During the focus group discussion, participants described their current and desired practice, whilst identifying issues regarding reinterpretation and reclassification. Further challenges came to light through discrepant practices described by participants. Based on these accounts, the authors identified the following five challenges.
1. Should active reinterpretation of variants be conducted by the laboratory as a routine practice?
Participants indicated that in current Dutch clinical genetics laboratory practice, reinterpretation mainly takes place reactively, i.e. in response to an event [Table 2]. There is no routine practice in Dutch laboratories of actively reviewing the latest available evidence on variants. In terms of possible, future implementation of active, periodic reinterpretation, a distinction can be made between what is desirable and feasible laboratory practice.
Desirability
Active, periodic reinterpretation of variants could be an important strategy for ensuring patients are alerted to changing evidence surrounding genetic variants. Participants indicated that rare variants in particular are currently less likely to be reinterpreted, as the chance that they are detected in other patients is relatively small. One laboratory was investigating ways to implement active, periodic reinterpretation of all VUS in their in-house database. However, active, periodic reinterpretation was not unequivocally considered a desirable practice by participants; they expressed concerns whether such reinterpretation could be adequately understood and consented to by patients during the process of genetic counseling (see section 2). Active reinterpretation of patient data was not seen by participants as a responsibility of the laboratory, as the laboratory’s aim, similar to other medical fields, was seen as providing service at request, rather than actively initiating services. Therefore the main responsibility to initiate reinterpretation of genetic data was considered to lie with patients (in tandem with their clinicians).
“We are service-oriented, in other words: at the request of the patient.” (Participant 11)
“In my opinion it’s comparable to a doctor: they don’t re-evaluate their pictures and files every year either to see if there is any new information. (Participant 3)
Feasibility
In terms of feasibility, the main factor impeding active, periodic reinterpretation according to participants was work load. Such a procedure was currently deemed to be financially and logistically unachievable, as the process of reinterpretation was seen as laborious and unamenable to complete automation. Participants noted that since different conditions require different kinds of evidence to classify variants as either benign or pathogenic, combined with the fact that laboratory geneticists differed in what they considered as adequate evidence for (re)classification, automating reinterpretation and reclassification was considered challenging.
“As mentioned by many [others], automation is hard, because the correct application of data present in databases and literature is almost impossible to interpret in silico. For this, you’ll always need humans.” (Participant 2)
“Active [reinterpretation] of data is currently not feasible in terms of work [load]. (…) Possibly, if [these processes] are more automated in the future, there will be more possibilities to do this (and to want [to do this]).” (Participant 3)
Several guidelines indicate that laboratories do not have any responsibility to routinely reinterpret genetic data, so as not to be overloaded or compromised in other duties related to patient care [Bombard et al. 2019; Matthijs et al. 2016]. However, we propose that active, periodic reinterpretation does not necessarily need to be conceptualized in an all-or-nothing way: a middle ground may also be feasible. Actively reclassifying only those variants for which clear, new evidence is present regarding a pathogenic (or benign) effect could benefit some patients, without leading to extensive database and literature searches. In terms of bioinformatics, this would mean that the threshold to flag a variant for possible reclassification is set high. Commercial companies exist that provide active, periodic reclassification [Pathway Genomics]. Tools such as InterVar (Li and Wang, 2017), which use a formalized protocol, could possibly aid automation of the classification process. Furthermore, an infrastructure in which classifications and variant frequencies are automatically shared between laboratories [VKGL, n.d.], as well as an infrastructure to alert clinicians or laboratory geneticists to updated classifications, as discussed by Aronson et al. (2012), could aid management of reclassified variants. The field of artificial intelligence could offer interesting opportunities in this context. In terms of comparing genetics services to other medical services which tend to offer services at request, rather than actively initiating them, it is important to note that genetic data differs from most other kinds of medical data, as it is constant throughout a patient’s life; only its analysis and interpretation changes. As the laboratory knows more about current variant classifications than clinicians, let alone patients [David et al. 2019], it is important to explore to what degree laboratories could and should have systems in place for systematic reinterpretation. Many variants, especially relating to patients with rare diseases and patients from ancestries underrepresented in the genomic evidence base [Bombard & Mighton 2018], are found infrequently on a global level. Importantly, not conducting active, periodic reinterpretation might result in an inequitable service where some patients more than others are left unaware of new evidence, thereby posing an injustice issue.
