It is a long time since I have written anything on this blog about the above issue, I have an article in submission to the real scientific literature with some illustrious co-authors, but I was prompted to write this blog post by a new publication (NICUSeq Study Group, et al. Effect of Whole-Genome Sequencing on the Clinical Management of Acutely Ill Infants With Suspected Genetic Disease: A Randomized Clinical Trial. JAMA Pediatr. 2021).
One feature of the redesign of Pubmed from a couple of years ago was that the abstract view now gives a place for the conflict of interest statement, which in this case is very long. The article has 44(!) authors of whom 10 were employees of Illumina, and 4 others had some potential conflicts. The study was apparently designed by a group of people 4 of whom were Illumina employees, and the corresponding author is an Illumina employee. That really makes a mockery of the statement in the small print after the article that “Outside the sponsor employees listed as authors, the sponsor organization had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.” I have no idea what that means when company employees were intimately involved in every single stage of the trial, which was entirely funded by the company.
One might guess that this potentially leads to some bias. Over-interpretation of the value of WGS is likely in a study designed to determine whether WGS leads to clinically important changes in clinical management (COM, as they abbreviate it) if the very existence of the company funding the trial depends on clinicians believing that WGS is often beneficial.
The trial was designed in a fairly innovative way, probably as a way of trying to ensure high consent rates, they performed WGS in all of the babies and families in the trial, but told them the results at different times. Infants and families (at least one family member had to participate, so either duo, trio or more WGS was performed) were randomized to have sequencing results revealed either at 15 days after randomization or at 60 days. Infants less than 120 days of age were eligible if they had a strong clinical suspicion of a monogenic disorder. They were then followed to see if a COM occurred up to 60 days after enrolment for the primary outcome, and then up to 90 days to see what happened to the delayed reveal group.
I have a question about the ethics of this; if WGS revealed a treatable condition (for example a metabolic disorder requiring a diet change) would it be ethical to continue to hide the results for 45 days? I don’t know if there was a possibility in the protocol to do that, but I think that question does point out that it is extremely unlikely that WGS will reveal a condition that needs, and is responsive to, a specific treatment. Most such conditions will be rapidly screened out by the testing we already do in the NICU. In the ethics section of the protocol there is no mention of this as an issue.
One thing I like about this study is that they acknowledge that those other tests are going on, and therefore compare the COM in the 2 groups, but that starts to get a bit murky, what does COM mean in the delayed group? These were critically ill babies in the NICU (mostly, 7% from the PICU and 10% from cardiovascular ICU) so changes in management were happening all the time.
There is no actual calculation of sample size. They decided to enrol 300 infants and then state that this gives an 80% power to detect “a significant increase in the proportion of cases with COM on the assumption of a 20% difference in diagnostic yield between the early and delayed groups”. Which is a strange way to express it, and doesn’t state what difference in COM they are powered to detect. In the published protocol accompanying the article they state that 300 infants gives a 90% power to detect “a significant increase in proportion of cases with” COM.
I think that determining an arbitrary sample size based on resources is not unreasonable in trial design, sample size is always a compromise between the ideal and the possible, but the power should then be calculated appropriately and presented clearly.
Sample size has to be based on a clear primary outcome. I was very surprised that the protocol has no clear definition of the primary outcome. The section of the protocol on determination of test outcomes is all about the WGS results, there is no definition of what is meant by COM, and in particular what is meant by COM in the delayed group. It was apparently the site PI who decided about the presence and classification of a COM, and they were of course, not blinded, but knew what genetic results were available. In the second supplement to the publication there is a table that looks like this
This looks like a reasonable way of defining COM after genetic testing, and I will assume that this also was used to determine COM in the delayed group if they received a genetic diagnosis prior to getting the WGS results.
That being so, there were finally 354 babies randomized; why there were more than 300 is never explained. The primary outcome of the study was that, at day 60, 21% of the early WGS group had a change of management and in the delayed group 10% had a COM, which using chi-square with Yates’ correction gives a p-value of 0.011.
Because I am cynical I wonder if there are more than 300 infants in the study because after 300 it wasn’t quite significant, so they decided to enrol another 50 or so… That is probably libel to write that, and as I am against reliance on p-values in any case, I wouldn’t dream of making that statement. But to go 17% over the planned sample size is unusual.
