By Gabriel Watts and Ainsley J. Newson.

Data obtained from genomic sequencing has an interesting quality. Unlike most other kinds of health results, the stored information remains accurate over time, because it reflects a largely stable property of our bodies: our DNA.

Of course, during this time, sequencing methods themselves are likely to have advanced further such that new sequence data will be of better quality. Much in the same way that the resolution on a phone camera picture from 2013 is not as good as one from today – indeed, we are already seeing changes to high-throughput DNA sequencing quality with the advent of long-read sequencing. But still, that present day sequence data can retain diagnostic validity for as long as a decade is already exceptional. In our paper in the Journal of Medical Ethics, we refer to this property as the ‘diagnostic durability’ of genomic data.

Another important aspect of genomic information is that while the sequencing data itself is stable, the interpretation of that data can be quite dynamic, and may change within a short space of time. This means a result delivered to a patient at one point in time may have a different interpretation later on.

For patients who receive results such as ‘no pathogenic (disease-causing) variant identified’ or who are told they have a ‘variant of uncertain significance (VUS)’, the changing status of this result can be significant. If a finding is re-graded, it may open up new treatment options that they couldn’t previously access. A result can also go the other way, to benign from VUS.

These attributes of genomic data have important implications for the responsible implementation of genomic testing in health. One question is: should laboratories or clinicians routinely go back to the data they hold for patients with null or VUS results, to see if a new interpretation is possible? We consider this question in our paper.

An immediate issue here is whether routine reanalysis is even feasible. On the one hand, doing this is known to increase the ‘diagnostic yield’ of genomic testing: more patients receive definitive information that can inform their treatment. Yet on the other hand, until automation of reanalysis is in place (and this is coming) this process is time- and resource-intensive, and likely beyond the majority of health systems to provide at large scale.

One way around this limitation is to only provide reanalysis to those who ask for it. But this is likely to limit this benefit to those who know to do it, or to ask for it, and so raises equity concerns.

One part of reanalysis, however, is reinterpretation of variant classifications. This process can achieve increased diagnostic yields in a comparable way to a detailed individual reanalysis. But it is more sustainable because it occurs at the level of classes of variant rather than individual patient DNA sequences. As such, routine reinterpretation of variant classifications may be more feasible at scale, at least in the short to medium term.

Given this, do laboratories or clinicians have an obligation to undertake routine reinterpretation of variant classifications as a part of the responsible implementation of genomic health care?

In our paper we argue against the existence of any general duty to reinterpret genomic variant classifications. Yet, we contend that a restricted duty to reinterpret ought to be recognised.

Our initial motivation was drawn from the intuitive pull of the position we argue against. It is undoubtedly ideal that diagnostic laboratories routinely reinterpret all their variant classifications, in order to keep up with the rapid changes in our understanding of genomic testing results.

It is a different question, however, whether there is a moral duty to do so. At issue here is whether the potential benefits of routinely reinterpreting genomic variant classifications is likely to lead to a valid diagnosis for any particular patient. If not, then it is arguably better to invest resources in preparing patients for the high likelihood that genomic testing will produce results that are uncertain, and that are statistically unlikely to become clinically relevant in the future, than to hold out of hope of a statistically unlikely diagnosis through regular variant reinterpretation.

To be clear, we are not arguing that we should not aim to develop diagnostic systems on which all genomic variant classifications are routinely reinterpreted. For instance, developing diagnostic systems that automate the various elements of reanalysis – including reinterpretation of variants classifications, but also the re-prioritisation of previously unanalysed sections of a patient’s genome, as well as the re-annotation of sequence data – is a morally laudable aim.

What we do argue is that the best healthcare systems need to be developed within the limits of what is currently or imminently feasible. As automation expands, the obligation to reanalyse may become actual. Our point is that we best not confuse this with an obligation arising from certain peculiar properties of genomic sequencing data. For any obligations here only stretch so far, and better warrant investment in patient counselling concerning the inherent uncertainty of genomic testing than investment in the routine reinterpretation of all variant classifications.

Paper title: Is there a duty to routinely reinterpret genomic variant classifications?

Authors: Gabriel Watts, Ainsley J Newson

Affiliations: Sydney Health Ethics, University of Sydney – Australian Genomics

Competing interests: None declared