2. How does reinterpretation initiated by the laboratory relate to patient expectations and consent?
Participants indicated that, in current Dutch laboratory practice, variant reinterpretation is initiated by laboratory geneticists during the analysis of new patients, without updated consent from previous patients with the same variants [Table 2, case b]. Concerns existed among laboratory geneticists regarding the autonomy of these patients, as well as the durability of consent. Even if the possibility of reinterpretation was discussed during initial counseling, participants felt unclear about patients’ ability to fathom that consent regarding reinterpretation of their genetic information is given for an undetermined period of time. This may result in unexpected and possibly, in that moment, unwanted recontact.
"When the same VUS is detected in a new patient, it will routinely be evaluated again. The question whether patients desire [reinterpretation] does not play a role in my decision making. […] [Because] the patient is informed about this possibility and also gave consent. However, I do wonder if they realize that they can be contacted two years later.” (Participant 9).
Concerns particularly existed regarding patient expectations, understanding, and wishes in discussions regarding active reinterpretation.
“Very rarely you have to send a corrected letter to the first patient, who thinks his/her [DNA] examination has been concluded and often is not looking for [new information]. Therefore, you should not be doing this routinely for each detected variant.” (Participant 6)
“It seems to me that it’s of main importance that the patient wants [active, routine reinterpretation]!” (Participant 9)
We pose that for laboratory geneticists, an important conceptual difference between active and reactive reinterpretation might be the generation of new information (active reinterpretation) versus reporting of information that has already been generated (reactive reinterpretation and reclassification following identification of a previously detected variant in a new patient). They may feel a moral responsibility to act on new information, after it has been generated. This may explain why more hesitance was expressed regarding respecting patient autonomy in the context of active reinterpretation in comparison with reactive reinterpretation. However, from the perspective of the patient, this conceptual difference may not exist. For the patient, being recontacted based on reactive reinterpretation initiated by the laboratory does not differ from being recontacted based on active reinterpretation initiated by the laboratory. Therefore, implementing active, periodic reinterpretation does not diverge from current practice, in terms of the possible negative consequence of unexpected recontact from a patient perspective, except that it would likely occur on a larger scale.
It is important not to alleviate the moral distress around patient autonomy as expressed by laboratory geneticists by automatically precluding active, periodic reinterpretation. Rather, these concerns should be addressed through empirical inquiry into what general patient expectations and understanding are regarding reinterpretation of their genetic data and through investigation of proper means of counseling and consent suitable for the genomic era. This could include dynamic consent, allowing patients to halt or continue active reinterpretation of their data at their preference; a larger emphasis during pre- and post-test counseling on the possibility of change in genetic knowledge and its implications for (family) medical management; and the exploration of technological infrastructures that allow the possibility, if desired by patients, of not reinterpreting or reclassifying their data. Particularly, an important question is whether it is good practice to give patients the opportunity to broadly opt out of potentially medically actionable information, as this is a complex decision regarding information that cannot be envisaged accurately by patients at the time. A potential additional problem with this approach is that different family members carrying the same variant(s) might then give different kinds of consent, leading to new reports being issued for some family members in case of variant reclassification, but not for others. This may give rise to difficult and tense family situations, as well as different medical treatment and screening protocols within one family, begging the question whether patient choice in this context is feasible and desirable in practice.
3. When should reinterpreted data be considered clinically significant and communicated from laboratory to clinician?
When patients and their clinicians ask a laboratory for reinterpretation of a previously reported variant, the laboratory communicates any new information or lack thereof to the clinician of the patient in question, participants pointed out.