More important than the p-values is the question of whether there was a real added value to the parents or the infant of getting early WGS results. One thing I really appreciate about this, and a few other publications about the possible value of WGS, is the individual patient data that are available (obviously without any details that could allow identification of the individual). Which means that some of the purported COM can be investigated in detail. For example subject 312, who had a diaphragmatic hernia and a secundum ASD. This patient was in the early group and is listed as having a COM type S after a non-contributory WGS. I think this is pushing it a bit, I presume the idea is that they proved the baby probably did not have a lethal genetic condition, and that was the added value of WGS. Patient 17 had a long QT interval, and the specific gene defect found was in RYR2, which apparently leads to a malignant form of recurrent ventricular tachydysrhythmias. Treatment is probably the same as other prolonged QT patients, but I can certainly go along with the evaluation that this WGS result helped the infant and their family, the standard panel for long QT apparently doesn’t pick this up.
Another infant in the trial, listed as having a COM, had a diagnosis of a hypomyelinisation myopathy, for which there is no specific therapy, did knowing the gene really help the infant? Surely the supportive care the baby needs is related to his hypotonia, feeding difficulties etc. and not to his genetic abnormality.
Another example is an infant with multiple anomalies who had a diagnosis of Cornelia de Lange syndrome after WGS, for whom the COM is noted as both M and S. However, there is no specific treatment for the syndrome, and supportive care depends on the symptomatology, not on which gene is abnormal. I’ve seen several babies with this syndrome, and not knowing which gene was abnormal did not, I think, have any impact on how we looked after them.
One thing that geneticists are all too aware of, but which seems often lacking in publications of the proposed benefits of WGS, is the problem of phenotype-genotype correlation (or the lack of correlation). It is vital to remember that exactly the same genetic defect can cause a range of phenotypic expression. In relatively common disorders, for example CHARGE there are enough cases to have publications about the issue, and there have even been cases of monozygotic twins with CHARGE who have dramatically different phenotypes. As the article I just linked to concludes, “the phenotype cannot be predicted from the genotype”.
Although I am being very critical of this article, I do think that WGS is sometimes indicated, and I do think there may sometimes be important benefits for families. Just knowing the diagnosis can be a benefit, having some idea of prognosis is important, even though it is important to always be aware of the limitations of predictions based on genotype. For genetic counselling, a diagnosis, and sometimes the identification of the gene, may be essential. And, for future possible pre-implantation diagnosis, knowing the precise genetic abnormality is essential. For some parents, however, knowing that their infant’s problems are due to genes inherited from them can cause them to feel guilty.
In a study which Annie Janvier and our genetics group at Sainte Justine did, 14% of parents, when asked about their experience with Chromosomal Micro-Array testing, expressed concern that they may be responsible for their child’s genetic condition and 12% of parents did actually feel an increase in guilt after the test. (although others, also about 10%, will feel less guilty).
The benefits of WGS testing, however, very rarely include a specific change in management. In this trial, unlike other studies, there are very few instances of COM involving palliative care or end-of-life care. Two of the cases listed as having a COM involving palliative or end-of-life care had a non-contributory WGS, so I don’t understand how that helped to make the decision. In other cases reported in other studies it seems to me that performing WGS only delayed the decision to redirect care. Often, the clinical details presented for those other cases suggest that redirection of care would have been appropriate without WGS. If parents and care teams continue intensive care in the hope that the WGS is going to produce a result which will dramatically change their infants treatment and prognosis, then that is going to be an extremely rare occurrence, and parents should know that. It may be that the most common impact of WGS testing is continuing invasive care and life support while waiting for a result, care which may be painful, stressful, and ultimately prove futile.
Informed consent for WGS testing requires that we have an honest evaluation of the potential benefits and risks. Some of the investigators who have published on the issue seem to have vested interests in proving how beneficial it is to clinical care. In my personal experience, the most common benefit of a positive WGS to the individual baby has been that we can stop searching for other diagnoses, and concentrate on their symptomatic management. The majority of WGS are non-contributory, however, and we just carry on doing what we were before. I think our group’s scepticism about the likelihood of finding a treatable diagnosis means that we have not delayed redirection of care while waiting for a WGS result.
All of our advanced diagnostic testing, from genomes to PET scans to whatever is next, needs to be accompanied by an unbiased evaluation of benefits and risks (and costs). Only then can we adequately inform parents.