would address, or fail to address, the needs and circumstances of the patient. Once results are returned, shared decision-making allows providers and the individuals receiving genomic results to make a considered and well-informed response to these findings accounting for the possibility of false-positives. This one-on-one relationship provides a number of other benefits that potentially increase the value and decrease the risks of genomic sequencing. For example, providers have the opportunity to carefully explain the technical limitations of genomic sequencing so that false reassurance is minimized, as well as an opportunity to discuss the implications of results for family members who might carry the same variants.

Population screening, on the other hand, typically takes place in a public health context. Rather than utilizing a shared decision-making process to select patients for testing, public health screening is generally conceptualized as applicable to everyone, often without the direct involvement of a personal healthcare provider. In this respect, the decision to screen is driven primarily by a protocol rather than shared decision-making. In newborn screening, for example, testing is guided by state- and hospital-level protocols that direct health care facilities to screen virtually all newborns. Since this approach does not utilize individualized decision-making to maximize benefits, the focus is on (1) ensuring that screening efforts are supported by substantial evidence for benefit within the screened population and (2) providing a coordinated infrastructure, including standard operating procedures and public health workers, to maximize the chances that the actions clinicians, patients, and families take in response to test results will bring net benefit. The infrastructure created in each state to support newborn screening programs is a good example of the substantial work that is required to ensure that population screening efforts provide more benefit than harm.

Indication-Driven Clinical Testing and Opportunistic Screening

In current clinical practice, genomic sequencing technologies are used far more frequently to answer specific clinical questions than they are to screen healthy individuals for conditions they have not yet developed. Sequencing may be performed, for example, to identify the molecular cause of an undiagnosed disease.¹ When sequencing is used to address a specific clinical indication, the provider and patient must also decide whether to search for any secondary findings, i.e. those results not directly related to the primary motivation for testing, for opportunistic screening. This is not a straightforward decision. Although the clinical indication that provided the motivation for testing will increase the likelihood that genomic variants related to this condition are true positives, these patients are similar to unselected populations when it comes to their prior probability for having unrelated conditions.

Current ACMG recommendations on this question reflect medical practice in other areas of medicine in that clinicians should select appropriate patients for testing and should choose the most focused and validated testing technology that will address the clinical question. The ACMG recommendations suggest that once laboratory data are generated, the default practice should be to also analyze at least a minimal list of genes where discovery of a known mutation could provide patients and providers with information that would change clinical management.^{4, 26}

As in other diagnostic settings, there may be compelling reasons to respect a patient’s preference not to receive secondary findings. The physician and patient could decide, for example, to decline the disclosure of secondary findings data because the patient’s medical condition would prevent secondary findings from being useful (such as when a patient is critically ill). Alternatively, the patient’s psychological state or past experiences may have resulted in a preference not to learn certain types of probabilistic risk information.

Provider-ordered Preventive Testing in Healthy Adults

The question of whether to analyze for and disclose secondary results is fundamentally different from the question of whether to initiate genomic testing exclusively for the purposes of prevention. When a clinician and patient (or parents) are considering genomic testing to address a specific indication, such as identifying a diagnosis for a child with intellectual disability, the anticipated value from addressing this clinical question fulfills the imperative to minimize harm by only ordering tests that provide potential benefits that outweigh the risks. Once that decision is made, the question of whether to also analyze for secondary findings generates a different decision structure. Whole exome and whole genome sequencing create large amounts of raw sequence data, and computer algorithms must then be used to “call” variants discoverable in this raw data. Since the decision to generate the raw data needed to call variants for secondary findings has already been made, the decision is reframed as one about how to responsibly analyze that data.

In comparison, when apparently healthy individuals want to pursue genomic screening in order to improve their health, there is no primary clinical question or indication that provides a clear probability of benefit. While the risks and benefits associated with specific preventive genomic results may be relatively well-defined, the overall balance of risks and benefits associated with preventive genomic testing is not yet well understood. Given current evidence, then, it is not clear that selecting genomic screening to address the healthcare needs of a healthy individual would meet a prudent standard for minimizing harms and maximizing benefits.

What if an apparently healthy adult individual and her provider, engaging in a shared decision-making process, decide that predispositional genome sequencing is an appropriate test to address her health goals? In this context, providers should engage with their patients in a process to consider which preventive tests and interventions are most likely to provide the health benefits they are seeking while minimizing potential harms. For persons especially focused on genomic screening, providers should provide appropriate counseling on the possible risks and benefits.¹⁵ And in rare cases, when the individual’s circumstances indicate a foreseeable risk of harm, the provider should be willing to decline the request for sequencing (or recommend against it if the patient will be pursuing testing through a direct-to-consumer (DTC) genetic testing service).