In contrast, participants indicated that when a variant is reclassified without prior inquiry by a patient or their clinician (e.g. when the variant is found in a new patient and reclassified for all individuals carrying this variant), communication of this new information from laboratory to clinician does not occur by default. Whether a reclassification potentially affects a patient’s (or their family’s) clinical management, i.e. its clinical significance, is the underlying principle of Dutch laboratories’ practices and decisions related to recontacting clinicians. Our focus group discussion showed that laboratory geneticists feel the clinician is responsible for establishing what is clinically significant. In practice, however, laboratory geneticists also make decisions on this matter by deciding which reclassifications are of clinical significance and therefore need to be communicated to clinicians. Despite absence of formal policy, participants indicated that it is standard practice to communicate the following reclassifications to clinicians, in case reinterpretation took place without request by patient or clinician: any variant that is upgraded to (likely) pathogenic (e.g. likely benign to likely pathogenic), as well as (likely) pathogenic variants that are downgraded (e.g. pathogenic to VUS) [Table 3]. An exception constitutes pathogenic variants that are downgraded to likely pathogenic. Often, these reclassifications are not reported to clinicians, as they are not expected to affect clinical management. However, practice differs most significantly for VUS that are downgraded to (likely) benign or vice versa: some laboratory geneticists do, whilst others do not communicate these reclassifications to clinicians, as laboratory geneticists differ in whether they consider these reclassifications as clinically significant.
Some laboratory geneticists argued that when a VUS is reclassified to a benign or likely benign variant, this does not merit recontact with the referring clinician, as it would not affect patients’ medical management. This is in line with ACMG recommendations that VUS should not be used in clinical decision-making (Richards et al. 2015). However, one participant observed that clinicians often communicate VUS to patients as possibly causative of their clinical symptoms, when they are detected and reported in the original laboratory report.
“In case of a reclassification from (…) [class] 3 to 2/1, in principle no [new lab report] will follow, because it doesn’t change [clinical] management.” (Participant 8)
“I do report a reclassification from [class] 3 to 1/2 [to the referring clinician] in case I come across it. It really depends on the condition. I can imagine that for onco[genetics], for example, it won’t change [clinical] management, but for many other conditions a class 3 [variant] is often counseled as possibly causative. If you then know for certain that it’s not pathogenic, I think you can report it.” (Participant 7)
Vos et al. (2012) and Solomon et al. (2016), among others, have documented the degree to which detected VUS that are communicated to patients as potentially causative of their phenotype can have a significant emotional impact on patients and their families. Furthermore, in some cases, a VUS might be tracked within families to see whether it segregates with the clinical phenotype. Therefore, we suggest that VUS cannot be seen as entities outside of clinical decision-making. In some cases, such as when VUS have previously been communicated to patients as highly suspicious in causing their symptoms, it may be important for the laboratory to report down-classifications of VUS to clinicians. As such, we propose that down-classifications of VUS should not be excluded by default, in contrast to what Chisholm et al. [2017] suggest with their proposed workflow for reinterpretation of variants.
4. Should reinterpretation, reanalysis or a new test be conducted?
In certain cases, participants believed that reanalysis may prove to be diagnostically more effective than reinterpretation. Reanalysis involves using a patient’s existing raw data (that has been generated as a result of genetic sequencing in the past, e.g. whole exome sequencing) in order to analyze all genes currently associated with the patient’s condition or symptoms, without having to conduct a new genetic test. This includes genes that were not analyzed previously, as a connection with the patient’s condition was not known at the time.
In light of fast-changing techniques, improved coverage, decreasing costs, and growing knowledge, it might even be more efficient to request a completely new test (i.e. resequencing of DNA), based on the latest laboratory and analysis standards, rather than performing reinterpretation or reanalysis of existing data.
"I believe that the patient (in consultation with the clinician) can actively request [reinterpretation], and, in my opinion, in the future, this could [change to] performing a completely new test, instead of reconsidering old data.” (Participant 11)
Therefore, we recommend that the discussion on reinterpretation in case no genetic diagnosis was made includes the comparison in (cost) efficiency between reinterpretation of previously detected variants, reanalysis of existing data, and redoing genetic tests (i.e. resequencing) based on the latest technologies. To better determine which genes are of interest for reanalysis or a new test, laboratories need to have optimal, recently updated information regarding patient and family phenotypes. For this, continuous feedback is needed from clinicians and patients. Currently, delivery of phenotypic information to the laboratory is considered too brief and unsystematic to reach this goal. Presently, no digital systems are in place to aid in the continuous updating of phenotypic data, before, during, or after the (initial) test. Electronic patient records could help update patient and family phenotypes in a standardized manner, for instance using Human Phenotype Ontology (HPO) or SNOMED CT terminology. Personal health records, which enable patients to digitally manage and update their own medical and family data, could also serve as an important tool in this context (Baldwin 2017).
