Traditional newborn screening generally looks for biochemical signatures associated with a defined set of disorders. A small blood sample, usually collected from a newborn’s heel, is analysed for metabolic products or other markers that indicate a potential disease risk. These programmes are designed around a relatively clear chain of action: identify a condition early, confirm it with diagnostic testing, and begin treatment or monitoring before symptoms cause irreversible harm. Genomic screening changes that model by reading DNA directly. Instead of searching only for the molecular consequences of a disease, it can scan for variants associated with many conditions at once, including disorders that may emerge years or decades after birth.

That expanded reach creates a technical problem known as variant interpretation. Human genomes contain millions of differences, most of which are harmless or have no established medical meaning. A sequencing system may identify a change in a gene, but the presence of a variant does not automatically establish that a child will develop a disease. Scientists must assess the strength of evidence linking the variant to a condition, determine whether the change disrupts gene function, consider the child’s biological context, and estimate how likely disease is to occur. Even when a variant is classified as pathogenic, penetrance—the proportion of people carrying it who actually develop the associated condition—may be incomplete. A positive genomic signal can therefore represent a confirmed risk, a possible risk, or a finding whose meaning is not yet known.

The paper’s title points to a distinction with major consequences: a case may be “missed,” or a signal may simply go “unreported.” These are not necessarily the same event. A missed case suggests that a condition was present but escaped detection, perhaps because the relevant gene was not included, the variant was difficult to identify, or the disease did not fit the screening algorithm. An unreported signal, by contrast, may have been detected in the data but excluded from communication because its interpretation was uncertain, its clinical value was judged insufficient, or the programme lacked a defined pathway for follow-up. The difference reveals that screening is not only a process of finding biological facts. It is also a process of selecting which facts become medically visible.

This selection begins long before a sequencing machine produces a result. Screening programmes must decide which genes and conditions to include, what level of disease risk justifies reporting, how severe or treatable a condition must be, and whether findings relevant only in adulthood should be returned to parents. They must also establish technical thresholds for calling variants, rules for confirming results, and procedures for handling findings that may affect relatives. Every threshold changes the apparent performance of the programme. A system designed to minimise false positives may fail to report signals that later prove important, while a system designed to maximise detection may identify many findings that do not lead to disease, creating anxiety and additional testing.

For families, these categories can be difficult to understand because the language of genomics often compresses uncertainty into simple labels. “Positive” may be interpreted as a diagnosis even when it means only that a genetic risk requires further assessment. “Negative” may be understood as a guarantee of health, although no screening test can exclude every disorder. Sequencing may overlook structural changes in DNA, repetitive regions, low-level mosaicism, or variants that current databases do not recognise. It may also identify a disease risk that has no immediate treatment. The technical limitations of a test therefore become communication challenges, and the way results are named can influence medical decisions, parental expectations, and a child’s future identity.

Genomic newborn screening also changes the meaning of evidence. Conventional programmes are often evaluated using measures such as sensitivity, specificity, positive predictive value, and the number of affected babies identified before symptoms appear. These measures remain essential, but genomic screening introduces additional questions. How should a programme assess a variant whose disease association is supported by limited evidence? How should it measure the value of detecting a condition that may never develop? What happens when a result prompts years of surveillance but no intervention? And how should researchers account for people who carry clinically important variants but are never reported because the programme’s framework excludes them? According to the article’s central concern, the architecture of reporting can shape the dataset used to judge whether screening works.

The problem is intensified by unequal representation in genomic databases. Variant interpretation depends partly on reference data gathered from people whose ancestry, health history, and clinical outcomes are known. If some populations are underrepresented, a variant may be incorrectly labelled uncertain, or a harmless population-specific change may be mistaken for a disease-causing one. Conversely, a genuinely harmful variant may be difficult to recognise because there are too few comparable cases. A screening system can therefore produce different levels of clarity for different families, not because their genomes are inherently more or less interpretable, but because scientific knowledge has been built unevenly. The question of what counts is consequently connected to questions of whose data are available and whose risks are made legible.

The authors’ analysis arrives as health services worldwide consider whether genomic newborn screening should complement or replace selected forms of biochemical testing. Sequencing is becoming faster and more affordable, and computational tools can compare a newborn’s DNA with expanding libraries of disease-associated variants. Yet technology alone cannot determine the boundaries of a responsible programme. Those boundaries require decisions about consent, privacy, data storage, reanalysis, return of results, clinical capacity, and long-term follow-up. They also require transparency about what is not being searched for, what cannot be interpreted, and what may be discovered later as scientific knowledge changes. A result that is uncertain today could become clinically meaningful tomorrow, while a result reported with confidence may later be reclassified.

The broader message is that genomic newborn screening should not be judged only by its ability to generate more findings. Its success may depend on whether it can distinguish between detection and diagnosis, between absence of evidence and evidence of absence, and between a signal that was never seen and one that was seen but withheld from the report. By focusing attention on the “architecture of what counts,” Carley and Lucassen place the debate beyond the familiar promise of catching disease earlier. The future of newborn screening will be shaped not only by the sensitivity of sequencing platforms, but also by the rules, values, and institutions that decide which genetic information becomes part of a child’s medical story.

Subject of Research: Genomic newborn screening and how genetic findings are classified, interpreted, and reported.

Article Title: Missed cases or unreported signals? Genomic newborn screening and the architecture of what counts.

Article References: Carley, H.M., Lucassen, A.M. “Missed cases or unreported signals? Genomic newborn screening and the architecture of what counts.” European Journal of Human Genetics (2026). https://doi.org/10.1038/s41431-026-02218-3

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41431-026-02218-3

Keywords: genomic newborn screening, genetic testing, variant interpretation, newborn screening, genomics, clinical genetics, public health, genetic uncertainty, precision medicine