Given the current state of evidence, translational research studies provide a useful opportunity for interested institutions and providers to explore the potential for genomic screening to improve preventive care in healthy persons.³⁰ Offering screening in a hybrid clinical/research setting, such as the MedSeq Project,¹⁴ provides the opportunity to engage patients in an appropriate informed consent process. Perhaps most importantly, this approach, in conjunction with trans-institutional efforts like the National Institutes of Health-funded Clinical Sequencing and Exploratory Research consortium,³¹ will provide the opportunity to develop the evidence base to establish whether this technology really does carry utility as a screening test.

Population Screening in the Public Health Context

In contrast with the clinical context, shared decision-making does not play as prominent a role in the public health context in decisions to pursue testing. In the public health context, decisions to screen are made at the population level, and are based on a careful consideration of the risks and benefits of the screening test across the population, ideally on the strength of substantial empirical evidence. Local efforts are focused on maximizing access and uptake rather than a careful consideration of individual risks and benefits.

Population screening does not always result in net benefit. In one well-known example, hundreds of Japanese children underwent unnecessary surgery when a national program designed to screen infants’ urine to identify neuroblastoma cases resulted in substantial overdiagnosis of this condition.³² Even though clinical experience supported the assumption that screening would provide benefit, and an appropriate infrastructure was developed to bolster this effort, the leaders of this effort ultimately concluded that their experience “underscores the importance of rigorous evaluation of potential benefit and harm before a screening program is adopted as public policy.”³²

At the moment, we simply do not know whether the benefits of genomic screening in the general population will outweigh its potential harms. Before we can prudently recommend genomic screening in the public health context, it will be necessary to not only develop a sophisticated infrastructure capable of evaluating and applying high standards of evidence across a broad range of genomic variants, addressing frequent false positives, and ensuring safe and appropriate responses to findings, but also to demonstrate that such an infrastructure is adequate to mitigate harm. Until that evidence exists, professional and regulatory bodies such as the ACMG and the Food and Drug Administration should continue to encourage restraint in the application of genomic screens in the public health setting, and funding agencies should continue to support research that rigorously evaluates the risks and benefits of this new technology across multiple clinical and public health contexts.

Although a detailed discussion of direct-to-consumer (DTC) genetic testing is beyond the scope of this paper, it is worth mentioning that many of the factors that raise concerns about pursuing population genomic screening in the public health setting could also be said of DTC genetic testing. Some DTC genetic testing companies offer access to a genetic counselor, and increasingly healthcare providers report being asked to provide help with interpreting these results.^{33, 34} By its nature, however, DTC genetic testing is not offered as a part of either a shared decision-making process with a healthcare provider or in the context of a public health infrastructure designed to maximize benefits and minimize harms. These factors raise significant concerns that DTC genetic testing may create risks that could otherwise have been mitigated had genomic sequencing been pursued in a clinical, or even a public health, context. However, the policy considerations for DTC genetic testing are quite different from the three potential applications of genomic technologies to preventive medicine discussed in this paper. In particular, decisions about the implementation of genomic sequencing in clinical and public health contexts fall within the scope of judgements that healthcare providers, healthcare systems, payers, and professional organizations need make about their own values and priorities. These stakeholders do not play a direct role in making decisions about whether DTC genetic testing is safe enough to be offered to consumers.

Conclusion

There can be little question that preventing an illness would be preferable to reacting to an illness once it has developed. This is a powerful narrative that has driven important work in recent years to reorient healthcare to be more proactive, and genomics is rightly playing an important role in this movement. However, not all proactive efforts are equal. Some preventive applications of genetic testing will reveal predispositions that can be used to take action and forestall devastating health effects. Others will lead patients and providers on a wild goose chase, possibly resulting in more iatrogenic harm or cost than preventive benefit. Among the thousands of results that can be generated using genome sequencing, we do not yet know which results fall into which category.

As investigators continue to study the downstream implications of clinical genome sequencing,⁹ the present uncertainty about the benefits and harms of genomic testing will be reduced, and it will become possible for providers, patients, and policymakers to make betterinformed decisions about the way these tests should be applied to preventive healthcare. For the present, however, it remains critical to remember the important differences between clinical and public health contexts and the parallel distinction between opportunistic screening on someone who has already been tested for a medical indication and population screening on an individual who has no clinical indication for testing.

Abbreviations

ACMG

American College of Medical Genetics and Genomics

DTC

Direct to Consumer