5. How are reclassifications perceived and how might this affect laboratory practice?
A final topic raised during the focus group discussion that warrants further investigation in future research is how reclassifications are perceived by laboratory geneticists. For example, are they viewed as the correction of an error that was made in the initial classification or merely as inevitable progression of knowledge? The way reclassifications are perceived may shape the actions taken by laboratory geneticists and their (legal) responsibilities regarding reinterpretation and reclassification of variants. For instance, the degree to which laboratory geneticists need to keep up with evolving evidence on variants within scientific literature.
“If the classification is downgraded, then immediate contact (…) must be sought with the referring clinician, because it possibly concerns a calamity, comparable to entirely missing a mutation. (…) I [mean] downgrading of [a class] 5 [variant to a class] 3 [variant]. Of course, [downgrading of a class] 3 [variant to a class] 1 [variant] is not a calamity but rather an insight. […] When you miss a mutation, you have made an error, and [this is] also [the case] when you classify [a variant] too high.” (Participant 8)
A United States court case (Williams vs Quest/Athena, currently in progress, illustrates this issue. Here, a mother (Williams) claims that a genetic misclassification led doctors to administer a pharmaceutical treatment that worsened her son’s condition and caused his death (aged 2, due to an epileptic seizure). A genetic test had been performed in 2007 which had identified a variant in SCN1A, agene implicated in Dravet syndrome. The variant was classified by the laboratory as a VUS and therefore did not affect clinical-decision making by the boy’s clinicians. Two research papers, available publicly at the time of the laboratory report, already stated that this presumed VUS was probably a pathogenic variant [Thorogood et al., 2017]. In 2015, the laboratory issued a new report with the variant reclassified as known to be disease causing. The prosecution argues that the initial report contained a mistaken classification and that had this variant been “correctly” classified in 2007, thereby pointing to the diagnosis of Dravet Syndrome, different medication would have been given to the child and the fatal seizure would not have occurred [Genomeweb, 2018]. This case raises important questions in the context of variant reclassification. Are reclassifications a correction of a previously made mistake (i.e. the initial classification) in some cases and should laboratory geneticists be held liable for making it, if the knowledge suggesting an alternate classification was available at first interpretation or even afterwards?
A factor complicating the question of error, such as in the Williams vs Quest/Athena case, is that there is no golden standard regarding what is considered to be enough evidence for an initial classification or for a reclassification. It is not simply a question of whether the evidence for making a certain classification was there and whether it was adequately taken into account. As discussed earlier (section 1), the amount of evidence needed differs for different conditions and variant types, and what is considered sufficient evidence for a certain variant classification differs between laboratory geneticists. As such, it may be difficult to establish at what point the act of making an initial classification should stray into the territory of error.
Yet, some reclassifications may more likely be considered by laboratory geneticists as the correction of an error in initial classification than others. In the public database ClinVar, variants that used to have a seemingly clear pathogenic or benign classification are currently reclassified or conflicting interpretations are mentioned [Mighton et al. 2019; Shah et al. 2018]. This means that in rare cases it is possible that during reinterpretation, pathogenic variants are reclassified to for instance VUS. When a result is initially reported to be pathogenic, reclassification means that confidence regarding its pathogenicity had been unwarranted. Whereas when a result is reported as a VUS, it is implicit that there is currently not sufficient evidence available, but that additional evidence may be obtained in the future, thereby inherently subjecting this variant to possible future reclassification.
We suggest that, rather than considering pathogenic down-classifications or benign up-classifications as constituting an event that was not supposed to happen (i.e. an error), perhaps an increase in awareness is needed that classifications are not necessarily as fixed as they might seem. Currently, this awareness mainly exists for VUS: VUS are expected to change at some point in the future. However, almost all variants are continuously subject to reinterpretation and reclassification. When considering trends on reclassification, it has been shown that progression of information on variants often leads to a shift towards conflicting interpretations of pathogenicity, rather than to more clear and correct classifications [Shah et al. 2018]. This emphasizes the argument that both initial and reclassifications should be inherently considered as subject to possible change. This does not imply, however, that clinical decisions such as initiating cascade family screening on presumed pathogenic variants should be abolished, based on the argument that one cannot be fully certain that these variants are as pathogenic as we currently think they are. Rather, both healthcare professionals and patients need to be cognizant that evolving knowledge is intrinsic to genetics and that, therefore, changes in a patient’s diagnosis or genetic risk profile are a possibility with increasing knowledge.